Transfer Learning Larochelle
Offered By: Association for Computing Machinery (ACM) via YouTube
Course Description
Overview
Explore transfer learning in deep learning with Hugo Larochelle's 44-minute conference talk from KDD2020. Delve into the importance of transfer learning, few-shot learning, and modern fine-tuning techniques. Examine prototypical networks, model-agnostic meta-learning, and the concept of training a fine-tuning procedure. Evaluate the meta-dataset approach and investigate universal representations and their transformers. Gain insights into the latest results in the field and discuss remaining challenges in transfer learning for deep neural networks.
Syllabus
Intro
WHYTRANSFER LEARNING • Deep leaming successes have required a lot of labeled training data
FEW-SHOT LEARNING
MODERN TRANSFER LEARNING: FINE-TUNING
LEARNING PROBLEM STATEMENT
PROTOTYPICAL NETWORKS
MODEL-AGNOSTIC META-LEARNING • Training a 'fine-tuning procedure
EVALUATION: META-DATASET
UNIVERSAL REPRESENTATIONS
UNIVERSAL REPRESENTATION TRANSFORMER
RESULTS
REMAINING CHALLENGES
Taught by
Association for Computing Machinery (ACM)
Related Courses
Stanford Seminar - Enabling NLP, Machine Learning, and Few-Shot Learning Using Associative ProcessingStanford University via YouTube GUI-Based Few Shot Classification Model Trainer - Demo
James Briggs via YouTube HyperTransformer - Model Generation for Supervised and Semi-Supervised Few-Shot Learning
Yannic Kilcher via YouTube GPT-3 - Language Models Are Few-Shot Learners
Yannic Kilcher via YouTube IMAML- Meta-Learning with Implicit Gradients
Yannic Kilcher via YouTube