Hardness of noise-free learning for two-hidden-layer neural networks
We give superpolynomial statistical query (SQ) lower bounds for learning two-hidden-layer
ReLU networks with respect to Gaussian inputs in the standard (noise-free) model. No …
ReLU networks with respect to Gaussian inputs in the standard (noise-free) model. No …
Tricking the hashing trick: A tight lower bound on the robustness of countsketch to adaptive inputs
Abstract CountSketch and Feature Hashing (the``hashing trick'') are popular randomized
dimensionality reduction methods that support recovery of l2-heavy hitters and approximate …
dimensionality reduction methods that support recovery of l2-heavy hitters and approximate …
Invertibility aware integration of static and time-series data: An application to lake temperature modeling
Accurate predictions of water temperature are the foundation for many decisions and
regulations, with direct impacts on water quality, fishery yields, and power production …
regulations, with direct impacts on water quality, fishery yields, and power production …
Agnostically Learning Multi-index Models with Queries
We study the power of query access for the task of agnostic learning under the Gaussian
distribution. In the agnostic model, no assumptions are made on the labels and the goal is to …
distribution. In the agnostic model, no assumptions are made on the labels and the goal is to …
New computational and statistical characterizations of neural network learning
A Gollakota - 2023 - repositories.lib.utexas.edu
A foundational goal of machine learning theory is to characterize the inherent computational
and statistical complexity of some of the most basic tasks in machine learning. In this thesis …
and statistical complexity of some of the most basic tasks in machine learning. In this thesis …