gradient methods

The Broximal Point Method and Its Impact on Optimization Theory and Machine Learning

-

B9 R2325

This talk introduces the Broximal Point Method (BPM), a condition-number-free alternative to classical proximal methods, that replaces proximal penalties with ball constraints, surveys its theoretical extensions and distributed variants, and explains how approximate non-Euclidean Brox updates underpin practical machine learning optimizers.

Unveiling Insights from "Gradient Descent Converges Linearly for Logistic Regression on Separable Data"

-

B1 L0 R0118

Abstract In this presentation, I will share a paper titled "Gradient Descent Converges Linearly for Logistic Regression on Separable Data", a work highly related to my ongoing research. I will explore its relevance to my current research topic and discuss the inspiration for our future works. Abstract of the paper: We show that running gradient descent with variable learning rate guarantees loss f(x) \leq 1.1f(x^*)+\epsilon for the logistic regression objective, where the error \epsilon decays exponentially with the number of iterations and polynomially with the magnitude of the entries of an

On the Natural Gradient Descent

-

KAUST

Abstract Numerous problems in scientific computing can be formulated as optimization problems of suitable parametric models over parameter spaces. Neural network and deep learning methods provide unique capabilities for building and optimizing such models, especially in high-dimensional settings. Nevertheless, neural networks and deep learning techniques are often opaque and resistant to precise control of their mathematical properties in terms of architectures, hyperparameters, etc. Consequently, optimizing neural network models can result in a laborious hyperparameter tuning process that