SANE: The phases of gradient descent through Sharpness Adjusted Number of Effective parameters

• URL: http://arxiv.org/abs/2305.18490v1
• Date: Mon, 29 May 2023 13:29:31 GMT
• Title: SANE: The phases of gradient descent through Sharpness Adjusted Number of Effective parameters
• Authors: Lawrence Wang, Stephen J. Roberts
• Abstract summary: We consider the Hessian matrix during network training. We show that Sharpness Adjusted Number of Effective parameters (SANE) is robust to large learning rates. We provide evidence and the Hessian shifts across "loss basins" at large learning rates.
• Score: 19.062678788410434
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Modern neural networks are undeniably successful. Numerous studies have investigated how the curvature of loss landscapes can affect the quality of solutions. In this work we consider the Hessian matrix during network training. We reiterate the connection between the number of "well-determined" or "effective" parameters and the generalisation performance of neural nets, and we demonstrate its use as a tool for model comparison. By considering the local curvature, we propose Sharpness Adjusted Number of Effective parameters (SANE), a measure of effective dimensionality for the quality of solutions. We show that SANE is robust to large learning rates, which represent learning regimes that are attractive but (in)famously unstable. We provide evidence and characterise the Hessian shifts across "loss basins" at large learning rates. Finally, extending our analysis to deeper neural networks, we provide an approximation to the full-network Hessian, exploiting the natural ordering of neural weights, and use this approximation to provide extensive empirical evidence for our claims.

Related papers

• Why do Learning Rates Transfer? Reconciling Optimization and Scaling Limits for Deep Learning [77.82908213345864]
  We find empirical evidence that learning rate transfer can be attributed to the fact that under $mu$P and its depth extension, the largest eigenvalue of the training loss Hessian is largely independent of the width and depth of the network. We show that under the neural tangent kernel (NTK) regime, the sharpness exhibits very different dynamics at different scales, thus preventing learning rate transfer.
  arXiv  Detail & Related papers  (2024-02-27T12:28:01Z)
• On the ISS Property of the Gradient Flow for Single Hidden-Layer Neural Networks with Linear Activations [0.0]
  We investigate the effects of overfitting on the robustness of gradient-descent training when subject to uncertainty on the gradient estimation. We show that the general overparametrized formulation introduces a set of spurious equilibria which lay outside the set where the loss function is minimized.
  arXiv  Detail & Related papers  (2023-05-17T02:26:34Z)
• Globally Optimal Training of Neural Networks with Threshold Activation Functions [63.03759813952481]
  We study weight decay regularized training problems of deep neural networks with threshold activations. We derive a simplified convex optimization formulation when the dataset can be shattered at a certain layer of the network.
  arXiv  Detail & Related papers  (2023-03-06T18:59:13Z)
• Implicit Stochastic Gradient Descent for Training Physics-informed Neural Networks [51.92362217307946]
  Physics-informed neural networks (PINNs) have effectively been demonstrated in solving forward and inverse differential equation problems. PINNs are trapped in training failures when the target functions to be approximated exhibit high-frequency or multi-scale features. In this paper, we propose to employ implicit gradient descent (ISGD) method to train PINNs for improving the stability of training process.
  arXiv  Detail & Related papers  (2023-03-03T08:17:47Z)
• Learning to Learn with Generative Models of Neural Network Checkpoints [71.06722933442956]
  We construct a dataset of neural network checkpoints and train a generative model on the parameters. We find that our approach successfully generates parameters for a wide range of loss prompts. We apply our method to different neural network architectures and tasks in supervised and reinforcement learning.
  arXiv  Detail & Related papers  (2022-09-26T17:59:58Z)
• Can pruning improve certified robustness of neural networks? [106.03070538582222]
  We show that neural network pruning can improve empirical robustness of deep neural networks (NNs) Our experiments show that by appropriately pruning an NN, its certified accuracy can be boosted up to 8.2% under standard training. We additionally observe the existence of certified lottery tickets that can match both standard and certified robust accuracies of the original dense models.
  arXiv  Detail & Related papers  (2022-06-15T05:48:51Z)
• Pruning in the Face of Adversaries [0.0]
  We evaluate the impact of neural network pruning on the adversarial robustness against L-0, L-2 and L-infinity attacks. Our results confirm that neural network pruning and adversarial robustness are not mutually exclusive. We extend our analysis to situations that incorporate additional assumptions on the adversarial scenario and show that depending on the situation, different strategies are optimal.
  arXiv  Detail & Related papers  (2021-08-19T09:06:16Z)
• Spline parameterization of neural network controls for deep learning [0.0]
  We choose a fixed number of B-spline basis functions whose coefficients are the trainable parameters of the neural network. We numerically show that the spline-based neural network increases robustness of the learning problem towards hyper parameters.
  arXiv  Detail & Related papers  (2021-02-27T19:35:45Z)
• The FaceChannel: A Fast & Furious Deep Neural Network for Facial Expression Recognition [71.24825724518847]
  Current state-of-the-art models for automatic Facial Expression Recognition (FER) are based on very deep neural networks that are effective but rather expensive to train. We formalize the FaceChannel, a light-weight neural network that has much fewer parameters than common deep neural networks. We demonstrate how our model achieves a comparable, if not better, performance to the current state-of-the-art in FER.
  arXiv  Detail & Related papers  (2020-09-15T09:25:37Z)
• HALO: Learning to Prune Neural Networks with Shrinkage [5.283963846188862]
  Deep neural networks achieve state-of-the-art performance in a variety of tasks by extracting a rich set of features from unstructured data. Modern techniques for inducing sparsity and reducing model size are (1) network pruning, (2) training with a sparsity inducing penalty, and (3) training a binary mask jointly with the weights of the network. We present a novel penalty called Hierarchical Adaptive Lasso which learns to adaptively sparsify weights of a given network via trainable parameters.
  arXiv  Detail & Related papers  (2020-08-24T04:08:48Z)
• Learning Rates as a Function of Batch Size: A Random Matrix Theory Approach to Neural Network Training [2.9649783577150837]
  We study the effect of mini-batching on the loss landscape of deep neural networks using spiked, field-dependent random matrix theory. We derive analytical expressions for the maximal descent and adaptive training regimens for smooth, non-Newton deep neural networks. We validate our claims on the VGG/ResNet and ImageNet datasets.
  arXiv  Detail & Related papers  (2020-06-16T11:55:45Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.