Explore cutting-edge applications of deep learning in speech recognition and collaborative filtering systems in this session from the NVIDIA AI Tech Workshop at NeurIPS Expo 2018. Delve into the intricacies of building speech recognition models using synthetic speech and optimizing neural collaborative filtering systems. Gain insights into semantic map filling, problem definition, and architecture overview. Examine various modules including wear module, neural networks, and auto encoders. Discover techniques for handling mode collapse and processing real images. Learn about complex image understanding, graph-based image representation, and structure prediction. Investigate graph permutation invariant architectures and their applications in synthetic examples and Visual Genome datasets. Enhance your knowledge of applied deep learning techniques presented by NVIDIA Research experts in this comprehensive 39-minute session.
Applied Deep Learning in Speech Recognition and Collaborative Filtering - Session 5
Overview
Syllabus
Intro
Semantic Map Filling
Problem Definition
Architecture Overview
Wear Module
Neural Network
Mode Collapse
Auto Encoder
Real Images
Animation
Neural Network Results
Copy and Paste
Understanding complex images
Images as a graph
Good enough graph
Structure prediction
Architecture
Graph permutation invariant
Synthetic example
Visual Genome
Visual Genome Example
Summary
Taught by
NVIDIA Developer
