Add two parallel implementations of EP with numpy and pytorch

EnsemblePursuitNumpy.py

@@ -0,0 +1,20 @@
+import numpy as np
+
+
+class EnsemblePursuitNumpy():
+    def __init__(self,n_ensembles,lambd,options_dict):
+        self.n_ensembles=n_ensembles
+        self.lambd=lambd
+        self.options_dict=options_dict
+
+
+    def zscore(self,X):
+        mean_stimuli=np.mean(X,axis=1)[...,np.newaxis]
+        std_stimuli=np.std(X,axis=1)[...,np.newaxis]+0.0000000001
+        X=np.subtract(X,mean_stimuli)
+        X=np.divide(X,std_stimuli)
+        return X
+
+    def fit_transform(self,X):
+        X=self.zscore(X)
+        print(X)

EnsemblePursuitPyTorch.py

@@ -0,0 +1,28 @@
+import torch
+
+class EnsemblePursuitPyTorch():
+    def __init__(self, n_ensembles, lambd, options_dict):
+        self.n_ensembles=n_ensembles
+        self.lambd=lambd
+        self.options_dict=options_dict
+
+    def zscore(self,X):
+        #Have to transpose X to make torch.sub and div work. Transpose back into 
+        #original shape when done with calculations. 
+        mean_stimuli=X.t().mean(dim=0)
+        std_stimuli=X.t().std(dim=0)+0.0000000001
+
+        X=torch.sub(X.t(),mean_stimuli)
+        X=X.div(std_stimuli)
+        return X.t()
+
+    def fit_transform(self,X):
+        '''
+             X-- shape (neurons, timepoints)
+        '''
+        X=torch.cuda.FloatTensor(X)
+        X=self.zscore(X)
+        print(X)
+
+
+