Abstract Knowledge Distillation (KD) is a popular technique to transfer knowledge from a
teacher model or ensemble to a student model. Its success is generally attributed to the …

The supplementary material contains the proof of Theorem 4.1, some of the proposed
algorithms from Section 3.2, and details about the exploratory data analysis of the Higgs …

Aiming at recovering an unknown tensor (ie, multi-way array) corrupted by both sparse
outliers and dense noises, robust tensor decomposition (RTD) serves as a powerful pre …

The deep active-subspace method is a neural-network based tool for the propagation of
uncertainty through computational models with high-dimensional input spaces. Unlike the …

In this work, we present an extension of genetic algorithm (GA) which exploits the
supervised learning technique called active subspaces (AS) to evolve the individuals on a …

Several recent empirical studies demonstrate that important machine learning tasks such as
training deep neural networks, exhibit a low-rank structure, where most of the variation in the …

We propose an algorithmic framework, that employs active subspace techniques, for
scalable global optimization of functions with low effective dimension (also referred to as low …

Bayesian inference for neural networks, or Bayesian deep learning, has the potential to
provide well-calibrated predictions with quantified uncertainty and robustness. However, the …

The focus of this work is on the application of classical Model Order Reduction techniques,
such as Active Subspaces and Proper Orthogonal Decomposition, to Deep Neural …

Memory footprint is one of the main limiting factors for large neural network training. In
backpropagation, one needs to store the input to each operation in the computational graph …