Learn about designing optimal adaptation strategies for few-shot recognition in this 30-minute AutoML seminar presented by researcher Lukasz Dudziak. Explore how neural architecture search (NAS) can be applied to develop classifiers that rapidly adapt and generalize to novel classes. Dive into key findings regarding the trade-offs between searching during meta-training and meta-testing time in multi-domain contexts, and discover how carefully crafted adaptation architectures can enhance fine-tuning approaches for few-shot learning scenarios.
Overview
Syllabus
Neural Fine-Tuning Search for Few-Shot Learning
Taught by
AutoML Seminars