Merge pull request #1 from Yue-Jiang/python3

try to make ZIFA python3 compatible
epierson9 · May 28, 2016 · eeeb170 · eeeb170
2 parents 627f779 3308177
commit eeeb170
Show file tree

Hide file tree

Showing 3 changed files with 17 additions and 17 deletions.
diff --git a/ZIFA/ZIFA.py b/ZIFA/ZIFA.py
@@ -360,9  360,9 @@ def fitModel(Y, K, singleSigma = False):
 	"""
 	N, D = Y.shape
 	if D > 2000:
-		print 'Warning: this dataset has a large number of genes. If ZIFA takes too long to run, try using block_ZIFA.py instead'
 		print('Warning: this dataset has a large number of genes. If ZIFA takes too long to run, try using block_ZIFA.py instead')
 	testInputData(Y)
-	print 'Running zero-inflated factor analysis with N = %i, D = %i, K = %i' % (N, D, K)
 	print('Running zero-inflated factor analysis with N = %i, D = %i, K = %i' % (N, D, K))
 	#initialize the parameters
 	np.random.seed(23)
 	A, mus, sigmas, decay_coef = initializeParams(Y, K, singleSigma = singleSigma)
@@ -388,10  388,10 @@ def fitModel(Y, K, singleSigma = False):
 		sigmas = new_sigmas
 		decay_coef = new_decay_coef
 		if paramsNotChanging:
-			print 'Param change below threshold %2.3e after %i iterations' % (param_change_thresh, n_iter)
 			print('Param change below threshold %2.3e after %i iterations' % (param_change_thresh, n_iter))
 			break
 		if n_iter >= max_iter:
-			print 'Maximum number of iterations reached; terminating loop'
 			print('Maximum number of iterations reached; terminating loop')
 		n_iter  = 1	
 	EZ, EZZT, EX, EXZ, EX2 = Estep(Y, A, mus, sigmas, decay_coef)
 	params = {'A':A, 'mus':mus, 'sigmas':sigmas, 'decay_coef':decay_coef}

diff --git a/ZIFA/block_ZIFA.py b/ZIFA/block_ZIFA.py
@@ -355,7  355,7 @@ def generateIndices(n_blocks, N, D):
 	y_indices_to_use[i][j] is the indices of block j in sample i. 
 	"""
 	y_indices_to_use = []
-	idxs = range(D)
 	idxs = list(range(D))
 	n_in_block = int(1.*D / n_blocks)
 	for i in range(N):
 		partition = []
@@ -372,7  372,7 @@ def generateIndices(n_blocks, N, D):
 			n_added  = len(idxs_in_block)
 		y_indices_to_use.append(partition)
 		if i == 0:
-			print 'Block sizes', [len(a) for a in partition]
 			print('Block sizes', [len(a) for a in partition])
 		assert(n_added == D)
 	return y_indices_to_use
 
@@ -419,20  419,20 @@ def testInputData(Y):
 	zero_fracs = Y_is_zero.mean(axis = 0)
 	column_is_all_zero = zero_fracs == 1.
 	if column_is_all_zero.sum() > 0:
-		print "Warning: Your Y matrix has %i columns which are entirely zero; filtering these out before continuing." % (column_is_all_zero.sum())
 		print("Warning: Your Y matrix has %i columns which are entirely zero; filtering these out before continuing." % (column_is_all_zero.sum()))
 		Y = Y[:, ~column_is_all_zero]
 	elif (zero_fracs > .9).sum() > 0:
-		print 'Warning: your Y matrix contains genes which are frequently zero. If the algorithm fails to converge, try filtering out genes which are zero more than 80 - 90% of the time, or using standard ZIFA.'
 		print('Warning: your Y matrix contains genes which are frequently zero. If the algorithm fails to converge, try filtering out genes which are zero more than 80 - 90% of the time, or using standard ZIFA.')
 	return Y
 
 def runEMAlgorithm(Y, K, singleSigma = False, n_blocks = None):
 	Y = testInputData(Y)
 	N, D = Y.shape
 	if n_blocks is None:
 		n_blocks = max(1, D / 500)
-		print 'Number of blocks has been set to', n_blocks
 		print('Number of blocks has been set to', n_blocks)
 
-	print 'Running block zero-inflated factor analysis with N = %i, D = %i, K = %i, n_blocks = %i' % (N, D, K, n_blocks)
 	print('Running block zero-inflated factor analysis with N = %i, D = %i, K = %i, n_blocks = %i' % (N, D, K, n_blocks))
 	#generate blocks. 
 	y_indices_to_use = generateIndices(n_blocks, N, D) 
 
@@ -473,7  473,7 @@ def runEMAlgorithm(Y, K, singleSigma = False, n_blocks = None):
 		try:
 			checkNoNans([EZ, EZZT, EX, EXZ, EX2, new_A, new_mus, new_sigmas, new_decay_coef])
 		except:
-			print "Error: algorithm failed to converge. Usual solutions to this problem: filtering out genes which are zero more than 80 - 90% of the time, or using standard ZIFA. Automatically retrying ZIFA when filtering out genes."
 			print("Error: algorithm failed to converge. Usual solutions to this problem: filtering out genes which are zero more than 80 - 90% of the time, or using standard ZIFA. Automatically retrying ZIFA when filtering out genes.")
 			return None
 
 		paramsNotChanging = True
@@ -489,10  489,10 @@ def runEMAlgorithm(Y, K, singleSigma = False, n_blocks = None):
 		sigmas = new_sigmas
 		decay_coef = new_decay_coef
 		if paramsNotChanging:
-			print 'Param change below threshold %2.3e after %i iterations' % (param_change_thresh, n_iter)
 			print('Param change below threshold %2.3e after %i iterations' % (param_change_thresh, n_iter))
 			break
 		if n_iter >= max_iter:
-			print 'Maximum number of iterations reached; terminating loop'
 			print('Maximum number of iterations reached; terminating loop')
 		n_iter  = 1	
 	params = {'A':A, 'mus':mus, 'sigmas':sigmas, 'decay_coef':decay_coef, 'X':EX}
 	return EZ, params
@@ -509,17  509,17 @@ def fitModel(Y, K, singleSigma = False, n_blocks = None, p0_thresh = .95):
 	params: a dictionary of model parameters. Throughout, we refer to lambda as "decay_coef". 
 	"""
 	assert(p0_thresh >= 0 and p0_thresh <= 1)
-	print 'Filtering out all genes which are zero in more than %2.1f%% of samples. To change this, change p0_thresh.' % (p0_thresh * 100)
 	print('Filtering out all genes which are zero in more than %2.1f%% of samples. To change this, change p0_thresh.' % (p0_thresh * 100))
 	Y = Y[:, (np.abs(Y) < 1e-6).mean(axis = 0) <= p0_thresh]
 	results = runEMAlgorithm(Y, K, singleSigma = singleSigma, n_blocks = n_blocks)
 
 	while results is None:
 		Y_is_zero = np.abs(Y) < 1e-6
 		max_zero_frac = Y_is_zero.mean(axis = 0).max()
 		new_max_zero_frac = max_zero_frac * .95
-		print 'Previously, maximum fraction of zeros for a gene was %2.3f; now lowering that to %2.3f and rerunning ZIFA' % (max_zero_frac, new_max_zero_frac)
 		print('Previously, maximum fraction of zeros for a gene was %2.3f; now lowering that to %2.3f and rerunning ZIFA' % (max_zero_frac, new_max_zero_frac))
 		Y = Y[:, Y_is_zero.mean(axis = 0) < new_max_zero_frac]
-		print 'After filtering out genes with too many zeros, %i samples and %i genes' % Y.shape
 		print('After filtering out genes with too many zeros, %i samples and %i genes' % Y.shape)
 		results = runEMAlgorithm(Y, K, singleSigma = singleSigma, n_blocks = n_blocks)
 	return results
 

diff --git a/example.py b/example.py
@@ -38,7  38,7 @@ def generateSimulatedDimensionalityReductionData(n_clusters, n, d, k, sigma, dec
 	cutoff = np.exp(-decay_coef * (Y ** 2))
 	zero_mask = rand_matrix < cutoff
 	Y[zero_mask] = 0
-	print 'Fraction of zeros: %2.3f; decay coef: %2.3f' % ((Y == 0).mean(), decay_coef)
 	print('Fraction of zeros: %2.3f; decay coef: %2.3f' % ((Y == 0).mean(), decay_coef))
 
 	return X, Y, Z.transpose(), cluster_ids