_id
stringlengths 2
7
| title
stringlengths 1
88
| partition
stringclasses 3
values | text
stringlengths 75
19.8k
| language
stringclasses 1
value | meta_information
dict |
|---|---|---|---|---|---|
q7000
|
Logger.subscribe
|
train
|
def subscribe(self, queue=None, *levels):
"""
Subscribe to the aggregated log stream. On subscribe a ledis queue will be fed with all running processes
logs. Always use the returned queue name from this method, even if u specified the queue name to use
Note: it is legal to subscribe to the same queue, but would be a bad logic if two processes are trying to
read from the same queue.
:param queue: Your unique queue name, otherwise, a one will get generated for your
:param levels:
:return: queue name to pull from
"""
args = {
'queue': queue,
'levels': list(levels),
}
self._subscribe_chk.check(args)
return self._client.json('logger.subscribe', args)
|
python
|
{
"resource": ""
}
|
q7001
|
AggregatorManager.query
|
train
|
def query(self, key=None, **tags):
"""
Query zero-os aggregator for current state object of monitored metrics.
Note: ID is returned as part of the key (if set) to avoid conflict with similar metrics that
has same key. For example, a cpu core nr can be the id associated with 'machine.CPU.percent'
so we can return all values for all the core numbers in the same dict.
U can filter on the ID as a tag
:example:
self.query(key=key, id=value)
:param key: metric key (ex: machine.memory.ram.available)
:param tags: optional tags filter
:return: dict of {
'key[/id]': state object
}
"""
args = {
'key': key,
'tags': tags,
}
self._query_chk.check(args)
return self._client.json('aggregator.query', args)
|
python
|
{
"resource": ""
}
|
q7002
|
CGroupManager.ensure
|
train
|
def ensure(self, subsystem, name):
"""
Creates a cgroup if it doesn't exist under the specified subsystem
and the given name
:param subsystem: the cgroup subsystem (currently support 'memory', and 'cpuset')
:param name: name of the cgroup to delete
"""
args = {
'subsystem': subsystem,
'name': name,
}
self._cgroup_chk.check(args)
return self._client.json('cgroup.ensure', args)
|
python
|
{
"resource": ""
}
|
q7003
|
dlmk
|
train
|
def dlmk(l,m,k,theta1):
"""
returns value of d^l_mk as defined in allen, ottewill 97.
Called by Dlmk
"""
if m >= k:
factor = sqrt(factorial(l-k)*factorial(l+m)/factorial(l+k)/factorial(l-m))
part2 = (cos(theta1/2))**(2*l+k-m)*(-sin(theta1/2))**(m-k)/factorial(m-k)
part3 = sp.hyp2f1(m-l,-k-l,m-k+1,-(tan(theta1/2))**2)
return factor*part2*part3
else:
return (-1)**(m-k) * dlmk(l,k,m,theta1)
|
python
|
{
"resource": ""
}
|
q7004
|
Dlmk
|
train
|
def Dlmk(l,m,k,phi1,phi2,theta1,theta2):
"""
returns value of D^l_mk as defined in allen, ottewill 97.
"""
return exp(complex(0.,-m*phi1)) * dlmk(l,m,k,theta1) * \
exp(complex(0.,-k*gamma(phi1,phi2,theta1,theta2)))
|
python
|
{
"resource": ""
}
|
q7005
|
gamma
|
train
|
def gamma(phi1,phi2,theta1,theta2):
"""
calculate third rotation angle
inputs are angles from 2 pulsars
returns the angle.
"""
if phi1 == phi2 and theta1 == theta2:
gamma = 0
else:
gamma = atan( sin(theta2)*sin(phi2-phi1) / \
(cos(theta1)*sin(theta2)*cos(phi1-phi2) - \
sin(theta1)*cos(theta2)) )
dummy_arg = (cos(gamma)*cos(theta1)*sin(theta2)*cos(phi1-phi2) + \
sin(gamma)*sin(theta2)*sin(phi2-phi1) - \
cos(gamma)*sin(theta1)*cos(theta2))
if dummy_arg >= 0:
return gamma
else:
return pi + gamma
|
python
|
{
"resource": ""
}
|
q7006
|
rotated_Gamma_ml
|
train
|
def rotated_Gamma_ml(m,l,phi1,phi2,theta1,theta2,gamma_ml):
"""
This function takes any gamma in the computational frame and rotates it to the
cosmic frame.
"""
rotated_gamma = 0
for ii in range(2*l+1):
rotated_gamma += Dlmk(l,m,ii-l,phi1,phi2,theta1,theta2).conjugate()*gamma_ml[ii]
return rotated_gamma
|
python
|
{
"resource": ""
}
|
q7007
|
real_rotated_Gammas
|
train
|
def real_rotated_Gammas(m,l,phi1,phi2,theta1,theta2,gamma_ml):
"""
This function returns the real-valued form of the Overlap Reduction Functions,
see Eqs 47 in Mingarelli et al, 2013.
"""
if m>0:
ans=(1./sqrt(2))*(rotated_Gamma_ml(m,l,phi1,phi2,theta1,theta2,gamma_ml) + \
(-1)**m*rotated_Gamma_ml(-m,l,phi1,phi2,theta1,theta2,gamma_ml))
return ans.real
if m==0:
return rotated_Gamma_ml(0,l,phi1,phi2,theta1,theta2,gamma_ml).real
if m<0:
ans=(1./sqrt(2)/complex(0.,1))*(rotated_Gamma_ml(-m,l,phi1,phi2,theta1,theta2,gamma_ml) - \
(-1)**m*rotated_Gamma_ml(m,l,phi1,phi2,theta1,theta2,gamma_ml))
return ans.real
|
python
|
{
"resource": ""
}
|
q7008
|
chisq
|
train
|
def chisq(psr,formbats=False):
"""Return the total chisq for the current timing solution,
removing noise-averaged mean residual, and ignoring deleted points."""
if formbats:
psr.formbats()
res, err = psr.residuals(removemean=False)[psr.deleted == 0], psr.toaerrs[psr.deleted == 0]
res -= numpy.sum(res/err**2) / numpy.sum(1/err**2)
return numpy.sum(res * res / (1e-12 * err * err))
|
python
|
{
"resource": ""
}
|
q7009
|
dchisq
|
train
|
def dchisq(psr,formbats=False,renormalize=True):
"""Return gradient of total chisq for the current timing solution,
after removing noise-averaged mean residual, and ignoring deleted points."""
if formbats:
psr.formbats()
res, err = psr.residuals(removemean=False)[psr.deleted == 0], psr.toaerrs[psr.deleted == 0]
res -= numpy.sum(res/err**2) / numpy.sum(1/err**2)
# bats already updated by residuals(); skip constant-phase column
M = psr.designmatrix(updatebats=False,fixunits=True,fixsigns=True)[psr.deleted==0,1:]
# renormalize design-matrix columns
if renormalize:
norm = numpy.sqrt(numpy.sum(M**2,axis=0))
M /= norm
else:
norm = 1.0
# compute chisq derivative, de-renormalize
dr = -2 * numpy.dot(M.T,res / (1e-12 * err**2)) * norm
return dr
|
python
|
{
"resource": ""
}
|
q7010
|
create_fourier_design_matrix
|
train
|
def create_fourier_design_matrix(t, nmodes, freq=False, Tspan=None,
logf=False, fmin=None, fmax=None):
"""
Construct fourier design matrix from eq 11 of Lentati et al, 2013
:param t: vector of time series in seconds
:param nmodes: number of fourier coefficients to use
:param freq: option to output frequencies
:param Tspan: option to some other Tspan
:param logf: use log frequency spacing
:param fmin: lower sampling frequency
:param fmax: upper sampling frequency
:return: F: fourier design matrix
:return: f: Sampling frequencies (if freq=True)
"""
N = len(t)
F = np.zeros((N, 2 * nmodes))
if Tspan is not None:
T = Tspan
else:
T = t.max() - t.min()
# define sampling frequencies
if fmin is not None and fmax is not None:
f = np.linspace(fmin, fmax, nmodes)
else:
f = np.linspace(1 / T, nmodes / T, nmodes)
if logf:
f = np.logspace(np.log10(1 / T), np.log10(nmodes / T), nmodes)
Ffreqs = np.zeros(2 * nmodes)
Ffreqs[0::2] = f
Ffreqs[1::2] = f
F[:,::2] = np.sin(2*np.pi*t[:,None]*f[None,:])
F[:,1::2] = np.cos(2*np.pi*t[:,None]*f[None,:])
if freq:
return F, Ffreqs
else:
return F
|
python
|
{
"resource": ""
}
|
q7011
|
powerlaw
|
train
|
def powerlaw(f, log10_A=-16, gamma=5):
"""Power-law PSD.
:param f: Sampling frequencies
:param log10_A: log10 of red noise Amplitude [GW units]
:param gamma: Spectral index of red noise process
"""
fyr = 1 / 3.16e7
return (10**log10_A)**2 / 12.0 / np.pi**2 * fyr**(gamma-3) * f**(-gamma)
|
python
|
{
"resource": ""
}
|
q7012
|
add_gwb
|
train
|
def add_gwb(psr, dist=1, ngw=1000, seed=None, flow=1e-8, fhigh=1e-5,
gwAmp=1e-20, alpha=-0.66, logspacing=True):
"""Add a stochastic background from inspiraling binaries, using the tempo2
code that underlies the GWbkgrd plugin.
Here 'dist' is the pulsar distance [in kpc]; 'ngw' is the number of binaries,
'seed' (a negative integer) reseeds the GWbkgrd pseudorandom-number-generator,
'flow' and 'fhigh' [Hz] determine the background band, 'gwAmp' and 'alpha'
determine its amplitude and exponent, and setting 'logspacing' to False
will use linear spacing for the individual sources.
It is also possible to create a background object with
gwb = GWB(ngw,seed,flow,fhigh,gwAmp,alpha,logspacing)
then call the method gwb.add_gwb(pulsar[i],dist) repeatedly to get a
consistent background for multiple pulsars.
Returns the GWB object
"""
gwb = GWB(ngw,seed,flow,fhigh,gwAmp,alpha,logspacing)
gwb.add_gwb(psr,dist)
return gwb
|
python
|
{
"resource": ""
}
|
q7013
|
add_dipole_gwb
|
train
|
def add_dipole_gwb(psr, dist=1, ngw=1000, seed=None, flow=1e-8,
fhigh=1e-5, gwAmp=1e-20, alpha=-0.66,
logspacing=True, dipoleamps=None,
dipoledir=None, dipolemag=None):
"""Add a stochastic background from inspiraling binaries distributed
according to a pure dipole distribution, using the tempo2
code that underlies the GWdipolebkgrd plugin.
The basic use is identical to that of 'add_gwb':
Here 'dist' is the pulsar distance [in kpc]; 'ngw' is the number of binaries,
'seed' (a negative integer) reseeds the GWbkgrd pseudorandom-number-generator,
'flow' and 'fhigh' [Hz] determine the background band, 'gwAmp' and 'alpha'
determine its amplitude and exponent, and setting 'logspacing' to False
will use linear spacing for the individual sources.
Additionally, the dipole component can be specified by using one of two
methods:
1) Specify the dipole direction as three dipole amplitudes, in the vector
dipoleamps
2) Specify the direction of the dipole as a magnitude dipolemag, and a vector
dipoledir=[dipolephi, dipoletheta]
It is also possible to create a background object with
gwb = GWB(ngw,seed,flow,fhigh,gwAmp,alpha,logspacing)
then call the method gwb.add_gwb(pulsar[i],dist) repeatedly to get a
consistent background for multiple pulsars.
Returns the GWB object
"""
gwb = GWB(ngw, seed, flow, fhigh, gwAmp, alpha, logspacing,
dipoleamps, dipoledir, dipolemag)
gwb.add_gwb(psr,dist)
return gwb
|
python
|
{
"resource": ""
}
|
q7014
|
add_efac
|
train
|
def add_efac(psr, efac=1.0, flagid=None, flags=None, seed=None):
"""Add nominal TOA errors, multiplied by `efac` factor.
Optionally take a pseudorandom-number-generator seed."""
if seed is not None:
N.random.seed(seed)
# default efacvec
efacvec = N.ones(psr.nobs)
# check that efac is scalar if flags is None
if flags is None:
if not N.isscalar(efac):
raise ValueError('ERROR: If flags is None, efac must be a scalar')
else:
efacvec = N.ones(psr.nobs) * efac
if flags is not None and flagid is not None and not N.isscalar(efac):
if len(efac) == len(flags):
for ct, flag in enumerate(flags):
ind = flag == N.array(psr.flagvals(flagid))
efacvec[ind] = efac[ct]
psr.stoas[:] += efacvec * psr.toaerrs * (1e-6 / day) * N.random.randn(psr.nobs)
|
python
|
{
"resource": ""
}
|
q7015
|
extrap1d
|
train
|
def extrap1d(interpolator):
"""
Function to extend an interpolation function to an
extrapolation function.
:param interpolator: scipy interp1d object
:returns ufunclike: extension of function to extrapolation
"""
xs = interpolator.x
ys = interpolator.y
def pointwise(x):
if x < xs[0]:
return ys[0] # +(x-xs[0])*(ys[1]-ys[0])/(xs[1]-xs[0])
elif x > xs[-1]:
return ys[-1] # +(x-xs[-1])*(ys[-1]-ys[-2])/(xs[-1]-xs[-2])
else:
return interpolator(x)
def ufunclike(xs):
return N.array(map(pointwise, N.array(xs)))
return ufunclike
|
python
|
{
"resource": ""
}
|
q7016
|
computeORFMatrix
|
train
|
def computeORFMatrix(psr):
"""
Compute ORF matrix.
:param psr: List of pulsar object instances
:returns: Matrix that has the ORF values for every pulsar
pair with 2 on the diagonals to account for the
pulsar term.
"""
# begin loop over all pulsar pairs and calculate ORF
npsr = len(psr)
ORF = N.zeros((npsr, npsr))
phati = N.zeros(3)
phatj = N.zeros(3)
ptheta = [N.pi/2 - p['DECJ'].val for p in psr]
pphi = [p['RAJ'].val for p in psr]
for ll in range(0, npsr):
phati[0] = N.cos(pphi[ll]) * N.sin(ptheta[ll])
phati[1] = N.sin(pphi[ll]) * N.sin(ptheta[ll])
phati[2] = N.cos(ptheta[ll])
for kk in range(0, npsr):
phatj[0] = N.cos(pphi[kk]) * N.sin(ptheta[kk])
phatj[1] = N.sin(pphi[kk]) * N.sin(ptheta[kk])
phatj[2] = N.cos(ptheta[kk])
if ll != kk:
xip = (1.-N.sum(phati*phatj)) / 2.
ORF[ll, kk] = 3.*( 1./3. + xip * ( N.log(xip) -1./6.) )
else:
ORF[ll, kk] = 2.0
return ORF
|
python
|
{
"resource": ""
}
|
q7017
|
plotres
|
train
|
def plotres(psr,deleted=False,group=None,**kwargs):
"""Plot residuals, compute unweighted rms residual."""
res, t, errs = psr.residuals(), psr.toas(), psr.toaerrs
if (not deleted) and N.any(psr.deleted != 0):
res, t, errs = res[psr.deleted == 0], t[psr.deleted == 0], errs[psr.deleted == 0]
print("Plotting {0}/{1} nondeleted points.".format(len(res),psr.nobs))
meanres = math.sqrt(N.mean(res**2)) / 1e-6
if group is None:
i = N.argsort(t)
P.errorbar(t[i],res[i]/1e-6,yerr=errs[i],fmt='x',**kwargs)
else:
if (not deleted) and N.any(psr.deleted):
flagmask = psr.flagvals(group)[~psr.deleted]
else:
flagmask = psr.flagvals(group)
unique = list(set(flagmask))
for flagval in unique:
f = (flagmask == flagval)
flagres, flagt, flagerrs = res[f], t[f], errs[f]
i = N.argsort(flagt)
P.errorbar(flagt[i],flagres[i]/1e-6,yerr=flagerrs[i],fmt='x',**kwargs)
P.legend(unique,numpoints=1,bbox_to_anchor=(1.1,1.1))
P.xlabel('MJD'); P.ylabel('res [us]')
P.title("{0} - rms res = {1:.2f} us".format(psr.name,meanres))
|
python
|
{
"resource": ""
}
|
q7018
|
plotgwsrc
|
train
|
def plotgwsrc(gwb):
"""
Plot a GWB source population as a mollweide projection.
"""
theta, phi, omega, polarization = gwb.gw_dist()
rho = phi-N.pi
eta = 0.5*N.pi - theta
# I don't know how to get rid of the RuntimeWarning -- RvH, Oct 10, 2014:
# /Users/vhaaster/env/dev/lib/python2.7/site-packages/matplotlib/projections/geo.py:485:
# RuntimeWarning: invalid value encountered in arcsin theta = np.arcsin(y / np.sqrt(2))
#old_settings = N.seterr(invalid='ignore')
P.title("GWB source population")
ax = P.axes(projection='mollweide')
foo = P.scatter(rho, eta, marker='.', s=1)
#bar = N.seterr(**old_settings)
return foo
|
python
|
{
"resource": ""
}
|
q7019
|
merge
|
train
|
def merge(data,skip=50,fraction=1.0):
"""Merge one every 'skip' clouds into a single emcee population,
using the later 'fraction' of the run."""
w,s,d = data.chains.shape
start = int((1.0 - fraction) * s)
total = int((s - start) / skip)
return data.chains[:,start::skip,:].reshape((w*total,d))
|
python
|
{
"resource": ""
}
|
q7020
|
cull
|
train
|
def cull(data,index,min=None,max=None):
"""Sieve an emcee clouds by excluding walkers with search variable 'index'
smaller than 'min' or larger than 'max'."""
ret = data
if min is not None:
ret = ret[ret[:,index] > min,:]
if max is not None:
ret = ret[ret[:,index] < max,:]
return ret
|
python
|
{
"resource": ""
}
|
q7021
|
make_ecc_interpolant
|
train
|
def make_ecc_interpolant():
"""
Make interpolation function from eccentricity file to
determine number of harmonics to use for a given
eccentricity.
:returns: interpolant
"""
pth = resource_filename(Requirement.parse('libstempo'),
'libstempo/ecc_vs_nharm.txt')
fil = np.loadtxt(pth)
return interp1d(fil[:,0], fil[:,1])
|
python
|
{
"resource": ""
}
|
q7022
|
best_kmers
|
train
|
def best_kmers(dt, response, sequence, k=6, consider_shift=True, n_cores=1,
seq_align="start", trim_seq_len=None):
"""
Find best k-mers for CONCISE initialization.
Args:
dt (pd.DataFrame): Table containing response variable and sequence.
response (str): Name of the column used as the reponse variable.
sequence (str): Name of the column storing the DNA/RNA sequences.
k (int): Desired k-mer length.
n_cores (int): Number of cores to use for computation. It can use up to 3 cores.
consider_shift (boolean): When performing stepwise k-mer selection. Is TATTTA similar to ATTTAG?
seq_align (str): one of ``{"start", "end"}``. To which end should we align sequences?
trim_seq_len (int): Consider only first `trim_seq_len` bases of each sequence when generating the sequence design matrix. If :python:`None`, set :py:attr:`trim_seq_len` to the longest sequence length, hence whole sequences are considered.
Returns:
string list: Best set of motifs for this dataset sorted with respect to
confidence (best candidate occuring first).
Details:
First a lasso model gets fitted to get a set of initial motifs. Next, the best
subset of unrelated motifs is selected by stepwise selection.
"""
y = dt[response]
seq = dt[sequence]
if trim_seq_len is not None:
seq = pad_sequences(seq, align=seq_align, maxlen=trim_seq_len)
seq = [s.replace("N", "") for s in seq]
dt_kmer = kmer_count(seq, k)
Xsp = csc_matrix(dt_kmer)
en = ElasticNet(alpha=1, standardize=False, n_splits=3)
en.fit(Xsp, y)
# which coefficients are nonzero?=
nonzero_kmers = dt_kmer.columns.values[en.coef_ != 0].tolist()
# perform stepwise selection
#
# TODO - how do we deal with the intercept?
# largest number of motifs where they don't differ by more than 1 k-mer
def find_next_best(dt_kmer, y, selected_kmers, to_be_selected_kmers, consider_shift=True):
"""
perform stepwise model selection while preventing to add a motif similar to the
already selected motifs.
"""
F, pval = f_regression(dt_kmer[to_be_selected_kmers], y)
kmer = to_be_selected_kmers.pop(pval.argmin())
selected_kmers.append(kmer)
def select_criterion(s1, s2, consider_shift=True):
if hamming_distance(s1, s2) <= 1:
return False
if consider_shift and hamming_distance(s1[1:], s2[:-1]) == 0:
return False
if consider_shift and hamming_distance(s1[:-1], s2[1:]) == 0:
return False
return True
to_be_selected_kmers = [ckmer for ckmer in to_be_selected_kmers
if select_criterion(ckmer, kmer, consider_shift)]
if len(to_be_selected_kmers) == 0:
return selected_kmers
else:
# regress out the new feature
lm = LinearRegression()
lm.fit(dt_kmer[selected_kmers], y)
y_new = y - lm.predict(dt_kmer[selected_kmers])
return find_next_best(dt_kmer, y_new, selected_kmers, to_be_selected_kmers, consider_shift)
selected_kmers = find_next_best(dt_kmer, y, [], nonzero_kmers, consider_shift)
return selected_kmers
|
python
|
{
"resource": ""
}
|
q7023
|
kmer_count
|
train
|
def kmer_count(seq_list, k):
"""
Generate k-mer counts from a set of sequences
Args:
seq_list (iterable): List of DNA sequences (with letters from {A, C, G, T})
k (int): K in k-mer.
Returns:
pandas.DataFrame: Count matrix for seach sequence in seq_list
Example:
>>> kmer_count(["ACGTTAT", "GACGCGA"], 2)
AA AC AG AT CA CC CG CT GA GC GG GT TA TC TG TT
0 0 1 0 1 0 0 1 0 0 0 0 1 1 0 0 1
1 0 1 0 0 0 0 2 0 2 1 0 0 0 0 0 0
"""
# generate all k-mers
all_kmers = generate_all_kmers(k)
kmer_count_list = []
for seq in seq_list:
kmer_count_list.append([seq.count(kmer) for kmer in all_kmers])
return pd.DataFrame(kmer_count_list, columns=all_kmers)
|
python
|
{
"resource": ""
}
|
q7024
|
generate_all_kmers
|
train
|
def generate_all_kmers(k):
"""
Generate all possible k-mers
Example:
>>> generate_all_kmers(2)
['AA', 'AC', 'AG', 'AT', 'CA', 'CC', 'CG', 'CT', 'GA', 'GC', 'GG', 'GT', 'TA', 'TC', 'TG', 'TT']
"""
bases = ['A', 'C', 'G', 'T']
return [''.join(p) for p in itertools.product(bases, repeat=k)]
|
python
|
{
"resource": ""
}
|
q7025
|
dict_to_numpy_dict
|
train
|
def dict_to_numpy_dict(obj_dict):
"""
Convert a dictionary of lists into a dictionary of numpy arrays
"""
return {key: np.asarray(value) if value is not None else None for key, value in obj_dict.items()}
|
python
|
{
"resource": ""
}
|
q7026
|
rec_dict_to_numpy_dict
|
train
|
def rec_dict_to_numpy_dict(obj_dict):
"""
Same as dict_to_numpy_dict, but recursive
"""
if type(obj_dict) == dict:
return {key: rec_dict_to_numpy_dict(value) if value is not None else None for key, value in obj_dict.items()}
elif obj_dict is None:
return None
else:
return np.asarray(obj_dict)
|
python
|
{
"resource": ""
}
|
q7027
|
compare_numpy_dict
|
train
|
def compare_numpy_dict(a, b, exact=True):
"""
Compare two recursive numpy dictionaries
"""
if type(a) != type(b) and type(a) != np.ndarray and type(b) != np.ndarray:
return False
# go through a dictionary
if type(a) == dict and type(b) == dict:
if not a.keys() == b.keys():
return False
for key in a.keys():
res = compare_numpy_dict(a[key], b[key], exact)
if res == False:
print("false for key = ", key)
return False
return True
# if type(a) == np.ndarray and type(b) == np.ndarray:
if type(a) == np.ndarray or type(b) == np.ndarray:
if exact:
return (a == b).all()
else:
return np.testing.assert_almost_equal(a, b)
if a is None and b is None:
return True
raise NotImplementedError
|
python
|
{
"resource": ""
}
|
q7028
|
BSpline.getS
|
train
|
def getS(self, add_intercept=False):
"""Get the penalty matrix S
Returns
np.array, of shape (n_bases + add_intercept, n_bases + add_intercept)
"""
S = self.S
if add_intercept is True:
# S <- cbind(0, rbind(0, S)) # in R
zeros = np.zeros_like(S[:1, :])
S = np.vstack([zeros, S])
zeros = np.zeros_like(S[:, :1])
S = np.hstack([zeros, S])
return S
|
python
|
{
"resource": ""
}
|
q7029
|
get_pwm_list
|
train
|
def get_pwm_list(motif_name_list, pseudocountProb=0.0001):
"""Get a list of ENCODE PWM's.
# Arguments
pwm_id_list: List of id's from the `PWM_id` column in `get_metadata()` table
pseudocountProb: Added pseudocount probabilities to the PWM
# Returns
List of `concise.utils.pwm.PWM` instances.
"""
l = _load_motifs()
l = {k.split()[0]: v for k, v in l.items()}
pwm_list = [PWM(l[m] + pseudocountProb, name=m) for m in motif_name_list]
return pwm_list
|
python
|
{
"resource": ""
}
|
q7030
|
auc
|
train
|
def auc(y_true, y_pred, round=True):
"""Area under the ROC curve
"""
y_true, y_pred = _mask_value_nan(y_true, y_pred)
if round:
y_true = y_true.round()
if len(y_true) == 0 or len(np.unique(y_true)) < 2:
return np.nan
return skm.roc_auc_score(y_true, y_pred)
|
python
|
{
"resource": ""
}
|
q7031
|
recall_at_precision
|
train
|
def recall_at_precision(y_true, y_pred, precision):
"""Recall at a certain precision threshold
Args:
y_true: true labels
y_pred: predicted labels
precision: resired precision level at which where to compute the recall
"""
y_true, y_pred = _mask_value_nan(y_true, y_pred)
precision, recall, _ = skm.precision_recall_curve(y_true, y_pred)
return recall[np.searchsorted(precision - precision, 0)]
|
python
|
{
"resource": ""
}
|
q7032
|
cor
|
train
|
def cor(y_true, y_pred):
"""Compute Pearson correlation coefficient.
"""
y_true, y_pred = _mask_nan(y_true, y_pred)
return np.corrcoef(y_true, y_pred)[0, 1]
|
python
|
{
"resource": ""
}
|
q7033
|
kendall
|
train
|
def kendall(y_true, y_pred, nb_sample=100000):
"""Kendall's tau coefficient, Kendall rank correlation coefficient
"""
y_true, y_pred = _mask_nan(y_true, y_pred)
if len(y_true) > nb_sample:
idx = np.arange(len(y_true))
np.random.shuffle(idx)
idx = idx[:nb_sample]
y_true = y_true[idx]
y_pred = y_pred[idx]
return kendalltau(y_true, y_pred)[0]
|
python
|
{
"resource": ""
}
|
q7034
|
mad
|
train
|
def mad(y_true, y_pred):
"""Median absolute deviation
"""
y_true, y_pred = _mask_nan(y_true, y_pred)
return np.mean(np.abs(y_true - y_pred))
|
python
|
{
"resource": ""
}
|
q7035
|
mse
|
train
|
def mse(y_true, y_pred):
"""Mean squared error
"""
y_true, y_pred = _mask_nan(y_true, y_pred)
return ((y_true - y_pred) ** 2).mean(axis=None)
|
python
|
{
"resource": ""
}
|
q7036
|
sample_params
|
train
|
def sample_params(params):
"""Randomly sample hyper-parameters stored in a dictionary on a predefined range and scale.
Useful for hyper-parameter random search.
Args:
params (dict): hyper-parameters to sample. Dictionary value-type parsing:
- :python:`[1e3, 1e7]` - uniformly sample on a **log10** scale from the interval :python:`(1e3,1e7)`
- :python:`(1, 10)` - uniformly sample on a **normal** scale from the interval :python:`(1,10)`
- :python:`{1, 2}` - sample from a **set** of values.
- :python:`1` - don't sample
Returns:
dict: Dictionary with the same keys as :py:attr:`params`, but with only one element as the value.
Examples:
>>> myparams = {
"max_pool": True, # allways use True
"step_size": [0.09, 0.005],
"step_decay": (0.9, 1),
"n_splines": {10, None}, # use either 10 or None
"some_tuple": {(1,2), (1)},
}
>>> concise.sample_params(myparams)
{'step_decay': 0.9288, 'step_size': 0.0292, 'max_pool': True, 'n_splines': None, 'some_tuple': (1, 2)}
>>> concise.sample_params(myparams)
{'step_decay': 0.9243, 'step_size': 0.0293, 'max_pool': True, 'n_splines': None, 'some_tuple': (1)}
>>> concise.sample_params(myparams)
{'step_decay': 0.9460, 'step_size': 0.0301, 'max_pool': True, 'n_splines': 10, 'some_tuple': (1, 2)}
Note:
- :python:`{[1,2], [3,4]}` is invalid. Use :python:`{(1,2), (3,4)}` instead.
- You can allways use :python:`{}` with a single element to by-pass sampling.
"""
def sample_log(myrange):
x = np.random.uniform(np.log10(myrange[0]), np.log10(myrange[1]))
return 10**x
def sample_unif(myrange):
x = np.random.uniform(myrange[0], myrange[1])
return x
def sample_set(myset):
x = random.sample(myset, 1)
return x[0]
def type_dep_sample(myrange):
if type(myrange) is list:
return sample_log(myrange)
if type(myrange) is tuple:
return sample_unif(myrange)
if type(myrange) is set:
return sample_set(myrange)
return myrange
return {k: type_dep_sample(v) for k, v in params.items()}
|
python
|
{
"resource": ""
}
|
q7037
|
cat_acc
|
train
|
def cat_acc(y, z):
"""Classification accuracy for multi-categorical case
"""
weights = _cat_sample_weights(y)
_acc = K.cast(K.equal(K.argmax(y, axis=-1),
K.argmax(z, axis=-1)),
K.floatx())
_acc = K.sum(_acc * weights) / K.sum(weights)
return _acc
|
python
|
{
"resource": ""
}
|
q7038
|
split_KFold_idx
|
train
|
def split_KFold_idx(train, cv_n_folds=5, stratified=False, random_state=None):
"""Get k-fold indices generator
"""
test_len(train)
y = train[1]
n_rows = y.shape[0]
if stratified:
if len(y.shape) > 1:
if y.shape[1] > 1:
raise ValueError("Can't use stratified K-fold with multi-column response variable")
else:
y = y[:, 0]
# http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.StratifiedKFold.html#sklearn.model_selection.StratifiedKFold.split
return model_selection.StratifiedKFold(n_splits=cv_n_folds, shuffle=True, random_state=random_state)\
.split(X=np.zeros((n_rows, 1)), y=y)
else:
return model_selection.KFold(n_splits=cv_n_folds, shuffle=True, random_state=random_state)\
.split(X=np.zeros((n_rows, 1)))
|
python
|
{
"resource": ""
}
|
q7039
|
prepare_data
|
train
|
def prepare_data(dt, features, response, sequence, id_column=None, seq_align="end", trim_seq_len=None):
"""
Prepare data for Concise.train or ConciseCV.train.
Args:
dt: A pandas DataFrame containing all the required data.
features (List of strings): Column names of `dt` used to produce the features design matrix. These columns should be numeric.
response (str or list of strings): Name(s) of column(s) used as a reponse variable.
sequence (str): Name of the column storing the DNA/RNA sequences.
id_column (str): Name of the column used as the row identifier.
seq_align (str): one of ``{"start", "end"}``. To which end should we align sequences?
trim_seq_len (int): Consider only first `trim_seq_len` bases of each sequence when generating the sequence design matrix. If :python:`None`, set :py:attr:`trim_seq_len` to the longest sequence length, hence whole sequences are considered.
standardize_features (bool): If True, column in the returned matrix matrix :py:attr:`X_seq` are normalied to have zero mean and unit variance.
Returns:
tuple: Tuple with elements: :code:`(X_feat: X_seq, y, id_vec)`, where:
- :py:attr:`X_feat`: features design matrix of shape :code:`(N, D)`, where N is :code:`len(dt)` and :code:`D = len(features)`
- :py:attr:`X_seq`: sequence matrix of shape :code:`(N, 1, trim_seq_len, 4)`. It represents 1-hot encoding of the DNA/RNA sequence.
- :py:attr:`y`: Response variable 1-column matrix of shape :code:`(N, 1)`
- :py:attr:`id_vec`: 1D Character array of shape :code:`(N)`. It represents the ID's of individual rows.
Note:
One-hot encoding of the DNA/RNA sequence is the following:
.. code:: python
{
"A": np.array([1, 0, 0, 0]),
"C": np.array([0, 1, 0, 0]),
"G": np.array([0, 0, 1, 0]),
"T": np.array([0, 0, 0, 1]),
"U": np.array([0, 0, 0, 1]),
"N": np.array([0, 0, 0, 0]),
}
"""
if type(response) is str:
response = [response]
X_feat = np.array(dt[features], dtype="float32")
y = np.array(dt[response], dtype="float32")
X_seq = encodeDNA(seq_vec=dt[sequence],
maxlen=trim_seq_len,
seq_align=seq_align)
X_seq = np.array(X_seq, dtype="float32")
id_vec = np.array(dt[id_column])
return X_feat, X_seq, y, id_vec
|
python
|
{
"resource": ""
}
|
q7040
|
EncodeSplines.fit
|
train
|
def fit(self, x):
"""Calculate the knot placement from the values ranges.
# Arguments
x: numpy array, either N x D or N x L x D dimensional.
"""
assert x.ndim > 1
self.data_min_ = np.min(x, axis=tuple(range(x.ndim - 1)))
self.data_max_ = np.max(x, axis=tuple(range(x.ndim - 1)))
if self.share_knots:
self.data_min_[:] = np.min(self.data_min_)
self.data_max_[:] = np.max(self.data_max_)
|
python
|
{
"resource": ""
}
|
q7041
|
EncodeSplines.transform
|
train
|
def transform(self, x, warn=True):
"""Obtain the transformed values
"""
# 1. split across last dimension
# 2. re-use ranges
# 3. Merge
array_list = [encodeSplines(x[..., i].reshape((-1, 1)),
n_bases=self.n_bases,
spline_order=self.degree,
warn=warn,
start=self.data_min_[i],
end=self.data_max_[i]).reshape(x[..., i].shape + (self.n_bases,))
for i in range(x.shape[-1])]
return np.stack(array_list, axis=-2)
|
python
|
{
"resource": ""
}
|
q7042
|
InputCodon
|
train
|
def InputCodon(seq_length, ignore_stop_codons=True, name=None, **kwargs):
"""Input placeholder for array returned by `encodeCodon`
Note: The seq_length is divided by 3
Wrapper for: `keras.layers.Input((seq_length / 3, 61 or 61), name=name, **kwargs)`
"""
if ignore_stop_codons:
vocab = CODONS
else:
vocab = CODONS + STOP_CODONS
assert seq_length % 3 == 0
return Input((seq_length / 3, len(vocab)), name=name, **kwargs)
|
python
|
{
"resource": ""
}
|
q7043
|
InputAA
|
train
|
def InputAA(seq_length, name=None, **kwargs):
"""Input placeholder for array returned by `encodeAA`
Wrapper for: `keras.layers.Input((seq_length, 22), name=name, **kwargs)`
"""
return Input((seq_length, len(AMINO_ACIDS)), name=name, **kwargs)
|
python
|
{
"resource": ""
}
|
q7044
|
InputRNAStructure
|
train
|
def InputRNAStructure(seq_length, name=None, **kwargs):
"""Input placeholder for array returned by `encodeRNAStructure`
Wrapper for: `keras.layers.Input((seq_length, 5), name=name, **kwargs)`
"""
return Input((seq_length, len(RNAplfold_PROFILES)), name=name, **kwargs)
|
python
|
{
"resource": ""
}
|
q7045
|
ConvSequence._plot_weights_heatmap
|
train
|
def _plot_weights_heatmap(self, index=None, figsize=None, **kwargs):
"""Plot weights as a heatmap
index = can be a particular index or a list of indicies
**kwargs - additional arguments to concise.utils.plot.heatmap
"""
W = self.get_weights()[0]
if index is None:
index = np.arange(W.shape[2])
fig = heatmap(np.swapaxes(W[:, :, index], 0, 1), plot_name="filter: ",
vocab=self.VOCAB, figsize=figsize, **kwargs)
# plt.show()
return fig
|
python
|
{
"resource": ""
}
|
q7046
|
ConvSequence._plot_weights_motif
|
train
|
def _plot_weights_motif(self, index, plot_type="motif_raw",
background_probs=DEFAULT_BASE_BACKGROUND,
ncol=1,
figsize=None):
"""Index can only be a single int
"""
w_all = self.get_weights()
if len(w_all) == 0:
raise Exception("Layer needs to be initialized first")
W = w_all[0]
if index is None:
index = np.arange(W.shape[2])
if isinstance(index, int):
index = [index]
fig = plt.figure(figsize=figsize)
if plot_type == "motif_pwm" and plot_type in self.AVAILABLE_PLOTS:
arr = pssm_array2pwm_array(W, background_probs)
elif plot_type == "motif_raw" and plot_type in self.AVAILABLE_PLOTS:
arr = W
elif plot_type == "motif_pwm_info" and plot_type in self.AVAILABLE_PLOTS:
quasi_pwm = pssm_array2pwm_array(W, background_probs)
arr = _pwm2pwm_info(quasi_pwm)
else:
raise ValueError("plot_type needs to be from {0}".format(self.AVAILABLE_PLOTS))
fig = seqlogo_fig(arr, vocab=self.VOCAB_name, figsize=figsize, ncol=ncol, plot_name="filter: ")
# fig.show()
return fig
|
python
|
{
"resource": ""
}
|
q7047
|
ConvSequence.plot_weights
|
train
|
def plot_weights(self, index=None, plot_type="motif_raw", figsize=None, ncol=1, **kwargs):
"""Plot filters as heatmap or motifs
index = can be a particular index or a list of indicies
**kwargs - additional arguments to concise.utils.plot.heatmap
"""
if "heatmap" in self.AVAILABLE_PLOTS and plot_type == "heatmap":
return self._plot_weights_heatmap(index=index, figsize=figsize, ncol=ncol, **kwargs)
elif plot_type[:5] == "motif":
return self._plot_weights_motif(index=index, plot_type=plot_type, figsize=figsize, ncol=ncol, **kwargs)
else:
raise ValueError("plot_type needs to be from {0}".format(self.AVAILABLE_PLOTS))
|
python
|
{
"resource": ""
}
|
q7048
|
_check_pwm_list
|
train
|
def _check_pwm_list(pwm_list):
"""Check the input validity
"""
for pwm in pwm_list:
if not isinstance(pwm, PWM):
raise TypeError("element {0} of pwm_list is not of type PWM".format(pwm))
return True
|
python
|
{
"resource": ""
}
|
q7049
|
heatmap
|
train
|
def heatmap(w, vmin=None, vmax=None, diverge_color=False,
ncol=1,
plot_name=None, vocab=["A", "C", "G", "T"], figsize=(6, 2)):
"""Plot a heatmap from weight matrix w
vmin, vmax = z axis range
diverge_color = Should we use diverging colors?
plot_name = plot_title
vocab = vocabulary (corresponds to the first axis)
"""
# Generate y and x values from the dimension lengths
assert len(vocab) == w.shape[0]
plt_y = np.arange(w.shape[0] + 1) + 0.5
plt_x = np.arange(w.shape[1] + 1) - 0.5
z_min = w.min()
z_max = w.max()
if vmin is None:
vmin = z_min
if vmax is None:
vmax = z_max
if diverge_color:
color_map = plt.cm.RdBu
else:
color_map = plt.cm.Blues
fig = plt.figure(figsize=figsize)
# multiple axis
if len(w.shape) == 3:
#
n_plots = w.shape[2]
nrow = math.ceil(n_plots / ncol)
else:
n_plots = 1
nrow = 1
ncol = 1
for i in range(n_plots):
if len(w.shape) == 3:
w_cur = w[:, :, i]
else:
w_cur = w
ax = plt.subplot(nrow, ncol, i + 1)
plt.tight_layout()
im = ax.pcolormesh(plt_x, plt_y, w_cur, cmap=color_map,
vmin=vmin, vmax=vmax, edgecolors="white")
ax.grid(False)
ax.set_yticklabels([""] + vocab, minor=False)
ax.yaxis.set_major_locator(MaxNLocator(integer=True))
ax.set_xticks(np.arange(w_cur.shape[1] + 1))
ax.set_xlim(plt_x.min(), plt_x.max())
ax.set_ylim(plt_y.min(), plt_y.max())
# nice scale location:
# http://stackoverflow.com/questions/18195758/set-matplotlib-colorbar-size-to-match-graph
divider = make_axes_locatable(ax)
cax = divider.append_axes("right", size="5%", pad=0.05)
fig.colorbar(im, cax=cax)
if plot_name is not None:
if n_plots > 0:
pln = plot_name + " {0}".format(i)
else:
pln = plot_name
ax.set_title(pln)
ax.set_aspect('equal')
return fig
|
python
|
{
"resource": ""
}
|
q7050
|
add_letter_to_axis
|
train
|
def add_letter_to_axis(ax, let, col, x, y, height):
"""Add 'let' with position x,y and height height to matplotlib axis 'ax'.
"""
if len(let) == 2:
colors = [col, "white"]
elif len(let) == 1:
colors = [col]
else:
raise ValueError("3 or more Polygons are not supported")
for polygon, color in zip(let, colors):
new_polygon = affinity.scale(
polygon, yfact=height, origin=(0, 0, 0))
new_polygon = affinity.translate(
new_polygon, xoff=x, yoff=y)
patch = PolygonPatch(
new_polygon, edgecolor=color, facecolor=color)
ax.add_patch(patch)
return
|
python
|
{
"resource": ""
}
|
q7051
|
seqlogo
|
train
|
def seqlogo(letter_heights, vocab="DNA", ax=None):
"""Make a logo plot
# Arguments
letter_heights: "motif length" x "vocabulary size" numpy array
Can also contain negative values.
vocab: str, Vocabulary name. Can be: DNA, RNA, AA, RNAStruct.
ax: matplotlib axis
"""
ax = ax or plt.gca()
assert letter_heights.shape[1] == len(VOCABS[vocab])
x_range = [1, letter_heights.shape[0]]
pos_heights = np.copy(letter_heights)
pos_heights[letter_heights < 0] = 0
neg_heights = np.copy(letter_heights)
neg_heights[letter_heights > 0] = 0
for x_pos, heights in enumerate(letter_heights):
letters_and_heights = sorted(zip(heights, list(VOCABS[vocab].keys())))
y_pos_pos = 0.0
y_neg_pos = 0.0
for height, letter in letters_and_heights:
color = VOCABS[vocab][letter]
polygons = letter_polygons[letter]
if height > 0:
add_letter_to_axis(ax, polygons, color, 0.5 + x_pos, y_pos_pos, height)
y_pos_pos += height
else:
add_letter_to_axis(ax, polygons, color, 0.5 + x_pos, y_neg_pos, height)
y_neg_pos += height
# if add_hline:
# ax.axhline(color="black", linewidth=1)
ax.set_xlim(x_range[0] - 1, x_range[1] + 1)
ax.grid(False)
ax.set_xticks(list(range(*x_range)) + [x_range[-1]])
ax.set_aspect(aspect='auto', adjustable='box')
ax.autoscale_view()
|
python
|
{
"resource": ""
}
|
q7052
|
get_cv_accuracy
|
train
|
def get_cv_accuracy(res):
"""
Extract the cv accuracy from the model
"""
ac_list = [(accuracy["train_acc_final"],
accuracy["test_acc_final"]
)
for accuracy, weights in res]
ac = np.array(ac_list)
perf = {
"mean_train_acc": np.mean(ac[:, 0]),
"std_train_acc": np.std(ac[:, 0]),
"mean_test_acc": np.mean(ac[:, 1]),
"std_test_acc": np.std(ac[:, 1]),
}
return perf
|
python
|
{
"resource": ""
}
|
q7053
|
one_hot2string
|
train
|
def one_hot2string(arr, vocab):
"""Convert a one-hot encoded array back to string
"""
tokens = one_hot2token(arr)
indexToLetter = _get_index_dict(vocab)
return [''.join([indexToLetter[x] for x in row]) for row in tokens]
|
python
|
{
"resource": ""
}
|
q7054
|
tokenize
|
train
|
def tokenize(seq, vocab, neutral_vocab=[]):
"""Convert sequence to integers
# Arguments
seq: Sequence to encode
vocab: Vocabulary to use
neutral_vocab: Neutral vocabulary -> assign those values to -1
# Returns
List of length `len(seq)` with integers from `-1` to `len(vocab) - 1`
"""
# Req: all vocabs have the same length
if isinstance(neutral_vocab, str):
neutral_vocab = [neutral_vocab]
nchar = len(vocab[0])
for l in vocab + neutral_vocab:
assert len(l) == nchar
assert len(seq) % nchar == 0 # since we are using striding
vocab_dict = _get_vocab_dict(vocab)
for l in neutral_vocab:
vocab_dict[l] = -1
# current performance bottleneck
return [vocab_dict[seq[(i * nchar):((i + 1) * nchar)]] for i in range(len(seq) // nchar)]
|
python
|
{
"resource": ""
}
|
q7055
|
encodeSequence
|
train
|
def encodeSequence(seq_vec, vocab, neutral_vocab, maxlen=None,
seq_align="start", pad_value="N", encode_type="one_hot"):
"""Convert a list of genetic sequences into one-hot-encoded array.
# Arguments
seq_vec: list of strings (genetic sequences)
vocab: list of chars: List of "words" to use as the vocabulary. Can be strings of length>0,
but all need to have the same length. For DNA, this is: ["A", "C", "G", "T"].
neutral_vocab: list of chars: Values used to pad the sequence or represent unknown-values. For DNA, this is: ["N"].
maxlen: int or None,
Should we trim (subset) the resulting sequence. If None don't trim.
Note that trims wrt the align parameter.
It should be smaller than the longest sequence.
seq_align: character; 'end' or 'start'
To which end should we align sequences?
encode_type: "one_hot" or "token". "token" represents each vocab element as a positive integer from 1 to len(vocab) + 1.
neutral_vocab is represented with 0.
# Returns
Array with shape for encode_type:
- "one_hot": `(len(seq_vec), maxlen, len(vocab))`
- "token": `(len(seq_vec), maxlen)`
If `maxlen=None`, it gets the value of the longest sequence length from `seq_vec`.
"""
if isinstance(neutral_vocab, str):
neutral_vocab = [neutral_vocab]
if isinstance(seq_vec, str):
raise ValueError("seq_vec should be an iterable returning " +
"strings not a string itself")
assert len(vocab[0]) == len(pad_value)
assert pad_value in neutral_vocab
assert encode_type in ["one_hot", "token"]
seq_vec = pad_sequences(seq_vec, maxlen=maxlen,
align=seq_align, value=pad_value)
if encode_type == "one_hot":
arr_list = [token2one_hot(tokenize(seq, vocab, neutral_vocab), len(vocab))
for i, seq in enumerate(seq_vec)]
elif encode_type == "token":
arr_list = [1 + np.array(tokenize(seq, vocab, neutral_vocab)) for seq in seq_vec]
# we add 1 to be compatible with keras: https://keras.io/layers/embeddings/
# indexes > 0, 0 = padding element
return np.stack(arr_list)
|
python
|
{
"resource": ""
}
|
q7056
|
encodeDNA
|
train
|
def encodeDNA(seq_vec, maxlen=None, seq_align="start"):
"""Convert the DNA sequence into 1-hot-encoding numpy array
# Arguments
seq_vec: list of chars
List of sequences that can have different lengths
maxlen: int or None,
Should we trim (subset) the resulting sequence. If None don't trim.
Note that trims wrt the align parameter.
It should be smaller than the longest sequence.
seq_align: character; 'end' or 'start'
To which end should we align sequences?
# Returns
3D numpy array of shape (len(seq_vec), trim_seq_len(or maximal sequence length if None), 4)
# Example
```python
>>> sequence_vec = ['CTTACTCAGA', 'TCTTTA']
>>> X_seq = encodeDNA(sequence_vec, seq_align="end", maxlen=8)
>>> X_seq.shape
(2, 8, 4)
>>> print(X_seq)
[[[0 0 0 1]
[1 0 0 0]
[0 1 0 0]
[0 0 0 1]
[0 1 0 0]
[1 0 0 0]
[0 0 1 0]
[1 0 0 0]]
[[0 0 0 0]
[0 0 0 0]
[0 0 0 1]
[0 1 0 0]
[0 0 0 1]
[0 0 0 1]
[0 0 0 1]
[1 0 0 0]]]
```
"""
return encodeSequence(seq_vec,
vocab=DNA,
neutral_vocab="N",
maxlen=maxlen,
seq_align=seq_align,
pad_value="N",
encode_type="one_hot")
|
python
|
{
"resource": ""
}
|
q7057
|
encodeRNA
|
train
|
def encodeRNA(seq_vec, maxlen=None, seq_align="start"):
"""Convert the RNA sequence into 1-hot-encoding numpy array as for encodeDNA
"""
return encodeSequence(seq_vec,
vocab=RNA,
neutral_vocab="N",
maxlen=maxlen,
seq_align=seq_align,
pad_value="N",
encode_type="one_hot")
|
python
|
{
"resource": ""
}
|
q7058
|
encodeCodon
|
train
|
def encodeCodon(seq_vec, ignore_stop_codons=True, maxlen=None, seq_align="start", encode_type="one_hot"):
"""Convert the Codon sequence into 1-hot-encoding numpy array
# Arguments
seq_vec: List of strings/DNA sequences
ignore_stop_codons: boolean; if True, STOP_CODONS are omitted from one-hot encoding.
maxlen: Maximum sequence length. See `pad_sequences` for more detail
seq_align: How to align the sequences of variable lengths. See `pad_sequences` for more detail
encode_type: can be `"one_hot"` or `token` for token encoding of codons (incremental integer ).
# Returns
numpy.ndarray of shape `(len(seq_vec), maxlen / 3, 61 if ignore_stop_codons else 64)`
"""
if ignore_stop_codons:
vocab = CODONS
neutral_vocab = STOP_CODONS + ["NNN"]
else:
vocab = CODONS + STOP_CODONS
neutral_vocab = ["NNN"]
# replace all U's with A's?
seq_vec = [str(seq).replace("U", "T") for seq in seq_vec]
return encodeSequence(seq_vec,
vocab=vocab,
neutral_vocab=neutral_vocab,
maxlen=maxlen,
seq_align=seq_align,
pad_value="NNN",
encode_type=encode_type)
|
python
|
{
"resource": ""
}
|
q7059
|
encodeAA
|
train
|
def encodeAA(seq_vec, maxlen=None, seq_align="start", encode_type="one_hot"):
"""Convert the Amino-acid sequence into 1-hot-encoding numpy array
# Arguments
seq_vec: List of strings/amino-acid sequences
maxlen: Maximum sequence length. See `pad_sequences` for more detail
seq_align: How to align the sequences of variable lengths. See `pad_sequences` for more detail
encode_type: can be `"one_hot"` or `token` for token encoding of codons (incremental integer ).
# Returns
numpy.ndarray of shape `(len(seq_vec), maxlen, 22)`
"""
return encodeSequence(seq_vec,
vocab=AMINO_ACIDS,
neutral_vocab="_",
maxlen=maxlen,
seq_align=seq_align,
pad_value="_",
encode_type=encode_type)
|
python
|
{
"resource": ""
}
|
q7060
|
_validate_pos
|
train
|
def _validate_pos(df):
"""Validates the returned positional object
"""
assert isinstance(df, pd.DataFrame)
assert ["seqname", "position", "strand"] == df.columns.tolist()
assert df.position.dtype == np.dtype("int64")
assert df.strand.dtype == np.dtype("O")
assert df.seqname.dtype == np.dtype("O")
return df
|
python
|
{
"resource": ""
}
|
q7061
|
get_pwm_list
|
train
|
def get_pwm_list(pwm_id_list, pseudocountProb=0.0001):
"""Get a list of Attract PWM's.
# Arguments
pwm_id_list: List of id's from the `PWM_id` column in `get_metadata()` table
pseudocountProb: Added pseudocount probabilities to the PWM
# Returns
List of `concise.utils.pwm.PWM` instances.
"""
l = load_motif_db(ATTRACT_PWM)
l = {k.split()[0]: v for k, v in l.items()}
pwm_list = [PWM(l[str(m)] + pseudocountProb, name=m) for m in pwm_id_list]
return pwm_list
|
python
|
{
"resource": ""
}
|
q7062
|
mask_loss
|
train
|
def mask_loss(loss, mask_value=MASK_VALUE):
"""Generates a new loss function that ignores values where `y_true == mask_value`.
# Arguments
loss: str; name of the keras loss function from `keras.losses`
mask_value: int; which values should be masked
# Returns
function; Masked version of the `loss`
# Example
```python
categorical_crossentropy_masked = mask_loss("categorical_crossentropy")
```
"""
loss_fn = kloss.deserialize(loss)
def masked_loss_fn(y_true, y_pred):
# currently not suppoerd with NA's:
# - there is no K.is_nan impolementation in keras.backend
# - https://github.com/fchollet/keras/issues/1628
mask = K.cast(K.not_equal(y_true, mask_value), K.floatx())
# we divide by the mean to correct for the number of done loss evaluations
return loss_fn(y_true * mask, y_pred * mask) / K.mean(mask)
masked_loss_fn.__name__ = loss + "_masked"
return masked_loss_fn
|
python
|
{
"resource": ""
}
|
q7063
|
get_pwm_list
|
train
|
def get_pwm_list(pwm_id_list, pseudocountProb=0.0001):
"""Get a list of HOCOMOCO PWM's.
# Arguments
pwm_id_list: List of id's from the `PWM_id` column in `get_metadata()` table
pseudocountProb: Added pseudocount probabilities to the PWM
# Returns
List of `concise.utils.pwm.PWM` instances.
"""
l = load_motif_db(HOCOMOCO_PWM)
l = {k.split()[0]: v for k, v in l.items()}
pwm_list = [PWM(_normalize_pwm(l[m]) + pseudocountProb, name=m) for m in pwm_id_list]
return pwm_list
|
python
|
{
"resource": ""
}
|
q7064
|
Concise._var_res_to_weights
|
train
|
def _var_res_to_weights(self, var_res):
"""
Get model weights
"""
# transform the weights into our form
motif_base_weights_raw = var_res["motif_base_weights"][0]
motif_base_weights = np.swapaxes(motif_base_weights_raw, 0, 2)
# get weights
motif_weights = var_res["motif_weights"]
motif_bias = var_res["motif_bias"]
final_bias = var_res["final_bias"]
feature_weights = var_res["feature_weights"]
# get the GAM prediction:
spline_pred = None
spline_weights = None
if self._param["n_splines"] is not None:
spline_pred = self._splines["X_spline"].dot(var_res["spline_weights"])
if self._param["spline_exp"] is True:
spline_pred = np.exp(spline_pred)
else:
spline_pred = (spline_pred + 1)
spline_pred.reshape([-1])
spline_weights = var_res["spline_weights"]
weights = {"motif_base_weights": motif_base_weights,
"motif_weights": motif_weights,
"motif_bias": motif_bias,
"final_bias": final_bias,
"feature_weights": feature_weights,
"spline_pred": spline_pred,
"spline_weights": spline_weights
}
return weights
|
python
|
{
"resource": ""
}
|
q7065
|
Concise._get_var_res
|
train
|
def _get_var_res(self, graph, var, other_var):
"""
Get the weights from our graph
"""
with tf.Session(graph=graph) as sess:
sess.run(other_var["init"])
# all_vars = tf.all_variables()
# print("All variable names")
# print([var.name for var in all_vars])
# print("All variable values")
# print(sess.run(all_vars))
var_res = self._get_var_res_sess(sess, var)
return var_res
|
python
|
{
"resource": ""
}
|
q7066
|
Concise._convert_to_var
|
train
|
def _convert_to_var(self, graph, var_res):
"""
Create tf.Variables from a list of numpy arrays
var_res: dictionary of numpy arrays with the key names corresponding to var
"""
with graph.as_default():
var = {}
for key, value in var_res.items():
if value is not None:
var[key] = tf.Variable(value, name="tf_%s" % key)
else:
var[key] = None
return var
|
python
|
{
"resource": ""
}
|
q7067
|
Concise.train
|
train
|
def train(self, X_feat, X_seq, y,
X_feat_valid=None, X_seq_valid=None, y_valid=None,
n_cores=3):
"""Train the CONCISE model
:py:attr:`X_feat`, :py:attr:`X_seq`, py:attr:`y` are preferrably returned by the :py:func:`concise.prepare_data` function.
Args:
X_feat: Numpy (float) array of shape :code:`(N, D)`. Feature design matrix storing :code:`N` training samples and :code:`D` features
X_seq: Numpy (float) array of shape :code:`(N, 1, N_seq, 4)`. It represents 1-hot encoding of the DNA/RNA sequence.(:code:`N`-seqeuences of length :code:`N_seq`)
y: Numpy (float) array of shape :code:`(N, 1)`. Response variable.
X_feat_valid: :py:attr:`X_feat` used for model validation.
X_seq_valid: :py:attr:`X_seq` used for model validation.
y: :py:attr:`y` used for model validation.
n_cores (int): Number of CPU cores used for training. If available, GPU is used for training and this argument is ignored.
"""
if X_feat_valid is None and X_seq_valid is None and y_valid is None:
X_feat_valid = X_feat
X_seq_valid = X_seq
y_valid = y
print("Using training samples also for validation ")
# insert one dimension - backcompatiblity
X_seq = np.expand_dims(X_seq, axis=1)
X_seq_valid = np.expand_dims(X_seq_valid, axis=1)
# TODO: implement the re-training feature
if self.is_trained() is True:
print("Model already fitted. Re-training feature not implemented yet")
return
# input check
assert X_seq.shape[0] == X_feat.shape[0] == y.shape[0]
assert y.shape == (X_feat.shape[0], self._num_tasks)
# extract data specific parameters
self._param["seq_length"] = X_seq.shape[2]
self._param["n_add_features"] = X_feat.shape[1]
# more input check
if not self._param["seq_length"] == X_seq_valid.shape[2]:
raise Exception("sequence lengths don't match")
# setup splines
if self._param["n_splines"] is not None:
padd_loss = self._param["motif_length"] - 1 # how much shorter is our sequence, since we don't use padding
X_spline, S, _ = splines.get_gam_splines(start=0,
end=self._param["seq_length"] - padd_loss - 1, # -1 due to zero-indexing
n_bases=self._param["n_splines"],
spline_order=3,
add_intercept=False)
self._splines = {"X_spline": X_spline,
"S": S
}
# setup graph and variables
self._graph = tf.Graph()
self._var = self._get_var_initialization(self._graph, X_feat_train=X_feat, y_train=y)
self._other_var = self._build_graph(self._graph, self._var)
# TODO: save the intialized parameters
var_res_init = self._get_var_res(self._graph, self._var, self._other_var)
self.init_weights = self._var_res_to_weights(var_res=var_res_init)
# finally train the model
# - it saves the accuracy
if self._param["optimizer"] == "adam":
_train = self._train_adam
elif self._param["optimizer"] == "lbfgs":
_train = self._train_lbfgs
else:
raise Exception("Optimizer {} not implemented".format(self._param["optimizer"]))
self._var_res = _train(X_feat, X_seq, y,
X_feat_valid, X_seq_valid, y_valid,
graph=self._graph, var=self._var, other_var=self._other_var,
early_stop_patience=self._param["early_stop_patience"],
n_cores=n_cores)
self._model_fitted = True
# TODO: maybe:
# - add y_train_accuracy
# - y_train
return True
|
python
|
{
"resource": ""
}
|
q7068
|
Concise._accuracy_in_session
|
train
|
def _accuracy_in_session(self, sess, other_var, X_feat, X_seq, y):
"""
Compute the accuracy from inside the tf session
"""
y_pred = self._predict_in_session(sess, other_var, X_feat, X_seq)
return ce.mse(y_pred, y)
|
python
|
{
"resource": ""
}
|
q7069
|
Concise._set_var_res
|
train
|
def _set_var_res(self, weights):
"""
Transform the weights to var_res
"""
if weights is None:
return
# layer 1
motif_base_weights_raw = np.swapaxes(weights["motif_base_weights"], 2, 0)
motif_base_weights = motif_base_weights_raw[np.newaxis]
motif_bias = weights["motif_bias"]
feature_weights = weights["feature_weights"]
spline_weights = weights["spline_weights"]
# filter
motif_weights = weights["motif_weights"]
final_bias = weights["final_bias"]
var_res = {
"motif_base_weights": motif_base_weights,
"motif_bias": motif_bias,
"spline_weights": spline_weights,
"feature_weights": feature_weights,
"motif_weights": motif_weights,
"final_bias": final_bias
}
# cast everything to float32
var_res = {key: value.astype(np.float32) if value is not None else None for key, value in var_res.items()}
self._var_res = var_res
|
python
|
{
"resource": ""
}
|
q7070
|
ConciseCV._get_folds
|
train
|
def _get_folds(n_rows, n_folds, use_stored):
"""
Get the used CV folds
"""
# n_folds = self._n_folds
# use_stored = self._use_stored_folds
# n_rows = self._n_rows
if use_stored is not None:
# path = '~/concise/data-offline/lw-pombe/cv_folds_5.json'
with open(os.path.expanduser(use_stored)) as json_file:
json_data = json.load(json_file)
# check if we have the same number of rows and folds:
if json_data['N_rows'] != n_rows:
raise Exception('N_rows from folds doesnt match the number of rows of X_seq, X_feat, y')
if json_data['N_folds'] != n_folds:
raise Exception('n_folds dont match', json_data['N_folds'], n_folds)
kf = [(np.array(train), np.array(test)) for (train, test) in json_data['folds']]
else:
kf = KFold(n_splits=n_folds).split(np.zeros((n_rows, 1)))
# store in a list
i = 1
folds = []
for train, test in kf:
fold = "fold_" + str(i)
folds.append((fold, train, test))
i = i + 1
return folds
|
python
|
{
"resource": ""
}
|
q7071
|
ConciseCV.train
|
train
|
def train(self, X_feat, X_seq, y, id_vec=None, n_folds=10, use_stored_folds=None, n_cores=1,
train_global_model=False):
"""Train the Concise model in cross-validation.
Args:
X_feat: See :py:func:`concise.Concise.train`
X_seq: See :py:func:`concise.Concise.train`
y: See :py:func:`concise.Concise.train`
id_vec: List of character id's used to differentiate the trainig samples. Returned by :py:func:`concise.prepare_data`.
n_folds (int): Number of CV-folds to use.
use_stored_folds (chr or None): File path to a .json file containing the fold information (as returned by :py:func:`concise.ConciseCV.get_folds`). If None, the folds are generated.
n_cores (int): Number of CPU cores used for training. If available, GPU is used for training and this argument is ignored.
train_global_model (bool): In addition to training the model in cross-validation, should the global model be fitted (using all the samples from :code:`(X_feat, X_seq, y)`).
"""
# TODO: input check - dimensions
self._use_stored_folds = use_stored_folds
self._n_folds = n_folds
self._n_rows = X_feat.shape[0]
# TODO: - fix the get_cv_accuracy
# save:
# - each model
# - each model's performance
# - each model's predictions
# - globally:
# - mean perfomance
# - sd performance
# - predictions
self._kf = self._get_folds(self._n_rows, self._n_folds, self._use_stored_folds)
cv_obj = {}
if id_vec is None:
id_vec = np.arange(1, self._n_rows + 1)
best_val_acc_epoch_l = []
for fold, train, test in self._kf:
X_feat_train = X_feat[train]
X_seq_train = X_seq[train]
y_train = y[train]
X_feat_test = X_feat[test]
X_seq_test = X_seq[test]
y_test = y[test]
id_vec_test = id_vec[test]
print(fold, "/", n_folds)
# copy the object
dc = copy.deepcopy(self._concise_model)
dc.train(X_feat_train, X_seq_train, y_train,
X_feat_test, X_seq_test, y_test,
n_cores=n_cores
)
dc._test(X_feat_test, X_seq_test, y_test, id_vec_test)
cv_obj[fold] = dc
best_val_acc_epoch_l.append(dc.get_accuracy()["best_val_acc_epoch"])
self._cv_model = cv_obj
# additionaly train the global model
if train_global_model:
dc = copy.deepcopy(self._concise_model)
# overwrite n_epochs with the best average number of best epochs
dc._param["n_epochs"] = int(np.array(best_val_acc_epoch_l).mean())
print("tranining global model with n_epochs = " + str(dc._param["n_epochs"]))
dc.train(X_feat, X_seq, y,
n_cores=n_cores
)
dc._test(X_feat, X_seq, y, id_vec)
self._concise_global_model = dc
|
python
|
{
"resource": ""
}
|
q7072
|
ConciseCV._from_dict
|
train
|
def _from_dict(self, obj_dict):
"""
Initialize a model from the dictionary
"""
self._n_folds = obj_dict["param"]["n_folds"]
self._n_rows = obj_dict["param"]["n_rows"]
self._use_stored_folds = obj_dict["param"]["use_stored_folds"]
self._concise_model = Concise.from_dict(obj_dict["init_model"])
if obj_dict["trained_global_model"] is None:
self._concise_global_model = None
else:
self._concise_global_model = Concise.from_dict(obj_dict["trained_global_model"])
self._kf = [(fold, np.asarray(train), np.asarray(test)) for fold, train, test in obj_dict["folds"]]
self._cv_model = {fold: Concise.from_dict(model_dict) for fold, model_dict in obj_dict["output"].items()}
|
python
|
{
"resource": ""
}
|
q7073
|
pwm_array2pssm_array
|
train
|
def pwm_array2pssm_array(arr, background_probs=DEFAULT_BASE_BACKGROUND):
"""Convert pwm array to pssm array
"""
b = background_probs2array(background_probs)
b = b.reshape([1, 4, 1])
return np.log(arr / b).astype(arr.dtype)
|
python
|
{
"resource": ""
}
|
q7074
|
pssm_array2pwm_array
|
train
|
def pssm_array2pwm_array(arr, background_probs=DEFAULT_BASE_BACKGROUND):
"""Convert pssm array to pwm array
"""
b = background_probs2array(background_probs)
b = b.reshape([1, 4, 1])
return (np.exp(arr) * b).astype(arr.dtype)
|
python
|
{
"resource": ""
}
|
q7075
|
load_motif_db
|
train
|
def load_motif_db(filename, skipn_matrix=0):
"""Read the motif file in the following format
```
>motif_name
<skip n>0.1<delim>0.2<delim>0.5<delim>0.6
...
>motif_name2
....
```
Delim can be anything supported by np.loadtxt
# Arguments
filename: str, file path
skipn_matrix: integer, number of characters to skip when reading
the numeric matrix (for Encode = 2)
# Returns
Dictionary of numpy arrays
"""
# read-lines
if filename.endswith(".gz"):
f = gzip.open(filename, 'rt', encoding='utf-8')
else:
f = open(filename, 'r')
lines = f.readlines()
f.close()
motifs_dict = {}
motif_lines = ""
motif_name = None
def lines2matrix(lines):
return np.loadtxt(StringIO(lines))
for line in lines:
if line.startswith(">"):
if motif_lines:
# lines -> matrix
motifs_dict[motif_name] = lines2matrix(motif_lines)
motif_name = line[1:].strip()
motif_lines = ""
else:
motif_lines += line[skipn_matrix:]
if motif_lines and motif_name is not None:
motifs_dict[motif_name] = lines2matrix(motif_lines)
return motifs_dict
|
python
|
{
"resource": ""
}
|
q7076
|
iter_fasta
|
train
|
def iter_fasta(file_path):
"""Returns an iterator over the fasta file
Given a fasta file. yield tuples of header, sequence
Code modified from Brent Pedersen's:
"Correct Way To Parse A Fasta File In Python"
# Example
```python
fasta = fasta_iter("hg19.fa")
for header, seq in fasta:
print(header)
```
"""
fh = open(file_path)
# ditch the boolean (x[0]) and just keep the header or sequence since
# we know they alternate.
faiter = (x[1] for x in groupby(fh, lambda line: line[0] == ">"))
for header in faiter:
# drop the ">"
headerStr = header.__next__()[1:].strip()
# join all sequence lines to one.
seq = "".join(s.strip() for s in faiter.__next__())
yield (headerStr, seq)
|
python
|
{
"resource": ""
}
|
q7077
|
write_fasta
|
train
|
def write_fasta(file_path, seq_list, name_list=None):
"""Write a fasta file
# Arguments
file_path: file path
seq_list: List of strings
name_list: List of names corresponding to the sequences.
If not None, it should have the same length as `seq_list`
"""
if name_list is None:
name_list = [str(i) for i in range(len(seq_list))]
# needs to be dict or seq
with open(file_path, "w") as f:
for i in range(len(seq_list)):
f.write(">" + name_list[i] + "\n" + seq_list[i] + "\n")
|
python
|
{
"resource": ""
}
|
q7078
|
read_RNAplfold
|
train
|
def read_RNAplfold(tmpdir, maxlen=None, seq_align="start", pad_with="E"):
"""
pad_with = with which 2ndary structure should we pad the sequence?
"""
assert pad_with in {"P", "H", "I", "M", "E"}
def read_profile(tmpdir, P):
return [values.strip().split("\t")
for seq_name, values in iter_fasta("{tmp}/{P}_profile.fa".format(tmp=tmpdir, P=P))]
def nelem(P, pad_width):
"""get the right neutral element
"""
return 1 if P is pad_with else 0
arr_hime = np.array([pad_sequences(read_profile(tmpdir, P),
value=[nelem(P, pad_with)],
align=seq_align,
maxlen=maxlen)
for P in RNAplfold_PROFILES_EXECUTE], dtype="float32")
# add the pairness column
arr_p = 1 - arr_hime.sum(axis=0)[np.newaxis]
arr = np.concatenate((arr_p, arr_hime))
# reshape to: seq, seq_length, num_channels
arr = np.moveaxis(arr, 0, 2)
return arr
|
python
|
{
"resource": ""
}
|
q7079
|
ism
|
train
|
def ism(model, ref, ref_rc, alt, alt_rc, mutation_positions, out_annotation_all_outputs,
output_filter_mask=None, out_annotation=None, diff_type="log_odds", rc_handling="maximum"):
"""In-silico mutagenesis
Using ISM in with diff_type 'log_odds' and rc_handling 'maximum' will produce predictions as used
in [DeepSEA](http://www.nature.com/nmeth/journal/v12/n10/full/nmeth.3547.html). ISM offers two ways to
calculate the difference between the outputs created by reference and alternative sequence and two
different methods to select whether to use the output generated from the forward or from the
reverse-complement sequences. To calculate "e-values" as mentioned in DeepSEA the same ISM prediction
has to be performed on a randomised set of 1 million 1000genomes, MAF-matched variants to get a
background of predicted effects of random SNPs.
# Arguments
model: Keras model
ref: Input sequence with the reference genotype in the mutation position
ref_rc: Reverse complement of the 'ref' argument
alt: Input sequence with the alternative genotype in the mutation position
alt_rc: Reverse complement of the 'alt' argument
mutation_positions: Position on which the mutation was placed in the forward sequences
out_annotation_all_outputs: Output labels of the model.
output_filter_mask: Mask of boolean values indicating which model outputs should be used.
Use this or 'out_annotation'
out_annotation: List of outputs labels for which of the outputs (in case of a multi-task model) the
predictions should be calculated.
diff_type: "log_odds" or "diff". When set to 'log_odds' calculate scores based on log_odds, which assumes
the model output is a probability. When set to 'diff' the model output for 'ref' is subtracted
from 'alt'. Using 'log_odds' with outputs that are not in the range [0,1] nan will be returned.
rc_handling: "average" or "maximum". Either average over the predictions derived from forward and
reverse-complement predictions ('average') or pick the prediction with the bigger absolute
value ('maximum').
# Returns
Dictionary with the key `ism` which contains a pandas DataFrame containing the calculated values
for each (selected) model output and input sequence
"""
seqs = {"ref": ref, "ref_rc": ref_rc, "alt": alt, "alt_rc": alt_rc}
assert diff_type in ["log_odds", "diff"]
assert rc_handling in ["average", "maximum"]
assert np.all([np.array(get_seq_len(ref)) == np.array(get_seq_len(seqs[k])) for k in seqs.keys() if k != "ref"])
assert get_seq_len(ref)[0] == mutation_positions.shape[0]
assert len(mutation_positions.shape) == 1
# determine which outputs should be selected
if output_filter_mask is None:
if out_annotation is None:
output_filter_mask = np.arange(out_annotation_all_outputs.shape[0])
else:
output_filter_mask = np.where(np.in1d(out_annotation_all_outputs, out_annotation))[0]
# make sure the labels are assigned correctly
out_annotation = out_annotation_all_outputs[output_filter_mask]
preds = {}
for k in seqs:
# preds[k] = model.predict(seqs[k])
preds[k] = np.array(model.predict(seqs[k])[..., output_filter_mask])
if diff_type == "log_odds":
if np.any([(preds[k].min() < 0 or preds[k].max() > 1) for k in preds]):
warnings.warn("Using log_odds on model outputs that are not bound [0,1]")
diffs = np.log(preds["alt"] / (1 - preds["alt"])) - np.log(preds["ref"] / (1 - preds["ref"]))
diffs_rc = np.log(preds["alt_rc"] / (1 - preds["alt_rc"])) - np.log(preds["ref_rc"] / (1 - preds["ref_rc"]))
elif diff_type == "diff":
diffs = preds["alt"] - preds["ref"]
diffs_rc = preds["alt_rc"] - preds["ref_rc"]
if rc_handling == "average":
diffs = np.mean([diffs, diffs_rc], axis=0)
elif rc_handling == "maximum":
replace_filt = np.abs(diffs) < np.abs(diffs_rc)
diffs[replace_filt] = diffs_rc[replace_filt]
diffs = pd.DataFrame(diffs, columns=out_annotation)
return {"ism": diffs}
|
python
|
{
"resource": ""
}
|
q7080
|
_train_and_eval_single
|
train
|
def _train_and_eval_single(train, valid, model,
batch_size=32, epochs=300, use_weight=False,
callbacks=[], eval_best=False, add_eval_metrics={}):
"""Fit and evaluate a keras model
eval_best: if True, load the checkpointed model for evaluation
"""
def _format_keras_history(history):
"""nicely format keras history
"""
return {"params": history.params,
"loss": merge_dicts({"epoch": history.epoch}, history.history),
}
if use_weight:
sample_weight = train[2]
else:
sample_weight = None
# train the model
logger.info("Fit...")
history = History()
model.fit(train[0], train[1],
batch_size=batch_size,
validation_data=valid[:2],
epochs=epochs,
sample_weight=sample_weight,
verbose=2,
callbacks=[history] + callbacks)
# get history
hist = _format_keras_history(history)
# load and eval the best model
if eval_best:
mcp = [x for x in callbacks if isinstance(x, ModelCheckpoint)]
assert len(mcp) == 1
model = load_model(mcp[0].filepath)
return eval_model(model, valid, add_eval_metrics), hist
|
python
|
{
"resource": ""
}
|
q7081
|
eval_model
|
train
|
def eval_model(model, test, add_eval_metrics={}):
"""Evaluate model's performance on the test-set.
# Arguments
model: Keras model
test: test-dataset. Tuple of inputs `x` and target `y` - `(x, y)`.
add_eval_metrics: Additional evaluation metrics to use. Can be a dictionary or a list of functions
accepting arguments: `y_true`, `y_predicted`. Alternatively, you can provide names of functions from
the `concise.eval_metrics` module.
# Returns
dictionary with evaluation metrics
"""
# evaluate the model
logger.info("Evaluate...")
# - model_metrics
model_metrics_values = model.evaluate(test[0], test[1], verbose=0,
batch_size=test[1].shape[0])
# evaluation is done in a single pass to have more precise metics
model_metrics = dict(zip(_listify(model.metrics_names),
_listify(model_metrics_values)))
# - eval_metrics
y_true = test[1]
y_pred = model.predict(test[0], verbose=0)
eval_metrics = {k: v(y_true, y_pred) for k, v in add_eval_metrics.items()}
# handle the case where the two metrics names intersect
# - omit duplicates from eval_metrics
intersected_keys = set(model_metrics).intersection(set(eval_metrics))
if len(intersected_keys) > 0:
logger.warning("Some metric names intersect: {0}. Ignoring the add_eval_metrics ones".
format(intersected_keys))
eval_metrics = _delete_keys(eval_metrics, intersected_keys)
return merge_dicts(model_metrics, eval_metrics)
|
python
|
{
"resource": ""
}
|
q7082
|
get_model
|
train
|
def get_model(model_fn, train_data, param):
"""Feed model_fn with train_data and param
"""
model_param = merge_dicts({"train_data": train_data}, param["model"], param.get("shared", {}))
return model_fn(**model_param)
|
python
|
{
"resource": ""
}
|
q7083
|
_delete_keys
|
train
|
def _delete_keys(dct, keys):
"""Returns a copy of dct without `keys` keys
"""
c = deepcopy(dct)
assert isinstance(keys, list)
for k in keys:
c.pop(k)
return c
|
python
|
{
"resource": ""
}
|
q7084
|
_mean_dict
|
train
|
def _mean_dict(dict_list):
"""Compute the mean value across a list of dictionaries
"""
return {k: np.array([d[k] for d in dict_list]).mean()
for k in dict_list[0].keys()}
|
python
|
{
"resource": ""
}
|
q7085
|
CMongoTrials.get_trial
|
train
|
def get_trial(self, tid):
"""Retrieve trial by tid
"""
lid = np.where(np.array(self.tids) == tid)[0][0]
return self.trials[lid]
|
python
|
{
"resource": ""
}
|
q7086
|
CMongoTrials.delete_running
|
train
|
def delete_running(self, timeout_last_refresh=0, dry_run=False):
"""Delete jobs stalled in the running state for too long
timeout_last_refresh, int: number of seconds
"""
running_all = self.handle.jobs_running()
running_timeout = [job for job in running_all
if coarse_utcnow() > job["refresh_time"] +
timedelta(seconds=timeout_last_refresh)]
if len(running_timeout) == 0:
# Nothing to stop
self.refresh_tids(None)
return None
if dry_run:
logger.warning("Dry run. Not removing anything.")
logger.info("Removing {0}/{1} running jobs. # all jobs: {2} ".
format(len(running_timeout), len(running_all), len(self)))
now = coarse_utcnow()
logger.info("Current utc time: {0}".format(now))
logger.info("Time horizont: {0}".format(now - timedelta(seconds=timeout_last_refresh)))
for job in running_timeout:
logger.info("Removing job: ")
pjob = job.to_dict()
del pjob["misc"] # ignore misc when printing
logger.info(pprint.pformat(pjob))
if not dry_run:
self.handle.delete(job)
logger.info("Job deleted")
self.refresh_tids(None)
|
python
|
{
"resource": ""
}
|
q7087
|
CMongoTrials.train_history
|
train
|
def train_history(self, tid=None):
"""Get train history as pd.DataFrame
"""
def result2history(result):
if isinstance(result["history"], list):
return pd.concat([pd.DataFrame(hist["loss"]).assign(fold=i)
for i, hist in enumerate(result["history"])])
else:
return pd.DataFrame(result["history"]["loss"])
# use all
if tid is None:
tid = self.valid_tid()
res = [result2history(t["result"]).assign(tid=t["tid"]) for t in self.trials
if t["tid"] in _listify(tid)]
df = pd.concat(res)
# reorder columns
fold_name = ["fold"] if "fold" in df else []
df = _put_first(df, ["tid"] + fold_name + ["epoch"])
return df
|
python
|
{
"resource": ""
}
|
q7088
|
CMongoTrials.as_df
|
train
|
def as_df(self, ignore_vals=["history"], separator=".", verbose=True):
"""Return a pd.DataFrame view of the whole experiment
"""
def add_eval(res):
if "eval" not in res:
if isinstance(res["history"], list):
# take the average across all folds
eval_names = list(res["history"][0]["loss"].keys())
eval_metrics = np.array([[v[-1] for k, v in hist["loss"].items()]
for hist in res["history"]]).mean(axis=0).tolist()
res["eval"] = {eval_names[i]: eval_metrics[i] for i in range(len(eval_metrics))}
else:
res["eval"] = {k: v[-1] for k, v in res["history"]["loss"].items()}
return res
def add_n_epoch(df):
df_epoch = self.train_history().groupby("tid")["epoch"].max().reset_index()
df_epoch.rename(columns={"epoch": "n_epoch"}, inplace=True)
return pd.merge(df, df_epoch, on="tid", how="left")
results = self.get_ok_results(verbose=verbose)
rp = [_flatten_dict(_delete_keys(add_eval(x), ignore_vals), separator) for x in results]
df = pd.DataFrame.from_records(rp)
df = add_n_epoch(df)
first = ["tid", "loss", "status"]
return _put_first(df, first)
|
python
|
{
"resource": ""
}
|
q7089
|
effect_from_model
|
train
|
def effect_from_model(model, ref, ref_rc, alt, alt_rc, methods, mutation_positions, out_annotation_all_outputs,
extra_args=None, **argv):
"""Convenience function to execute multiple effect predictions in one call
# Arguments
model: Keras model
ref: Input sequence with the reference genotype in the mutation position
ref_rc: Reverse complement of the 'ref' argument
alt: Input sequence with the alternative genotype in the mutation position
alt_rc: Reverse complement of the 'alt' argument
methods: A list of prediction functions to be executed, e.g.: from concise.effects.ism.ism. Using the same
function more often than once (even with different parameters) will overwrite the results of the
previous calculation of that function.
mutation_positions: Position on which the mutation was placed in the forward sequences
out_annotation_all_outputs: Output labels of the model.
extra_args: None or a list of the same length as 'methods'. The elements of the list are dictionaries with
additional arguments that should be passed on to the respective functions in 'methods'. Arguments
defined here will overwrite arguments that are passed to all methods.
**argv: Additional arguments to be passed on to all methods, e.g,: out_annotation.
# Returns
Dictionary containing the results of the individual calculations, the keys are the
names of the executed functions
"""
assert isinstance(methods, list)
if isinstance(extra_args, list):
assert(len(extra_args) == len(methods))
else:
extra_args = [None] * len(methods)
main_args = {"model": model, "ref": ref, "ref_rc": ref_rc, "alt": alt, "alt_rc": alt_rc,
"mutation_positions": mutation_positions,
"out_annotation_all_outputs": out_annotation_all_outputs}
pred_results = {}
for method, xargs in zip(methods, extra_args):
if xargs is not None:
if isinstance(xargs, dict):
for k in argv:
if k not in xargs:
xargs[k] = argv[k]
else:
xargs = argv
for k in main_args:
xargs[k] = main_args[k]
res = method(**xargs)
pred_results[method.__name__] = res
return pred_results
|
python
|
{
"resource": ""
}
|
q7090
|
trades
|
train
|
def trades(ctx, market, limit, start, stop):
""" List trades in a market
"""
market = Market(market, bitshares_instance=ctx.bitshares)
t = [["time", "quote", "base", "price"]]
for trade in market.trades(limit, start=start, stop=stop):
t.append(
[
str(trade["time"]),
str(trade["quote"]),
str(trade["base"]),
"{:f} {}/{}".format(
trade["price"],
trade["base"]["asset"]["symbol"],
trade["quote"]["asset"]["symbol"],
),
]
)
print_table(t)
|
python
|
{
"resource": ""
}
|
q7091
|
ticker
|
train
|
def ticker(ctx, market):
""" Show ticker of a market
"""
market = Market(market, bitshares_instance=ctx.bitshares)
ticker = market.ticker()
t = [["key", "value"]]
for key in ticker:
t.append([key, str(ticker[key])])
print_table(t)
|
python
|
{
"resource": ""
}
|
q7092
|
cancel
|
train
|
def cancel(ctx, orders, account):
""" Cancel one or multiple orders
"""
print_tx(ctx.bitshares.cancel(orders, account=account))
|
python
|
{
"resource": ""
}
|
q7093
|
orderbook
|
train
|
def orderbook(ctx, market):
""" Show the orderbook of a particular market
"""
market = Market(market, bitshares_instance=ctx.bitshares)
orderbook = market.orderbook()
ta = {}
ta["bids"] = [["quote", "sum quote", "base", "sum base", "price"]]
cumsumquote = Amount(0, market["quote"])
cumsumbase = Amount(0, market["base"])
for order in orderbook["bids"]:
cumsumbase += order["base"]
cumsumquote += order["quote"]
ta["bids"].append(
[
str(order["quote"]),
str(cumsumquote),
str(order["base"]),
str(cumsumbase),
"{:f} {}/{}".format(
order["price"],
order["base"]["asset"]["symbol"],
order["quote"]["asset"]["symbol"],
),
]
)
ta["asks"] = [["price", "base", "sum base", "quote", "sum quote"]]
cumsumquote = Amount(0, market["quote"])
cumsumbase = Amount(0, market["base"])
for order in orderbook["asks"]:
cumsumbase += order["base"]
cumsumquote += order["quote"]
ta["asks"].append(
[
"{:f} {}/{}".format(
order["price"],
order["base"]["asset"]["symbol"],
order["quote"]["asset"]["symbol"],
),
str(order["base"]),
str(cumsumbase),
str(order["quote"]),
str(cumsumquote),
]
)
t = [["bids", "asks"]]
t.append([format_table(ta["bids"]), format_table(ta["asks"])])
print_table(t)
|
python
|
{
"resource": ""
}
|
q7094
|
buy
|
train
|
def buy(ctx, buy_amount, buy_asset, price, sell_asset, order_expiration, account):
""" Buy a specific asset at a certain rate against a base asset
"""
amount = Amount(buy_amount, buy_asset)
price = Price(
price, base=sell_asset, quote=buy_asset, bitshares_instance=ctx.bitshares
)
print_tx(
price.market.buy(price, amount, account=account, expiration=order_expiration)
)
|
python
|
{
"resource": ""
}
|
q7095
|
openorders
|
train
|
def openorders(ctx, account):
""" List open orders of an account
"""
account = Account(
account or config["default_account"], bitshares_instance=ctx.bitshares
)
t = [["Price", "Quote", "Base", "ID"]]
for o in account.openorders:
t.append(
[
"{:f} {}/{}".format(
o["price"],
o["base"]["asset"]["symbol"],
o["quote"]["asset"]["symbol"],
),
str(o["quote"]),
str(o["base"]),
o["id"],
]
)
print_table(t)
|
python
|
{
"resource": ""
}
|
q7096
|
cancelall
|
train
|
def cancelall(ctx, market, account):
""" Cancel all orders of an account in a market
"""
market = Market(market)
ctx.bitshares.bundle = True
market.cancel([x["id"] for x in market.accountopenorders(account)], account=account)
print_tx(ctx.bitshares.txbuffer.broadcast())
|
python
|
{
"resource": ""
}
|
q7097
|
spread
|
train
|
def spread(ctx, market, side, min, max, num, total, order_expiration, account):
""" Place multiple orders
\b
:param str market: Market pair quote:base (e.g. USD:BTS)
:param str side: ``buy`` or ``sell`` quote
:param float min: minimum price to place order at
:param float max: maximum price to place order at
:param int num: Number of orders to place
:param float total: Total amount of quote to use for all orders
:param int order_expiration: Number of seconds until the order expires from the books
"""
from tqdm import tqdm
from numpy import linspace
market = Market(market)
ctx.bitshares.bundle = True
if min < max:
space = linspace(min, max, num)
else:
space = linspace(max, min, num)
func = getattr(market, side)
for p in tqdm(space):
func(p, total / float(num), account=account, expiration=order_expiration)
print_tx(ctx.bitshares.txbuffer.broadcast())
|
python
|
{
"resource": ""
}
|
q7098
|
updateratio
|
train
|
def updateratio(ctx, symbol, ratio, account):
""" Update the collateral ratio of a call positions
"""
from bitshares.dex import Dex
dex = Dex(bitshares_instance=ctx.bitshares)
print_tx(dex.adjust_collateral_ratio(symbol, ratio, account=account))
|
python
|
{
"resource": ""
}
|
q7099
|
bidcollateral
|
train
|
def bidcollateral(
ctx, collateral_symbol, collateral_amount, debt_symbol, debt_amount, account
):
""" Bid for collateral in the settlement fund
"""
print_tx(
ctx.bitshares.bid_collateral(
Amount(collateral_amount, collateral_symbol),
Amount(debt_amount, debt_symbol),
account=account,
)
)
|
python
|
{
"resource": ""
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.