We developed an automated method to detect solar radio bursts from the dynamical spectrographs, and analyzed the origin of the detected solar radio bursts.
We develop a de-novo method for predicting Short Linear Motifs (SLiMs) in protein sequences. We use Finding SeMo to find motifs across all species and use them to predict functions of uncharacterized protein sequences.
We study extreme events in time series with long-range correlations. We find that it is possible to infer the long memory of time series just by extreme events.
We analyze signal propagation in ReLU networks with anti-correlated weights and demonstrate that it has an order-to-chaos phase transition, unlike the uncorrelated case. Furthermore, we demonstrate that ReLU networks initialized at this criticality train faster.
In the past few years, there has been some progress in understanding Deep Neural Networks (DNNs) through ideas from statistical physics. These studies have shed light on various theoretical questions about DNNs. This includes, their function space and generalization properties. For my master’s thesis, I have worked on the information propagation in Deep ReLU networks with correlated weights. In particular, we show that ReLU networks with anti-correlated weights have an order-to-chaos criticality, unlike the uncorrelated weight case. Furthermore, we propose intializing ReLU networks at this criticality, and demonstrate that ReLU networks with anti-correlated intialization train faster. ArXiv link to the paper.
Published:
No talks yet.