Introduction to Self-supervised Learning

This presentation dive into applications of AI, progressing from modern machine learning techniques to more specific topic self-supervised learning.

author

Zhenyuan Lu

where

School of Medicine, University of Washington, Seattle

given

reading time

1 min

A diagram of semi-supervised learning: a large amount of unlabeled data trains an encoder φ by contrastive learning; fine-tuned on a few labeled data, it gives the unlabeled data pseudo labels, and the pseudo-labeled data trains the last fine-tuned model.

This presentation dive into applications of AI, progressing from modern machine learning techniques to more specific topic self-supervised learning. As the intricate nature of biological data demands sophisticated analytical tools, the gist points of this talk illustrating how self-supervised learning can be adeptly applied to RNA-seq analysis, offering a glimpse into the potential of AI in unraveling complex biological narratives.

A slide of the talk, Machine Learning, in three columns: supervised learning, a height in centimetres giving a weight in kilograms; unsupervised learning, a dog among dragons that “John Snow still doesn’t know”; reinforcement learning, training a dragon to fetch the ball and not burn it; the speaker’s webcam in the corner