Fix #16, #17

And apply updated black formatting

mofapy2/build_model/build_model.py

@@ -149,7 +149,6 @@ def build_ThetaZ(self):
         # initTheta_qE = 1.
 
         self.init_model.initThetaZ(
-            self.data_opts["samples_groups"],
             qa=initTheta_a,
             qb=initTheta_b,
             qE=initTheta_qE,

mofapy2/build_model/init_model.py

@@ -98,7 +98,6 @@ def initZ(
         # mean
         if qmean is not None:
             if isinstance(qmean, str):
-
                 # Random initialisation
                 if qmean == "random":
                     qmean = stats.norm.rvs(loc=0, scale=1, size=(self.N, self.K))
@@ -199,7 +198,6 @@ def initZ_smooth(
         # mean
         if qmean is not None:
             if isinstance(qmean, str):
-
                 # Random initialisation
                 if qmean == "random":
                     qmean = stats.norm.rvs(loc=0, scale=1, size=(self.N, self.K))
@@ -350,7 +348,6 @@ def initZgU(
         # mean
         if qmean is not None:
             if isinstance(qmean, str):
-
                 # Random initialisation
                 if qmean == "random":
                     qmean = stats.norm.rvs(loc=0, scale=1, size=(self.N, self.K))
@@ -415,7 +412,6 @@ def initSigma(
         model_groups=False,
         use_gpytorch=False,
     ):
-
         self.Sigma = Sigma_Node(
             dim=(self.K,),
             sample_cov=sample_cov,
@@ -437,7 +433,6 @@ def initSigma_sparse(
         model_groups=False,
         use_gpytorch=False,
     ):
-
         self.Sigma = Sigma_Node_sparse(
             dim=(self.K,),
             sample_cov=sample_cov,
@@ -464,7 +459,6 @@ def initSigma_warping(
         model_groups=False,
         use_gpytorch=False,
     ):
-
         self.Sigma = Sigma_Node_warping(
             dim=(self.K,),
             sample_cov=sample_cov,
@@ -514,7 +508,6 @@ def initSZ(
 
         ## Initialise variational distribution (Q)
         if isinstance(qmean_T1, str):
-
             if qmean_T1 == "random":
                 qmean_T1 = stats.norm.rvs(loc=0, scale=1, size=(self.N, self.K))
             elif qmean_T1 == "pca":
@@ -579,7 +572,6 @@ def initW(
         W_list = [None] * self.M
 
         for m in range(self.M):
-
             ## Initialise prior distribution (P) ##
 
             # mean
@@ -593,7 +585,6 @@ def initW(
             # mean
             if qmean is not None:
                 if isinstance(qmean, str):
-
                     # Random initialisation
                     if qmean == "random":
                         qmean_m = stats.norm.rvs(
@@ -665,12 +656,10 @@ def initSW(
         W_list = [None] * self.M
 
         for m in range(self.M):
-
             ## Initialise prior distribution (P)
 
             ## Initialise variational distribution (Q)
             if isinstance(qmean_S1, str):
-
                 if qmean_S1 == "random":
                     qmean_S1_tmp = stats.norm.rvs(
                         loc=0, scale=1.0, size=(self.D[m], self.K)
@@ -800,7 +789,6 @@ def initTau(self, pa=1e-3, pb=1e-3, qa=1.0, qb=1.0, qE=None):
         tau_list = [None] * self.M
 
         for m in range(self.M):
-
             # Poisson noise model for count data
             if self.lik[m] == "poisson":
                 tmp = 0.25 + 0.17 * np.nanmax(self.data[m], axis=0)

mofapy2/build_model/save_model.py

@@ -55,7 +55,6 @@ def __init__(
         sort_factors=True,
         compression_level=9,
     ):
-
         # Check that the model is trained
         # NOTE: it might be not trained if saving when training is interrupted
         if not model.trained:
@@ -265,7 +264,6 @@ def saveData(self):
             data_subgrp = data_grp.create_group(self.views_names[m])
             intercept_subgrp = intercept_grp.create_group(self.views_names[m])
             for g in range(len(self.groups_names)):
-
                 # Subset group
                 samples_idx = np.where(
                     np.array(self.samples_groups) == self.groups_names[g]
@@ -332,7 +330,6 @@ def saveImputedData(self, mean, variance):
         for m in range(len(mean)):
             view_subgrp = data_grp.create_group(self.views_names[m])
             for g in range(len(self.groups_names)):
-
                 # Subset group
                 samples_idx = np.where(
                     np.array(self.samples_groups) == self.groups_names[g]
@@ -391,7 +388,6 @@ def saveZpredictions(self, mean, variance, values, groups):
                 )
 
     def saveExpectations(self, nodes="all"):
-
         # Get nodes from the model
         nodes_dic = self.model.getNodes()
         if type(nodes) is str:
@@ -435,15 +431,12 @@ def saveExpectations(self, nodes="all"):
             # Multi-view nodes
             if isinstance(nodes_dic[n], Multiview_Node):
                 for m in range(nodes_dic[n].M):
-
                     # Multi-groups nodes (Tau, Y, and Z)
                     if n in multigroup_nodes:
-
                         # Create subgroup for the view
                         view_subgrp = node_subgrp.create_group(self.views_names[m])
 
                         for g in self.groups_names:
-
                             # Add missing values to Tau and Y nodes
                             exp[m][self.mask[m]] = np.nan
 
@@ -471,7 +464,6 @@ def saveExpectations(self, nodes="all"):
 
             # Single-view nodes
             else:
-
                 # Multi-group nodes (Z)
                 if n in multigroup_nodes:
                     for g in self.groups_names:
@@ -548,13 +540,11 @@ def saveParameters(self, nodes="all"):
             # Multi-view nodes
             if isinstance(nodes_dic[n], Multiview_Node):
                 for m in range(nodes_dic[n].M):
-
                     # Create subgroup for the view
                     view_subgrp = node_subgrp.create_group(self.views_names[m])
 
                     # Multi-groups nodes
                     if n in multigroup_nodes:
-
                         for g in self.groups_names:
                             grp_subgrp = view_subgrp.create_group(g)
 
@@ -586,7 +576,6 @@ def saveParameters(self, nodes="all"):
 
             # Single-view nodes
             else:
-
                 # Multi-group nodes
                 if n in multigroup_nodes:
                     for g in self.groups_names:
@@ -615,7 +604,6 @@ def saveParameters(self, nodes="all"):
         pass
 
     def saveModelOptions(self):
-
         # Subset model options
         options_to_save = [
             "likelihoods",
@@ -717,7 +705,6 @@ def saveSmoothOptions(self, smooth_opts):
         grp[k].attrs["names"] = np.asarray(list(opts.keys())).astype("S")
 
     def saveVarianceExplained(self):
-
         # Sort values by alphabetical order of views
         # order = np.argsort(self.views_names)
         # # order = [ i[0] for i in sorted(enumerate(self.views_names), key=lambda x:x[1]) ]

mofapy2/build_model/train_model.py

@@ -10,7 +10,6 @@
 
 
 def train_model(model):
-
     # Sanity check on the Bayesian Network
     assert isinstance(model, BayesNet), "'model' has to be a BayesNet class"
 

mofapy2/build_model/utils.py

@@ -53,9 +53,7 @@ def gaussianise(Y_m, axis=0):
 
 
 def process_data(data, likelihoods, data_opts, samples_groups):
-
     for m in range(len(data)):
-
         # For some wierd reason, when using reticulate from R, missing values are stored as -2147483648
         data[m][data[m] == -2147483648] = np.nan
 
@@ -79,7 +77,6 @@ def process_data(data, likelihoods, data_opts, samples_groups):
 
         # Centering and scaling is only appropriate for gaussian data
         if likelihoods[m] in ["gaussian"]:
-
             # Center features per group
             if data_opts["center_groups"]:
                 for g in data_opts["groups_names"]:

mofapy2/core/BayesNet.py

@@ -133,7 +133,6 @@ def getNodes(self):
         return self.nodes
 
     def calculate_variance_explained(self, total=False):
-
         # Collect relevant expectations
         Z = self.nodes["Z"].getExpectation()
         W = self.nodes["W"].getExpectation()
@@ -384,7 +383,7 @@ def iterate(self):
                 "time": iter_time,
                 "number_factors": number_factors,
                 "elbo": elbo["total"].values,
-                "elbo_terms": elbo.drop("total", 1),
+                "elbo_terms": elbo.drop("total", axis=1),
             }
             if "Sigma" in self.nodes.keys():
                 tmp = self.nodes["Sigma"].getParameters()  # save only last iteration

mofapy2/core/distributions/basic_distributions.py

@@ -19,6 +19,7 @@
 # from mofapy2.config import settings
 # from mofapy2.core.utils import *  # TODO prob not necessary ?
 
+
 # General class for probability distributions
 class Distribution(object):
     """General class for a statistical distribution"""

mofapy2/core/distributions/bernoulli_gaussian.py

@@ -107,7 +107,6 @@ def updateExpectations(self):
         # self.expectations = {'E':E, 'EB':EB, 'EN':EN, 'E2':E2, 'ENN':ENN, 'EXXT':EXXT }
 
     def removeDimensions(self, axis, idx):
-
         # Method to remove undesired dimensions
         # - axis (int): axis from where to remove the elements
         # - idx (numpy array): indices of the elements to remove

mofapy2/core/gpu_utils.py

@@ -11,6 +11,7 @@
 # --------------------------------------------------------
 gpu_mode = False
 
+
 # --------------------------------------------------------
 # operations on single matrices
 # --------------------------------------------------------

mofapy2/core/nodes/Kc_node.py

@@ -80,7 +80,6 @@ def compute_kernel(self, spectral_decomp=True):
             self.compute_kernel_at_gridpoint(i, spectral_decomp=spectral_decomp)
 
     def compute_kernel_at_gridpoint(self, i, spectral_decomp=True):
-
         # build kernel matrix based on given covariance function
         if not spectral_decomp:
             self.Kmat[i, :, :] = SE(self.covariates, self.l_grid[i], zeta=0)
@@ -118,7 +117,6 @@ def set_gridix(self, lidx, k):
         # no recomputation required as stored on grid
 
     def eval_at_newpoints_k(self, new_cov, k):
-
         Kc_new = SE(new_cov, self.l_grid[self.gridix[k]], zeta=0)
 
         return Kc_new

mofapy2/core/nodes/Kg_node.py

@@ -6,6 +6,7 @@
 from mofapy2.core.nodes.variational_nodes import *
 from mofapy2.core.gp_utils import *
 
+
 # TODO: for large number of groups avoid constructing Kmat if spectral decomp and only save x at the end (getParameters in SigmaNode)
 class Kg_Node(Node):
     """

mofapy2/core/nodes/Sigma_node.py

@@ -345,7 +345,6 @@ def removeFactors(self, idx, axis=1):
         self.Sigma_inv_logdet = np.delete(self.Sigma_inv_logdet, axis=0, obj=idx)
 
     def calc_neg_elbo_k(self, par, lidx, k, var):
-
         self.zeta[k] = par[0]
         self.Kc.set_gridix(lidx, k)
 
@@ -624,7 +623,6 @@ def optimise(self, var):
 
         # optimise hyperparamters of GP
         for k in range(K):
-
             best_zeta = -1
             best_elbo = -np.Inf
 
@@ -751,7 +749,6 @@ def __init__(
         rankx=None,
         model_groups=False,
     ):
-
         super().__init__(
             dim, sample_cov, groups, start_opt, opt_freq, n_grid, rankx, model_groups
         )
@@ -812,7 +809,6 @@ def __init__(
         model_groups=False,
         idx_inducing=None,
     ):
-
         super().__init__(
             dim, sample_cov, groups, start_opt, opt_freq, n_grid, rankx, model_groups
         )
@@ -946,7 +942,6 @@ def __init__(
         warping_open_end=True,
         warping_groups=None,
     ):
-
         super().__init__(
             dim, sample_cov, groups, start_opt, opt_freq, n_grid, rankx, model_groups
         )

mofapy2/core/nodes/Theta_nodes.py

@@ -66,7 +66,6 @@ def _updateParameters(self):
         self.Q.setParameters(a=Qa, b=Qb)
 
     def calculateELBO(self):
-
         # Collect parameters and expectations
         Qpar, Qexp = self.Q.getParameters(), self.Q.getExpectations()
         Pa, Pb, Qa, Qb = self.Ppar["a"], self.Ppar["b"], Qpar["a"], Qpar["b"]
@@ -91,7 +90,6 @@ class ThetaZ_Node(Beta_Unobserved_Variational_Node):
     """
 
     def __init__(self, dim, pa, pb, qa, qb, groups, qE=None):
-
         self.groups = groups
         self.factors_axis = 1
         self.N = len(self.groups)
@@ -159,7 +157,6 @@ def updateParameters(self, ix=None, ro=1.0):
         self.Q.setParameters(a=Qa, b=Qb)
 
     def _updateParameters(self, S, groups, ro):
-
         Q = self.Q.getParameters()
         Qa, Qb = Q["a"], Q["b"]
         Qa *= 1 - ro
@@ -186,7 +183,6 @@ def _updateParameters(self, S, groups, ro):
         return Qa, Qb
 
     def calculateELBO(self):
-
         # Collect parameters and expectations
         Qpar, Qexp = self.Q.getParameters(), self.Q.getExpectations()
         Pa, Pb, Qa, Qb = self.Ppar["a"], self.Ppar["b"], Qpar["a"], Qpar["b"]

mofapy2/core/nodes/U_nodes.py

@@ -10,6 +10,7 @@
 # TODO:
 # - integrate into Z node using ix
 
+
 # U_GP_Node_mv
 class U_GP_Node_mv(MultivariateGaussian_Unobserved_Variational_Node):
     """
@@ -64,7 +65,6 @@ def get_mini_batch(self):
             return self.mini_batch
 
     def updateParameters(self, ix=None, ro=1.0):
-
         # Get expectations from other nodes
         W = self.markov_blanket["W"].getExpectations()
         Y = self.markov_blanket["Y"].get_mini_batch()