Graph Constructions for Machine Learning Applications - New Insights and Algorithms

Explore graph constructions for machine learning applications in this IEEE Signal Processing Society webinar presented by Antonio Ortega from USC. Delve into basic definitions, graph signal variation, and semisupervised learning techniques. Examine graph signal sampling, active semisupervised learning, and theoretical analysis of conventional approaches. Investigate orthogonalization, linear embeddings, label propagation, and deep neural networks. Learn about supervised classification, smoothness regularization, local political interpolation, and motor model selection. Gain insights into the geometry of deep learning and discover new algorithms for enhancing machine learning applications using graph-based methods.

Introduction
Motivation
Outline
Basic definitions
Graph signal variation
Semisupervised learning
Graph signal sampling
Active semisupervised learning
Graph signal variations
Paper
Theoretical Analysis
Conventional Approach
orthogonalization
linear embeddings
label propagation
deep neural networks
supervised classification
smoothness
regularization
local political interpolation
local nonparametric approach
motor model selection
local interpolation
Geometry of deep learning