Permutation Invariant Learning with High-Dimensional Particle Filters
This addresses challenges in continual learning for AI systems, offering a principled solution to mitigate issues like catastrophic forgetting, though it appears incremental as it builds on existing particle filter and Bayesian methods.
The paper tackles the problem of catastrophic forgetting and loss of plasticity in sequential deep learning by introducing a permutation-invariant learning framework based on high-dimensional particle filters, which improves performance and reduces variance on benchmarks like SplitMNIST, SplitCIFAR100, and ProcGen.
Sequential learning in deep models often suffers from challenges such as catastrophic forgetting and loss of plasticity, largely due to the permutation dependence of gradient-based algorithms, where the order of training data impacts the learning outcome. In this work, we introduce a novel permutation-invariant learning framework based on high-dimensional particle filters. We theoretically demonstrate that particle filters are invariant to the sequential ordering of training minibatches or tasks, offering a principled solution to mitigate catastrophic forgetting and loss-of-plasticity. We develop an efficient particle filter for optimizing high-dimensional models, combining the strengths of Bayesian methods with gradient-based optimization. Through extensive experiments on continual supervised and reinforcement learning benchmarks, including SplitMNIST, SplitCIFAR100, and ProcGen, we empirically show that our method consistently improves performance, while reducing variance compared to standard baselines.