Learning with Limited Labels - ODSC West 2020

Explore cutting-edge techniques for machine learning with limited labeled data in this 40-minute conference talk from ODSC West 2020. Delve into domain adaptation, low-shot learning, and self-supervised learning algorithms that enable transfer of information across multiple domains and recognition of novel categories with few-shot samples. Discover how these approaches allow learning systems to automatically adapt to real-world variations and new environmental conditions. Examine specific topics such as adversarial multiple source domain adaptation, multi-source distilling domain adaptation, learning invariant risks and representations for domain transfer, compositional few-shot recognition with primitive discovery and enhancing, distant-domain few-shot recognition with mid-level patterns, and generalized zero-shot learning with dual adversarial networks. Gain insights into overcoming the challenges of obtaining large-scale labeled datasets and learn strategies to develop more efficient and adaptable machine learning models.

Introduction
Learning with Limited Labels
Domain Shifts
Deep Neural Network
Architecture
Multisource Distillation
Visualisation
Intrinsic Objective
Image Classification
Domain Adaptation
Type Text Classification