UMass CICS Highlights Broad Machine Learning Research Advances in 2021
The UMass Manning College of Information and Computer Sciences publishes an extensive retrospective showcasing numerous machine learning papers from 2021. The featured research spans top conferences like ICML and NeurIPS, covering topics from reinforcement learning to differential privacy.
The UMass Manning College of Information and Computer Sciences (CICS) releases a comprehensive retrospective detailing a wealth of machine learning research produced by its students and faculty in 2021. This overview highlights a wide array of papers accepted into prestigious conferences, demonstrating the breadth and depth of the college's ongoing work in the field.
Featured publications from the International Conference on Machine Learning (ICML) explore diverse challenges such as high-confidence generalization for reinforcement learning, observational causal inference, and innovative approaches to node embeddings. Additionally, the ICML selections investigate unbiased alpha divergence minimization and practical mean bounds for small sample sizes.
The retrospective also includes a strong presence at the Conference on Neural Information Processing Systems (NeurIPS), where researchers present advancements in off-policy evaluation, amortized variational inference, and differentially private query answering. Other notable NeurIPS contributions tackle cooperative stochastic bandits, the Turing completeness of bounded-precision recurrent neural networks, and optimized coresets for classification tasks.