Merge pull request #46 from abarton51/musicnet_preprocessing

tsne plots and code

src/musicNet/dim_reduction.py

@@ -0,0 +1,42 @@
+import numpy as np
+import matplotlib.pyplot as plt
+import os
+import sys
+from sklearn.decomposition import PCA
+from sklearn.manifold import TSNE
+
+
+fig_path = 'src/musicNet/figs/tsne_plots'
+data_path = 'src/musicNet/processed_data'
+X_train = np.load(data_path + '/train_data_midi.npy')
+X_test = np.load(data_path + '/test_data_midi.npy')
+y_train = np.load(data_path + '/train_labels_midi.npy')
+y_test = np.load(data_path + '/test_labels_midi.npy')
+
+X_full = np.concatenate([X_train, X_test]).astype('int')
+y_full = np.concatenate([y_train, y_test]).astype('int')
+
+# 0 -> 0 - Bach
+# 1 -> 1 - Beethoven
+# 2 -> 2 - Brahms
+# 7 -> 3 - Mozart
+# 9 -> 4 - Schubert
+
+y_full[y_full==7] = 3
+y_full[y_full==9] = 4
+
+tsne = TSNE(n_components=3)
+X_embedded = tsne.fit_transform(X_full)
+colors = ['dodgerblue', 'blueviolet', 'firebrick', 'darkslategray', 'forestgreen']
+
+def tsne_plot(embd_data, labels, colors=None):
+    classes = np.unique(labels)
+    dim = embd_data.shape[1]
+    fig = plt.figure()
+    ax = fig.add_subplot(projection='3d')
+    for i in classes:
+        embd_data_i = embd_data[labels==i]
+        ax.scatter(embd_data_i[:,0], embd_data_i[:,1], embd_data_i[:,2], color=colors[i])
+    plt.savefig(fig_path + '/tsne_plot_' + str(dim) + 'd')
+
+tsne_plot(X_embedded, y_full, colors)

src/musicNet/main.py

@@ -10,6 +10,17 @@
 y_train = np.load(path + '/train_labels_midi.npy')
 y_test = np.load(path + '/test_labels_midi.npy')
 
+# 0 -> 0 - Bach
+# 1 -> 1 - Beethoven
+# 2 -> 2 - Brahms
+# 7 -> 3 - Mozart
+# 9 -> 4 - Schubert
+
+y_train[y_train==7] = 3
+y_test[y_test==7] = 3
+y_train[y_train==9] = 4
+y_test[y_test==9] = 4
+
 dt_clf = DecisionTreeClassifier(random_state=42)
 
 dt_clf.fit(X_train, y_train)