Gathering & Poster Session
Prof. David Malah, Head of SIPL
Nonparametric Canonical Correlation Analysis
Prof. Tomer Michaeli
Canonical correlation analysis (CCA) is a classical representation learning technique for finding correlated variables in multi-view data. This tool has found widespread use in various fields, including recent application to natural language processing, speech recognition, genomics, and cross-modal retrieval. One of the shortcomings of CCA is its restriction to linear mappings, since many real-world multi-view datasets exhibit highly nonlinear relationships. In recent years, several nonlinear extensions of the original linear CCA have been proposed, including kernel and deep neural network methods. These approaches significantly improve upon linear CCA in many practical applications, but have two limitations. First, they are still restricted to families of projection functions which the user must specify (by choosing a kernel or neural network structure). Second, they are computationally demanding.
In this work, we derive a closed form solution to the nonlinear CCA problem without any functional restrictions. We show that the optimal projections can be obtained from the singular value decomposition of a certain operator associated with the joint density of the views. Thus, by estimating the population density from training data, we obtain a practical nonparametric CCA (NCCA) algorithm, which reduces to solving an eigenvalue system. Superficially, this is similar to kernel CCA, but importantly, NCCA does not require the inversion of any kernel matrix. We also derive a partially linear CCA (PLCCA) variant in which one of the views undergoes a linear projection while the other is nonparametric. Finally, we show how our algorithms can be constrained to output non-redundant projections, a feature not possessed by any other nonlinear CCA algorithm.
As we demonstrate on several test cases, our NCCA and PLCCA algorithms are memory-efficient, often run much faster, and perform better than kernel CCA and comparable to deep CCA.
This is joint work with Weiran Wang, Karen Livescu and Yochai Blau.
Wilk Family Award Ceremony
Advanced Learning for Deep Reinforcement Learning
2nd prize winner in the Kasher undergraduate project contest in the EE faculty
Shai Reozenberg, Nadav Bhonker
Supervisor: Itai Hubara
Break & Poster Session
Review of Teaching Activity in SIPL
Audio QR Over Streaming Media
Wilk family award winner
Gal Binyamin, Itai Dagan
Supervisor: Alon Eilam
In cooperation with:
Usage of Surface EMG on the Forearm to Classify Palm Movement
Wilk family award winner
Aviv Peleg, Or Dicker
Supervisor: Tal Shnitzer
In cooperation with: Dr. Oscar Lichtenstein, Faculty of BioMedical Engineering, Technion
Manifold Learning for Anomaly Detection in High-Dimensional Data
Gal Mishne, Ph.D. student
Advisor: Prof. Israel Cohen
In this talk, I will present manifold learning-based methods for anomaly and target detection in supervised and unsupervised settings. Our approach achieves impressive results on various remote sensing image datasets and is shown to be independent of the imaging sensor and noise model. I will also present Diffusion Nets, a geometric auto-encoder that incorporates a manifold embedding of the data in the deep learning framework.