K-means recovers ICA filters when independent components are sparse

Keyphrases

• K-means 100%
• Analysis Filter 100%
• International Cooperative Ataxia Rating Scale (ICARS) 100%
• Unsupervised Feature Learning 42%
• Feature Learning 28%
• Clustering Algorithm 14%
• Learning Methods 14%
• Closest Connection 14%
• Simple Features 14%
• Linear Classifier 14%
• Learning Approaches 14%
• Mixing Matrix 14%
• Neural Network Training 14%
• K-means Clustering Algorithm 14%
• Unsupervised Pre-training 14%

Computer Science

• Independent Component Analysis 100%
• Independent Component 100%
• Analysis Filter 100%
• Representation Learning 83%
• Clustering Algorithm 16%
• Neural Network 16%
• Linear Classifier 16%
• Learning Approach 16%
• Unlabeled Example 16%
• k-means Clustering 16%