Moscow Deep November: Weekend school on Deep learning models of human vision

The members of the 'Attention, Computational Models and Eye Movements' (ACME) Research and Study Group conducted the Weekend school / workshop on Deep learning models of human vision.

Deep learning is a new approach to neural networks that allows stacked layers to improve performance. Individual early layers can be trained unsupervised, and reused for multiple applications with the addition of specialised upper layers. More important for psychology, these networks claim to process images in a layered way that is similar to human visual cortex.

This weekend school provided an introduction into various deep learning algorithms such as autoencoders, Restricted Boltzman machines, Deep convolution nets, Deep recurrant networks, and deep belief networks. The second day was a workshop for practicing these algorithms. Practice was in MATLAB and include DeeBNnet and Matlab Neural Network Toolbox. Debugging help was provided for all ongoing Research and Study Group projects.

The school was supervised by the head of the ACME Group Dr. Joe MacInnes (Associate Professor, School of Psychology) and Georgi Zhulikov, a member of the ACME Group.

The event took place November 24–25, 2017, at 4/2 Armyanskyj per., rooms 414 and 118.

The School materials are available here:  Zhulikov_NeuralNets.pdf
                                                          MacInnes_RBMguide.pdf