Introduces self-regularization for analyzing learning algorithms.
problem Analyzing and optimizing learning algorithms without explicit regularization.
method Develops a self-regularization framework for learning algorithms.
result Provides statistical analysis and minmax-optimal rates for self-regularized algorithms.
GRAC improves reinforcement learning by self-guiding and self-regularizing.
problem Learning divergence and slow updates in reinforcement learning algorithms.
method Self-regularized TD-learning and self-guided policy improvement.
result Achieved or outperformed state-of-the-art results on OpenAI gym tasks.
Random Matrix Theory (RMT) is applied to analyze the weight matrices of Deep Neural Networks (DNNs), including both production quality, pre-trained models such as AlexNet and Inception, and smaller models trained from scratch, such as LeNet5 and a miniature-AlexNet. Empirical and theoretical results clearly indicate th…
Random Matrix Theory (RMT) is applied to analyze weight matrices of Deep Neural Networks (DNNs), including both production quality, pre-trained models such as AlexNet and Inception, and smaller models trained from scratch, such as LeNet5 and a miniature-AlexNet. Empirical and theoretical results clearly indicate that t…
AlphaPruning optimizes LLM pruning using HT-SR theory for better performance.
problem Improving pruning of large language models to reduce size without sacrificing performance.
method AlphaPruning uses HT-SR theory to allocate layerwise sparsity ratios more theoretically.
result AlphaPruning prunes LLaMA-7B to 80% sparsity with reasonable perplexity.
NPMLE estimator automatically chooses the right model complexity for Gaussian mixtures.
problem Learning mixture models and empirical Bayes estimation with non-convex likelihoods.
method Nonparametric maximum likelihood estimator (NPMLE) using complex-analytic techniques.
result NPMLE solution has O(logn) atoms with high probability, improving model complexity. CalNF models rare failures with limited data, improving safety in autonomous systems.
problem Challenges in modeling and debugging rare safety-critical failures due to limited data.
method CalNF, a self-regularized framework for posterior learning from limited data.
result Achieves state-of-the-art performance on data-limited failure modeling and inverse problems.
We propose Mish, a novel self-regularized non-monotonic activation function which can be mathematically defined as: f(x)=xtanh(softplus(x)). As activation functions play a crucial role in the performance and training dynamics in neural networks, we validated experimentally on several well-known benchmarks…
Paper explains how tree ensembles improve predictions by smoothing and regulating smoothness.
problem Understanding why tree ensembles perform well despite their complexity.
method Interpreting tree ensembles as adaptive and self-regularizing smoothers.
result Ensemble trees make more smooth predictions than individual trees and adjust smoothness based on input dissimilarity.
LEAK learns from mistakes to improve point cloud segmentation.
problem Improving point cloud semantic segmentation performance.
method Coarse-to-fine clustering, class-conditional prototypical feature alignment, fairness weighting.
result State-of-the-art performances on different architectures, datasets, and tasks.
TempBalance boosts model performance with low data.
problem Low-data training and fine-tuning in model alignment.
method Inspired by HT-SR theory, TempBalance balances training quality across layers.
result TempBalance improves model performance as data decreases.
TempBalance improves neural network training by balancing learning rates across layers.
problem Improving neural network training performance.
method Layer-wise learning rate scheduling based on HT-SR Theory.
result TempBalance significantly outperforms ordinary SGD and state-of-the-art optimizers.
Study analyzes neural network models to understand generalization performance.
problem Understanding good generalization in neural networks.
method Analyzed a corpus of models from a public contest, breaking ALPHAHAT into scale and shape metrics.
result Identified a Simpson's paradox in metric performance across different model depths and regularization hyperparameters.
Focusing on the grand-canonical extension of the ordinary restricted Boltzmann machine, we suggest an energy-based model for feature extraction that uses a layer of hidden units with varying size. By an appropriate choice of the chemical potential and given a sufficiently large number of hidden resources the generative…
Working in high-dimensional latent spaces, the internal encoding of data in Variational Autoencoders becomes naturally sparse. We discuss this known but controversial phenomenon sometimes refereed to as overpruning, to emphasize the under-use of the model capacity. In fact, it is an important form of self-regularizatio…
Convex neural networks enforce convex constraints on weights and activations, improving generalization.
problem Improving generalization and reducing overfitting in neural networks.
method Enforce convex constraints on weights and activations, using non-negative weights and non-decreasing convex activation functions.
result Convex neural networks self-regularize, outperforming base architectures and achieving similar performance to convolutional architectures.
Untrained neural networks can recover natural images from few measurements.
problem Recovering natural images from a small number of measurements.
method Gradient descent on un-trained convolutional neural networks.
result Untrained neural networks can approximate reconstruct signals/images from a near minimal number of random measurements.
Analog arrays are a promising upcoming hardware technology with the potential to drastically speed up deep learning. Their main advantage is that they compute matrix-vector products in constant time, irrespective of the size of the matrix. However, early convolution layers in ConvNets map very unfavorably onto analog a…
In many applications, one works with neural network models trained by someone else. For such pretrained models, one may not have access to training data or test data. Moreover, one may not know details about the model, e.g., the specifics of the training data, the loss function, the hyperparameter values, etc. Given on…
Unified theory for kernel regression generalizes well under realistic assumptions.
problem Analyzing kernel regression under realistic conditions.
method Unified theory providing rigorous bounds for various settings.
result Self-regularization phenomenon in kernel matrices enables good generalization.
Unified framework detects overfitting in crash classification models.
problem Evaluation metrics fail to detect overfitting in crash classification models.
method Random Matrix Theory and Heavy-Tailed Self-Regularization framework applied to various model types.
result Power-law exponent α reliably distinguishes well-regularized from overfit models.
New findings on neural networks with non-negative weights and low training error.
problem Does a low training error imply a small outer norm for two-layer neural networks?
method Covering number argument and fat-shattering dimension analysis.
result For non-negative output weights, low training error guarantees a well-controlled outer norm.
Given two or more Deep Neural Networks (DNNs) with the same or similar architectures, and trained on the same dataset, but trained with different solvers, parameters, hyper-parameters, regularization, etc., can we predict which DNN will have the best test accuracy, and can we do so without peeking at the test data? In …
Novel anti-grokking phase discovered in neural networks, revealed by HTSR layer quality metric.
problem Understanding and detecting overfitting in neural networks.
method 3-layer MLP, weight decay, HTSR layer quality metric α, correlation traps, Kolmogorov–Smirnov tests. result Anti-grokking phase detected late in training, revealed by α<2 and correlation traps. New results show neural networks generalize well due to polynomial regression, not just overparametrization.
problem Generalization of neural networks despite overparametrization.
method Teacher/Student model with polynomial regression.
result Student networks interpolating teacher-generated data generalize well with a minimal sample size.