Learning with not Enough Data Part 2: Active Learning

The performance of supervised learning tasks improves with more high-quality labels available. However, it is expensive to collect a large number of labeled samples. Active learning is one paradigm to deal with not enough labeled data, when there are resources for labeling more data samples but under a limited budget....

This is part 2 of what to do when facing a limited amount of labeled data for supervised learning tasks. This time we will get some amount of human labeling work involved, but within a budget limit, and therefore we need to be smart when selecting which samples to label. Notations Symbol Meaning K Number of unique class labels. (xl,y)∼X,y∈{0,1}K Labeled dataset. y is a one-hot representation of the true label. u∼U Unlabeled dataset. D=X∪U The entire dataset, including both labeled and unlabeled examples. x Any sample which can be either labeled or unlabeled. xi The i-th sample. U(x)…

related reading

• Learning with not Enough Data Part 1: Semi-Supervised Learning | Lil'Loglilianweng.github.io
• Growing or Compressing Datasets · Introduction to Data-Centric AIdcai.csail.mit.edu
• Bayesian Neural Networkscs.toronto.edu
• Thinking about High-Quality Human Data | Lil'Loglilianweng.github.io
• Unifying Approaches in Active Learning and Active Sampling via Fisher Information and Information-Theoretic Quantities | PDFarxiv.org
• CS231n Deep Learning for Computer Visioncs231n.github.io
• An Overview of Deep Semi-Supervised Learningarxiv.org
• A Course in Machine Learningciml.info
• Self-supervised learning: The dark matter of intelligenceai.facebook.com
• Semi-Supervised Learning: Techniques & Examples [2024]v7labs.com
• IsoCompute Playbook: Optimally Scaling Sampling Compute for RL Training of LLMscompute-optimal-rl-llm-scaling.github.io
• sparseAutoencoder.pdfweb.stanford.edu