Proposes deep weight prior for improving neural network performance.
problem Improving neural network performance with limited training data.
method Defines deep weight prior (DWP) as an implicit distribution and proposes variational inference methods.
result Improves performance of Bayesian neural networks with limited data and accelerates conventional CNN training.
Proposes MOPED method for choosing priors in Bayesian DNNs.
problem Challenges in specifying meaningful priors for deep neural networks.
method Two-stage hierarchical modeling with empirical Bayes.
result MOPED enables scalable variational inference and reliable uncertainty quantification.
We study deep Bayesian neural networks with Gaussian priors, revealing heavy-tailed unit activations.
problem Characterizing regularization effects in deep Bayesian neural networks.
method Investigation of deep Bayesian neural networks with Gaussian weight priors and ReLU-like nonlinearities.
result The prior distribution on units becomes increasingly heavy-tailed with depth, influencing activation patterns.
Bayesian deep learning with heavy-tailed weights achieves near-optimal performance.
problem Deep neural networks with heavy-tailed weights achieve near-optimal performance in various contexts.
method Introduced a Bayesian deep learning prior based on heavy-tailed weights and ReLU activation, showing near-optimal minimax contraction rates.
result Posterior distribution achieves near-optimal minimax contraction rates, adaptive to smoothness and intrinsic dimension.
We propose a novel method for compressed sensing recovery using untrained deep generative models. Our method is based on the recently proposed Deep Image Prior (DIP), wherein the convolutional weights of the network are optimized to match the observed measurements. We show that this approach can be applied to solve any…
Stabilizes deep Bayesian neural networks with self-stabilizing priors.
problem Brittleness and difficulty in training deep Bayesian neural networks.
method Signal propagation theory, reformulated ELBO, self-stabilizing priors.
result Improved convergence and robustness in training deeper networks and noisier settings.
New method learns priors for Bayesian neural networks from datasets.
problem Lack of prior beliefs in Bayesian deep learning.
method Amortised variational inference to learn priors from datasets.
result Flexible Bayesian neural networks for meta-learning and within-task minibatching.
One-dimensional CNNs improve signal recovery from sparse measurements.
problem Recovering signals from limited data.
method One-dimensional Deep Image Prior (DIP) using CNNs with regularization.
result One-dimensional CNNs outperform traditional methods in signal recovery.
Bayesian sparsification reduces deep neural network complexity.
problem Complexity of deep neural networks limits their performance.
method Combines Bayesian shrinkage priors with stochastic variational inference.
result Bayesian model reduction (BMR) is a more efficient alternative for pruning model weights.
Bayesian deep learning uses function-space priors to improve model uncertainty and robustness.
problem Bayesian deep learning struggles with model-specific weight-space priors that are hard to interpret and specify.
method Apply a Dirichlet prior in predictive space and perform approximate function-space variational inference.
result The approach improves uncertainty quantification, scalability, and adversarial robustness in large-scale image classification.
Hybrid Bayesian neural networks use function uncertainty for probabilistic inference.
problem Uncertainty in neural network weights is hard to specify and interpret.
method Integrates probabilistic layers with standard deterministic layers for function uncertainty.
result Improves probabilistic inference by encoding function uncertainty.
Soft diamond regularizers improve deep learning performance and sparsity.
problem Improving deep learning performance and sparsity of trained weights.
method New soft diamond synaptic weight priors based on thick-tailed symmetric alpha stable probability curves.
result Soft diamond regularizers outperform state-of-the-art methods in deep learning tasks.
Compression and computational efficiency in deep learning have become a problem of great significance. In this work, we argue that the most principled and effective way to attack this problem is by adopting a Bayesian point of view, where through sparsity inducing priors we prune large parts of the network. We introduc…
Study on Bayesian transformers finds issues with weight-space inference and prior specification.
problem Challenges in obtaining meaningful uncertainty estimates for transformer models.
method Proposed a novel method based on implicit reparameterization of the Dirichlet distribution for variational inference on attention weights.
result Proposed method performs competitively with baselines in estimating predictive uncertainty.
New filter bank regularization improves DCNNs by incorporating image priors.
problem Improving DCNNs' robustness and generality.
method Structured filter bank regularization of DCNN kernels.
result Filter bank regularization leads to faster convergence and better generalization.
This paper uses reference priors to improve deep learning models with unlabeled and labeled data.
problem Improving deep learning models with limited labeled data and unlabeled data from the same or related tasks.
method Develops and applies generalizations of reference priors for deep networks to exploit unlabeled and labeled data.
result Demonstrates new semi-supervised learning and pretraining methods for transfer learning.
New model separates object attributes for better perceptual grouping.
problem Perceptual grouping of complex visual scenes.
method Spatial mixture models with learnable priors.
result Outperforms state-of-the-art methods in perceptual grouping.
DSARF models complex spatio-temporal data with deep switching auto-regressive factors.
problem Forecasting complex spatio-temporal data with recurring patterns.
method Deep switching auto-regressive factorization (DSARF) with stochastic variational inference.
result DSARF outperforms state-of-the-art methods in long- and short-term prediction accuracy.
This work tackles the challenge of Bayesian deep learning by proposing a new framework for matching Gaussian process priors with neural network parameters.
problem The challenge of specifying priors over neural network parameters, which affects the induced functional prior and is uncontrolled.
method The approach involves defining functional priors using Gaussian processes and matching these priors with the functional prior of neural networks through the minimization of Wasserstein distance.
result The proposed framework offers systematic performance improvements over alternative priors and approximate Bayesian deep learning approaches.
New theory for BNNs with Gaussian priors achieves optimal posterior concentration rates.
problem Lack of theoretical results for BNNs with Gaussian priors.
method New approximation theory for non-sparse DNNs with bounded parameters.
result BNNs with non-sparse general priors can achieve near-minimax optimal posterior concentration rates.
Infinite CNNs lose spatial correlations, but can be restored by correlated weights.
problem Infinite CNNs lose spatial correlations, which are crucial for their performance.
method Introduced correlated weights to restore spatial correlations in infinite CNNs.
result Optimal performance is achieved with a moderate level of weight correlation.
DMSTF models spatio-temporal data with deep Markov priors.
problem Analyzing nonlinear multimodal spatio-temporal dynamics.
method Deep Markov spatio-temporal factorization with stochastic variational inference.
result DMSTF outperforms other methods in predictive performance and clustering.
Gradient descent aligns weights in deep linear networks for binary classification.
problem Aligning weights in deep linear networks for binary classification.
method Gradient descent applied to strictly decreasing loss functions.
result Normalized weight matrices align across layers, converging to the maximum margin solution.
Heavy-tailed regularization improves deep neural network performance.
problem Improving generalization of deep neural networks.
method Introducing Heavy-Tailed Regularization, using differentiable penalty terms and Bayesian statistics.
result Heavy-tailed regularization outperforms conventional regularization techniques.
Deep Gaussian Processes are reinterpreted as deep trigonometric networks for tractable inference.
problem Challenging inference in DGPs due to intractable marginalization in latent function space.
method Viewing DGPs as deep trigonometric networks with Bochner's theorem, and using the wide limit with a bottleneck to translate DGPs into deep trigonometric networks.
result The weight space view yields the same effective covariance functions as obtained in function space, and varying prior distributions over network parameters is equivalent to employing different kernels.
Bayesian deep neural networks converge to processes with α-stable marginals under infinite variance weights.
problem Representation learning in deep kernel processes is hindered by deterministic covariance kernels.
method Showed convergence to α-stable processes with conditionally Gaussian representations in infinite-width networks.
result Conditional random covariance kernels can be recursively linked, even if the process is α-stable.
Dropout is explained as a structured shrinkage prior in neural networks.
problem Understanding the effectiveness of dropout in preventing overfitting.
method Proposes a novel framework to explain dropout as a structured shrinkage prior, considering continuous distributions and Bernoulli noise.
result Dropout's Monte Carlo training objective approximates marginal MAP estimation.
Bayesian priors for neural networks are improved by incorporating weight correlations and tail behavior.
problem Improving Bayesian priors for neural networks to better reflect true beliefs and performance.
method Analyzed summary statistics of neural network weights in different architectures and incorporated these observations into new priors.
result Improved performance on image classification datasets by using new priors that account for weight correlations and tail behavior.
New method improves deep RL by combining emphatic weightings with replay data.
problem Improving sample efficiency and scaling model-free RL methods.
method Developed a multi-step emphatic weighting and time-reversed n-step TD learning algorithm. result The new approach reduces variance and provides convergence guarantees.
The paper introduces a portfolio construction method using Black-Litterman model and factors.
problem Developing an efficient portfolio construction method using Black-Litterman model and factors.
method The method involves selecting 20 factors based on global market, asset class, and stock characteristics, applying various weight allocation methods including Black-Litterman model, and incorporating deep learning for dynamic weight updates.
result The model using Black-Litterman and deep learning outperforms other weight allocation schemes.
Deep Gaussian processes can have non-degenerate and non-Gaussian limits.
problem Understanding the behavior of deep Gaussian processes as depth grows.
method Studying the limit of compositional Gaussian processes where each layer is a Gaussian process.
result Identified a sharp bandwidth threshold above which the limit is degenerate, and proved that for bandwidths below this threshold, the limit is a non-degenerate and non-Gaussian distribution.
Sharp asymptotics derived for phase retrieval and compressed sensing with random generative priors.
problem Phase retrieval and compressed sensing with random measurement matrices.
method Sharp asymptotics derived for optimal performance and polynomial algorithm for random generative priors.
result Compressed phase retrieval becomes tractable with random generative priors, unlike sparse priors.
Radial BNNs offer a scalable, continuous weight distribution for Bayesian deep learning.
problem Discrete support in Bayesian deep learning methods like MC dropout.
method Radial BNNs with full support over weight-space.
result Radial BNNs outperform discrete-support methods in real-world applications.
Bayesian neural networks with dependent weights converge to Gaussian mixtures.
problem Limitations of standard Gaussian priors in neural networks.
method Posterior analysis with Gaussian likelihood for networks with dependent weights.
result Posterior distribution identified in the wide-width limit, ensuring invertibility of random covariance matrix.
New method prevents deep learning forgetting past by remembering key examples.
problem Catastrophic forgetting in continual learning.
method Functional regularisation using Gaussian Process formulation.
result Achieves state-of-the-art performance on benchmarks.
Researchers derive exact priors for finite Bayesian neural networks.
problem Understanding non-Gaussian priors in finite Bayesian neural networks.
method Analytical derivation of function space priors for finite fully-connected feedforward networks.
result Exact solutions for priors of finite networks, including Meijer G-function for linear networks and mixtures for ReLU networks.
This paper introduces hierarchical Gaussian process priors for neural networks to capture weight correlations and inductive biases.
problem Capturing weight correlations and inductive biases in neural networks.
method Hierarchical Gaussian process priors with unit embeddings and input-dependent kernels.
result Hierarchical Gaussian process priors provide competitive predictive performance and desirable uncertainty estimates.
SPIDER uses deep neural networks for streaming tensor factorization.
problem Lack of effective approach for deep tensor factorization of streaming data.
method Bayesian neural networks with spike-and-slab prior, Taylor expansions, moment matching, and EPI framework.
result Effective incremental updates for latent factors and NN weights.
Algorithms for Magnetic Resonance (MR) image reconstruction from undersampled measurements exploit prior information to compensate for missing k-space data. Deep learning (DL) provides a powerful framework for extracting such information from existing image datasets, through learning, and then using it for reconstructi…
Proposes a new method to initialize neural networks by estimating global curvature of weights.
problem Improving the initialization of neural networks for better training and convergence.
method Estimates the global curvature of weights across layers using the Hessian matrix norm.
result The proposed method helps in more rigorously initializing weights, leading to better performance.
New methods improve deep learning on imbalanced datasets.
problem Poor performance of deep learning on imbalanced datasets.
method Label-distribution-aware margin (LDAM) loss and a training schedule.
result Combination of methods achieves significant performance gains.
Bayesian weight priors improve neural network learning of identity relations.
problem Neural networks struggle to learn abstract and systematic relations, especially identity relations.
method Extended RBP approach using Bayesian weight priors as a regularization term.
result Bayesian weight priors lead to perfect generalization for identity relations and do not hinder standard neural network learning.
Bayesian approach adapts deep network structure for continual learning.
problem Training neural networks with sequential or streaming tasks.
method Bayesian approach to learn deep network structure for each task.
result Model performs comparably or better than recent advances in continual learning.
We consider the task of one-shot learning of visual categories. In this paper we explore a Bayesian procedure for updating a pretrained convnet to classify a novel image category for which data is limited. We decompose this convnet into a fixed feature extractor and softmax classifier. We assume that the target weights…
DICCA maps multi-view data into a shared latent space with interpretable components.
problem Learning from multiple related but distinct data views.
method DICCA extends CCA to deep generative networks and uses sparsity-inducing priors for interpretability.
result DICCA effectively disentangles shared and view-specific variations in multi-view data.
In the past years, Deep convolution neural network has achieved great success in many artificial intelligence applications. However, its enormous model size and massive computation cost have become the main obstacle for deployment of such powerful algorithm in the low power and resource-limited mobile systems. As the c…
Open problem: Establishing bounds for Cayley-table completion to discover discrete algorithmic axioms.
problem Discovering discrete algorithmic axioms missing in deep learning.
method Cayley-table completion as a testbed for algorithmic complexity minimization.
result Formal exact recovery bounds for Cayley-table completion.
Paper proposes new Bayesian neural network models for efficient learning.
problem Efficient learning and model compression in deep neural networks.
method Proposes Spike-and-Slab Group Lasso (SS-GL) and Spike-and-Slab Group Horseshoe (SS-GHS) priors for structured sparsity in Bayesian neural networks.
result Establishes competitive performance in prediction accuracy, model compression, and inference latency compared to baseline models.