Myopic Bayesian Decision Theory for Batch Active Learning with Partial Batch Label Sampling
This work addresses a computational bottleneck in batch active learning for practitioners, though it is incremental as it builds on existing Bayesian Decision Theory frameworks.
The paper tackles the challenge of scaling Bayesian active learning to large batch sizes by deriving Partial Batch Label Sampling (ParBaLS) for the EPIG algorithm, showing experimentally that ParBaLS EPIG achieves superior performance on several datasets with Bayesian Logistic Regression on Neural Embeddings.
Over the past couple of decades, many active learning acquisition functions have been proposed, leaving practitioners with an unclear choice of which to use. Bayesian Decision Theory (BDT) offers a universal principle to guide decision-making. In this work, we derive BDT for (Bayesian) active learning in the myopic framework, where we imagine we only have one more point to label. This derivation leads to effective algorithms such as Expected Error Reduction (EER), Expected Predictive Information Gain (EPIG), and other algorithms that appear in the literature. Furthermore, we show that BAIT (active learning based on V-optimal experimental design) can be derived from BDT and asymptotic approximations. A key challenge of such methods is the difficult scaling to large batch sizes, leading to either computational challenges (BatchBALD) or dramatic performance drops (top-$B$ selection). Here, using a particular formulation of the decision process, we derive Partial Batch Label Sampling (ParBaLS) for the EPIG algorithm. We show experimentally for several datasets that ParBaLS EPIG gives superior performance for a fixed budget and Bayesian Logistic Regression on Neural Embeddings. Our code is available at https://github.com/ADDAPT-ML/ParBaLS.