I. Melnyk

We propose a hybrid model called R2N2 (Residual RNN), which first models the time series with a simple linear model (like VAR) and then models its residual errors using RNNs. Expand

We propose a new method for training encoder-decoders using non-parallel data that combines a collaborative classifier, attention and the cycle consistency loss that outperforms a state-of-the-art text style transfer system in two out of three quantitative metrics. Expand

We consider estimating structured VAR (vector auto-regressive model), where the structure can be captured by any suitable norm, and establish bounds on the non-asymptotic estimation error. Expand

We present a deep learning model, DE-LSTM, for the simulation of a stochastic process with an underlying nonlinear dynamics. Expand

We proposed a framework for detecting precursors to aviation safety incidents due to human factors based on Hidden Semi-Markov Models (HSMM). Expand

We study the estimation of the mutual information I(X;T`) between the input X to a deep neural network (DNN) and the output vector T of its `th hidden layer (an “internal representation”). Expand

We study the flow of information and the evolution of internal representations during deep neural network (DNN) training, aiming to demystify the compression aspect of the information bottleneck theory. Expand

We propose a new conditional GAN for image captioning that enforces semantic alignment between images and captions through a co-attentive discriminator and a context-aware LSTM sequence generator. Expand

We propose a novel algorithm for text attribute transfer with non-parallel corpora based on the encoder-decoder architecture with attention, augmented with the collaborative classifier and a set of content preservation losses. Expand