To explain gating of memory encoding, magnetoencephalography (MEG) was analyzed over multi-regional network of negative correlations between alpha band power during cue (cue-alpha) and gamma band power during item presentation (item-gamma) in Remember (R) and No-remember (NR) condition. Persistent homology with graph filtration on alpha-gamma correlation disclosed topological invariants to expl...

Deep convolutional network has been the state-ofthe-art approach for a wide variety of tasks over the last few years. Its successes have, in many cases, turned it into the default model in quite a few domains. In this work we will demonstrate that convolutional networks have limitations that may, in some cases, hinder it from learning properties of the data, which are easily recognizable by tra...

The driving force behind convolutional networks – the most successful deep learning architecture to date, is their expressive power. Despite its wide acceptance and vast empirical evidence, formal analyses supporting this belief are scarce. The primary notions for formally reasoning about expressiveness are efficiency and inductive bias. Expressive efficiency refers to the ability of a network ...

Deep convolutional networks are well-known for their high computational and memory demands. Given limited resources, how does one design a network that balances its size, training time, and prediction accuracy? A surprisingly effective approach to trade accuracy for size and speed is to simply reduce the number of channels in each convolutional layer by a fixed fraction and retrain the network....

The classification of imbalanced datasets has recently attracted significant attention due to its implications in several real-world use cases. In such scenarios, the datasets have skewed class distributions while very few data instances are associated with certain classes. The classifiers developed on such datasets tend to favor the majority classes and are biased against the minority class. D...

Geometric matrix completion (GMC) has been proposed for recommendation by integrating the relationship (link) graphs among users/items into matrix completion (MC) . Traditional GMC methods typically adopt graph regularization to impose smoothness priors for MC. Recently, geometric deep learning on graphs (GDLG) is proposed to solve the GMC problem, showing better performance than existing GMC m...

motivated by the Convolutional Neural Networks about digit recognition and ImageNet deep neural network by Krizhevsky et al. [1], I did this project on Guqin notation recognition, which classified reduced characters with positioned 1-10 (一 -十) in handwritten Chinese characters and translated to other music recording scores. I built a four-layer convolutional neural network using adjusted CaffeN...

Images captured in low-light conditions usually suffer from very low contrast, which increases the difficulty of subsequent computer vision tasks in a great extent. In this paper, a low-light image enhancement model based on convolutional neural network and Retinex theory is proposed. Firstly, we show that multi-scale Retinex is equivalent to a feedforward convolutional neural network with diff...