Can Local Learning Match Self-Supervised Backpropagation?
This work addresses the challenge of making local learning rules competitive with global backpropagation for self-supervised learning, which is incremental but important for more biologically plausible AI systems.
The paper tackled the problem of local self-supervised learning (local-SSL) struggling to match the performance of global backpropagation-based self-supervised learning (global BP-SSL) in deep neural networks, and developed novel local-SSL variants that approximate global BP-SSL, achieving matching performance on image datasets like CIFAR-10, STL-10, and Tiny ImageNet.
While end-to-end self-supervised learning with backpropagation (global BP-SSL) has become central for training modern AI systems, theories of local self-supervised learning (local-SSL) have struggled to build functional representations in deep neural networks. To establish a link between global and local rules, we first develop a theory for deep linear networks: we identify conditions for local-SSL algorithms (like Forward-forward or CLAPP) to implement exactly the same weight update as a global BP-SSL. Starting from the theoretical insights, we then develop novel variants of local-SSL algorithms to approximate global BP-SSL in deep non-linear convolutional neural networks. Variants that improve the similarity between gradient updates of local-SSL with those of global BP-SSL also show better performance on image datasets (CIFAR-10, STL-10, and Tiny ImageNet). The best local-SSL rule with the CLAPP loss function matches the performance of a comparable global BP-SSL with InfoNCE or CPC-like loss functions, and improves upon state-of-the-art for local SSL on these benchmarks.