Transformer Invariances
Yannic and Hugo discuss the surprising ability of transformers to learn invariances without explicit biases, challenging traditional notions of data efficiency and inductive biases in machine learning models. The conversation delves into the potential of transformers to adapt to different data scenarios and the trade-offs between bias and variance in model performance.In this clip
From this podcast
Machine Learning Street Talk (MLST)
#044 - Data-efficient Image Transformers (Hugo Touvron)
Related Questions