Add files via upload

UnionCom.py

@@ -34,7 +34,6 @@ class params():
 	col = []
 	row = []
 	output_dim = 32
-
 
 def fit_transform(dataset, datatype=None, epoch_pd=20000, epoch_DNN=200, \
 	epsilon=0.001, lr=0.001, batch_size=100, rho=10, beta=1,\
@@ -131,7 +130,6 @@ def fit_transform(dataset, datatype=None, epoch_pd=20000, epoch_DNN=200, \
 	else:
 		integrated_data = project_barycentric(dataset, match_result)	
 
-
 	print("---------------------------------")
 	print("unionCom Done!")
 	time2 = time.time()
@@ -141,3 +139,24 @@ def fit_transform(dataset, datatype=None, epoch_pd=20000, epoch_DNN=200, \
 			test_UnionCom(integrated_data, datatype, params, device, test)
 
 	return integrated_data
+
+def PCA_visualize(data, integrated_data, datatype=None):
+
+	if datatype is not None:
+		visualize(data, integrated_data, datatype)
+	else:
+		visualize(data, integrated_data)
+
+def test_label_transfer_accuracy(integrated_data, datatype):
+
+	test_UnionCom(integrated_data, datatype)
+
+
+
+
+
+
+
+
+
+

test.py

@@ -1,13 +1,10 @@
 import os
 import random
-import torch
-import torch.backends.cudnn as cudnn
-from torch.autograd import Variable
 import numpy as np
 import scipy.sparse as sp 
 from sklearn.neighbors import NearestNeighbors, KNeighborsClassifier
 
-def align_fraction(data1, data2, params):
+def align_fraction(data1, data2):
 	row1, col1 = np.shape(data1)
 	row2, col2 = np.shape(data2)
 	fraction = 0
@@ -34,13 +31,13 @@ def transfer_accuracy(domain1, domain2, type1, type2):
 			count += 1
 	return count / len(type1)
 
-def test_UnionCom(integrated_data, datatype, params, device, test):
+def test_UnionCom(integrated_data, datatype):
 
 	for i in range(len(integrated_data)-1):
-		# fraction = align_fraction(data[i], data[-1], params)
+		# fraction = align_fraction(data[i], data[-1])
 		# print("average fraction:")
 		# print(fraction)
 
 		acc = transfer_accuracy(integrated_data[i], integrated_data[-1], datatype[i], datatype[-1])
-		print("label transfer accuracy:")
+		print("label transfer accuracy of data{:d}:".format(i+1))
 		print(acc)

version.md

@@ -4,6 +4,10 @@
 
 ## version 0.2.0
 + Software optimization.
-+ Split function "train" into functions "Match" and "Project".
-+ Use Kuhn-Munkres algorithm to find optimal pairs between datasets instead of parbabilistic matrix matching.
++ Split function train into functions Match and Project.
++ Use Kuhn-Munkres to find optimal pairs between datasets instead of parbabilistic matrix matching.
 + Add a new parameter "project" to provide options for barycentric projection.
+
+## version 0.2.1 
++ Separate "test_label_transfer_accuracy" function from "fit_transform" function
++ fix some bugs

visualization.py

@@ -6,44 +6,36 @@
 def visualize(data, data_integrated, datatype=None):
 
     dataset_num = len(data)
-    font_label = {
-             'weight': 'normal',
-             'size': 22,
-         }
-
-    font_title = {
-             'weight': 'normal',
-             'size': 25,
-         }
 
     styles = ['g', 'r', 'b', 'y', 'k', 'm', 'c'] 
 
     embedding = []
     dataset_xyz = []
     for i in range(dataset_num):
-        dataset_xyz.append("dataset{:d}".format(i+1))
+        dataset_xyz.append("data{:d}".format(i+1))
         embedding.append(PCA(n_components=2).fit_transform(data[i]))
 
+    fig = plt.figure()
     if datatype is not None:
         for i in range(dataset_num):
-            fig = plt.figure()
+            plt.subplot(1,dataset_num,i+1)
             for j in set(datatype[i]):
                 index = np.where(datatype[i]==j) 
                 plt.scatter(embedding[i][index,0], embedding[i][index,1], c=styles[j], s=5.)
-            plt.title(dataset_xyz[i], font_title)
-            plt.xlabel('Dimension-1', font_label)
-            plt.ylabel('Dimension-2', font_label)
+            plt.title(dataset_xyz[i])
+            plt.xlabel('PCA-1')
+            plt.ylabel('PCA-2')
             plt.legend()
     else:
         for i in range(dataset_num):
-            fig = plt.figure()
+            plt.subplot(1,dataset_num,i+1)
             plt.scatter(embedding[i][:,0], embedding[i][:,1],c=styles[i], s=5.)
-            plt.title(dataset_xyz[i], font_title)
-            plt.xlabel('Dimension-1', font_label)
-            plt.ylabel('Dimension-2', font_label)
+            plt.title(dataset_xyz[i])
+            plt.xlabel('PCA-1')
+            plt.ylabel('PCA-2')
             plt.legend()
 
-
+    plt.tight_layout()
 
     data_all = np.vstack((data_integrated[0], data_integrated[1]))
     for i in range(2, dataset_num):
@@ -64,31 +56,38 @@ def visualize(data, data_integrated, datatype=None):
     color = [[1,0.5,0], [0.2,0.4,0.1], [0.1,0.2,0.8], [0.5, 1, 0.5], [0.1, 0.8, 0.2]]
     # marker=['x','^','o','*','v']
 
+    fig = plt.figure()
     if datatype is not None:
-        fig = plt.figure()
+
+        plt.subplot(1,2,1)
         for i in range(dataset_num):
-            plt.scatter(embedding[i][:,0], embedding[i][:,1], c=color[i], s=5., alpha=0.8)
-        plt.title('Embeddings', font_title)
-        plt.xlabel('Dimension-1', font_label)
-        plt.ylabel('Dimension-2', font_label)
+            plt.scatter(embedding[i][:,0], embedding[i][:,1], c=color[i], label='data{:d}'.format(i+1), s=5., alpha=0.8)
+        plt.title('Integrated Embeddings')
+        plt.xlabel('PCA-1')
+        plt.ylabel('PCA-2')
         plt.legend()
 
-        fig = plt.figure()
+        plt.subplot(1,2,2)
         for i in range(dataset_num):  
             for j in set(datatype[i]):
                 index = np.where(datatype[i]==j) 
-                plt.scatter(embedding[i][index,0], embedding[i][index,1], c=styles[j], s=5., alpha=0.8)
-        plt.title('Cell Types', font_title)
-        plt.xlabel('Dimension-1', font_label)
-        plt.ylabel('Dimension-2', font_label)
+                if i < dataset_num-1:
+                    plt.scatter(embedding[i][index,0], embedding[i][index,1], c=styles[j], s=5., alpha=0.8)
+                else:
+                    plt.scatter(embedding[i][index,0], embedding[i][index,1], c=styles[j], s=5., alpha=0.8)  
+        plt.title('Integrated Cell Types')
+        plt.xlabel('PCA-1')
+        plt.ylabel('PCA-2')
         plt.legend()
+
     else:
-        fig = plt.figure()
+
         for i in range(dataset_num):
-            plt.scatter(embedding[i][:,0], embedding[i][:,1], c=color[i], s=5., alpha=0.8)
-        plt.title('Embeddings', font_title)
-        plt.xlabel('Dimension-1', font_label)
-        plt.ylabel('Dimension-2', font_label)
+            plt.scatter(embedding[i][:,0], embedding[i][:,1], c=styles[i], label='data{:d}'.format(i+1), s=5., alpha=0.8)
+        plt.title('Integrated Embeddings')
+        plt.xlabel('PCA-1')
+        plt.ylabel('PCA-2')
         plt.legend()
 
+    plt.tight_layout()
     plt.show()