Projects

Project description

Scientific investigation advances brain-like intelligence. Inspired by the discovery in neuroscience, this project proposes a sparse binary projection model for general intelligent information processing tasks. 

We will establish a unified mathematical theory to analyze the model, develop efficient and effective machine learning algorithms to seek high-quality projection methods, and, focusing on practical tasks, design a general data representation scheme based on sparse binary vectors.

We will study the effect of competitive learning in the proposed sparse binary projection model.

We will apply the proposed models and algorithms in empirical applications, specifically in natural language processing and image processing applications.

Key contribution

Our proposed study aims at achieving novel theories, influential algorithms, and successful applications.

We found and proved the random sparse binary projection theorem which lays the foundation of the study.

We designed the sparse lifting algorithm for data representation tasks and reported improved results in information retrieval and sentiment analysis applications.

Next steps

Make the package and code open-source.

Investigate new applications for the proposed model.

Team/contributors

LI Wenye, MA Changyi, YU Fangchen, YU Yueyao