Non-quadratic Mirror Descent exhibits exponential initialization sensitivity in convex settings, shown via 3D constructions and KL-regularized simplex examples, with Bregman anchoring proposed for stabilization.
Train longer, generalize better: closing the generalization gap in large batch training of neural networks
6 Pith papers cite this work. Polarity classification is still indexing.
6
Pith papers citing it
abstract
Background: Deep learning models are typically trained using stochastic gradient descent or one of its variants. These methods update the weights using their gradient, estimated from a small fraction of the training data. It has been observed that when using large batch sizes there is a persistent degradation in generalization performance - known as the "generalization gap" phenomena. Identifying the origin of this gap and closing it had remained an open problem. Contributions: We examine the initial high learning rate training phase. We find that the weight distance from its initialization grows logarithmically with the number of weight updates. We therefore propose a "random walk on random landscape" statistical model which is known to exhibit similar "ultra-slow" diffusion behavior. Following this hypothesis we conducted experiments to show empirically that the "generalization gap" stems from the relatively small number of updates rather than the batch size, and can be completely eliminated by adapting the training regime used. We further investigate different techniques to train models in the large-batch regime and present a novel algorithm named "Ghost Batch Normalization" which enables significant decrease in the generalization gap without increasing the number of updates. To validate our findings we conduct several additional experiments on MNIST, CIFAR-10, CIFAR-100 and ImageNet. Finally, we reassess common practices and beliefs concerning training of deep models and suggest they may not be optimal to achieve good generalization.
representative citing papers
Continual classification in homogeneous models is sequential projections onto margin sets, with local linear convergence under regularity properties for random and cyclic tasks, extended to regression.
Effective data transferred from pre-training to fine-tuning is described by a power law in model parameter count and fine-tuning dataset size, acting like a multiplier on the fine-tuning data.
LAMB optimizer trains BERT with batch size 32868, reducing training time to 76 minutes on TPUv3 Pod without performance loss.
Language models show good calibration when asked to estimate the probability that their own answers are correct, with performance improving as models get larger.
Ranked preference modeling outperforms imitation learning for language model alignment and scales more favorably with model size.
citing papers explorer
-
Mirror Descent Beyond Euclidean Stability: An Exponential Separation in Initialization Sensitivity
Non-quadratic Mirror Descent exhibits exponential initialization sensitivity in convex settings, shown via 3D constructions and KL-regularized simplex examples, with Bregman anchoring proposed for stabilization.
-
Convergence of Continual Learning in Homogeneous Deep Networks
Continual classification in homogeneous models is sequential projections onto margin sets, with local linear convergence under regularity properties for random and cyclic tasks, extended to regression.
-
Scaling Laws for Transfer
Effective data transferred from pre-training to fine-tuning is described by a power law in model parameter count and fine-tuning dataset size, acting like a multiplier on the fine-tuning data.
-
Large Batch Optimization for Deep Learning: Training BERT in 76 minutes
LAMB optimizer trains BERT with batch size 32868, reducing training time to 76 minutes on TPUv3 Pod without performance loss.
-
Language Models (Mostly) Know What They Know
Language models show good calibration when asked to estimate the probability that their own answers are correct, with performance improving as models get larger.
-
A General Language Assistant as a Laboratory for Alignment
Ranked preference modeling outperforms imitation learning for language model alignment and scales more favorably with model size.