Differentially Quantized Gradient Methods
Offered By: Simons Institute via YouTube
Course Description
Overview
Explore a 34-minute lecture by Victoria Kostina from the California Institute of Technology, presented at the Simons Institute, focusing on Differentially Quantized Gradient Methods. Delve into information-theoretic approaches for developing trustworthy machine learning systems, gaining insights into advanced techniques for optimizing gradient-based algorithms through quantization methods.
Syllabus
Differentially Quantized Gradient Methods
Taught by
Simons Institute
Related Courses
Information-Theoretic Foundations of Generative Adversarial ModelsSimons Institute via YouTube Generalization Bounds for Neural Network Based Decoders
Simons Institute via YouTube Optimal Neural Network Compressors and the Manifold Hypothesis
Simons Institute via YouTube Improving Accuracy-Privacy Tradeoff via Model Reprogramming
Simons Institute via YouTube Fundamental Trade-Offs in FL-FA - Sparsity - DP - Communication Constraints
Simons Institute via YouTube