NADS improves OoD detection accuracy by 57%.
problem Uncertainty in machine learning models when encountering out-of-distribution data.
method NADS searches for a distribution of architectures that perform well on a given task, optimizing a stochastic OoD detection objective.
result NADS achieves up to 57% improvement in accuracy over state-of-the-art methods.
This work maps Boltzmann distributions to ARNNs for better physics-based model approximations.
problem Approximating Boltzmann distributions of binary systems.
method Exact mapping of Boltzmann distribution to autoregressive neural network architecture.
result New ARNN architectures derived from physical models show superior performance.
The paper proposes a neural network architecture inspired by Langevin Monte Carlo for sampling from target distributions.
problem Sampling from complex target distributions efficiently.
method A neural network architecture inspired by Langevin Monte Carlo is proposed to map samples from a simple reference distribution to samples from the target.
result The proposed neural network architecture achieves approximation rates in the Wasserstein-2 distance for smooth, log-concave target distributions.
This work proposes searching for optimal operation distribution in neural architecture search.
problem Finding optimal neural architecture with specific operations and connections.
method Search for the optimal operation distribution, providing a stochastic and approximate solution.
result Operation distribution holds enough discriminating power to reliably identify a solution and is easier to optimise than traditional encodings.
Study reduces NAS search cost by generating multiple complex architectures in one shot.
problem Finding multiple neural architectures with varying complexities efficiently.
method Uses importance sampling to generate and update multiple distributions of architectures.
result Reduces search cost by finding multiple architectures with different complexities in a single search.
This study evaluates different normalizing flow architectures for MCMC.
problem Lack of systematic comparison of normalizing flow architectures in MCMC.
method Extensive evaluation of various normalizing flow architectures on different MCMC methods and target distributions.
result Contractive residual flows are the best general-purpose models for MCMC.
DrNAS improves neural architecture search with Dirichlet distribution and progressive learning.
problem Efficiently search for neural architectures with improved generalization and exploration.
method Formulates architecture search as a distribution learning problem using Dirichlet distribution and gradient-based optimization. Introduces a progressive learning scheme to handle large-scale tasks.
result Achieves state-of-the-art results on CIFAR-10 and ImageNet, demonstrating improved generalization and exploration.
Secure blockchain architectures protect data privacy in distributed learning.
problem Data privacy and trust in distributed learning across organizations.
method Adequate encryption and blockchain mechanisms ensure data privacy and trust in iterative learning.
result Secure sharing of a learned model among coalition members without revealing data.
A two-network architecture learns intractable exponential family models.
problem Learning a model itself, not just optimizing parameters of a single distribution.
method Two-network architecture and optimization procedure for exponential family models.
result Accurately learns exponential family models, enabling generic operations.
A Bayesian approach to neural architecture search improves efficiency and accuracy.
problem Improving the efficiency and accuracy of neural architecture search.
method Formulating NAS from a Bayesian perspective, explicitly estimating the joint posterior distribution over architectures and weights, using Variational Dropout, and posterior-guided sampling.
result Posterior-guided NAS (PGNAS) achieves a good trade-off between precision and speed of search.
PARSEC uses a probabilistic approach to reduce memory usage in neural architecture search.
problem Efficiently search over large and complex neural architectures with reduced memory usage.
method Probabilistic sampling to learn a distribution over high-performing architectures, enabling transfer learning.
result Our approach outperforms state-of-the-art methods with significantly less computational cost.
In this work, we perform an exploratory study on synthesizing deep neural networks using biological synaptic strength distributions, and the potential influence of different distributions on modelling performance particularly for the scenario associated with small data sets. Surprisingly, a CNN with convolutional layer…
Paper introduces privacy-preserving deep-learning service.
problem Privacy loss in deep-learning-as-a-service.
method Homomorphic Encryption tailored for CNNs.
result Effectiveness of proposed privacy-preserving architecture.
BGANs enable multi-agent learning with distributed private datasets.
problem Learning from distributed private datasets in multi-agent scenarios.
method Brainstorming GAN architecture for fully distributed learning.
result BGANs generate high-quality data samples with lower JSD and FID.
This paper studies parallelization schemes for stochastic Vector Quantization algorithms in order to obtain time speed-ups using distributed resources. We show that the most intuitive parallelization scheme does not lead to better performances than the sequential algorithm. Another distributed scheme is therefore intro…
New insights suggest larger early layers improve CNN performance.
problem Common assumption of monotonously increasing feature counts in CNNs is challenged.
method Used a skew normal distribution to investigate feature amounts in CNN layers.
result Architectures favoring larger early layers yield better accuracy.
SNAS efficiently searches neural architectures using stochastic optimization.
problem Efficiently searching for optimal neural architectures.
method SNAS trains parameters of both neural operations and architecture distribution in a single round of backpropagation, using a novel search gradient and locally decomposable rewards.
result SNAS achieves state-of-the-art accuracy with fewer training epochs compared to other NAS methods.
Wide and Deep GNN learns from distributed graphs and retrain online.
problem Decentralized graph support changes over time, causing mismatch between training and testing graphs.
method Wide and Deep GNN architecture with distributed online learning.
result Convergence guarantees for online retraining of the wide part of the GNN.
Conventional Neural Architecture Search (NAS) aims at finding a single architecture that achieves the best performance, which usually optimizes task related learning objectives such as accuracy. However, a single architecture may not be representative enough for the whole dataset with high diversity and variety. Intuit…
Bayesian method learns neural network architecture parameters.
problem Estimating optimal neural network architecture parameters.
method Bayesian learning of concrete distributions over layer size and network depth.
result Regular networks with learnt structure generalize better on small datasets, while stochastic networks are more robust to initialisation.
Sparse transformer architecture improves accuracy and speed in generative modeling and inverse problems.
problem Improving accuracy and speed in generative modeling and inverse problems.
method Proposes a sparse transformer architecture using regularized Wasserstein proximal operator with L1 prior. result Sparse transformer achieves higher accuracy and faster convergence than classical methods.
Study reveals how neural network architectures bias their learning based on feature directions.
problem Understanding how neural network architectures bias their learning based on feature directions.
method Defined neural anisotropy directions (NADs) to encapsulate the directional inductive bias of architectures and provided an efficient method to identify them.
result NADs characterize the features used by CNNs to discriminate between different classes for the CIFAR-10 dataset.
Study shows LLMs can extrapolate rules from out-of-distribution prompts.
problem Understanding LLMs' ability to generalize from unexpected inputs.
method Formal languages and rule-based scenarios to evaluate LLMs' OOD behavior.
result LLMs can extrapolate rules from out-of-distribution prompts, even in complex scenarios.
Researchers analyze neural process architectures and their representational capacities.
problem Understanding what functions can be represented by different neural process architectures.
method Analyzing four types of neural process architectures: CNPs, ANPs, TNPs, and their latent variants.
result Prove these architectures form a strict hierarchy and characterize their representational capabilities.
This paper distills Bayesian posterior expectations for deep neural networks.
problem Improving deep neural network performance and uncertainty quantification.
method Develops a framework for distilling expectations from Bayesian posterior distributions using Monte Carlo samples.
result The framework successfully distills posterior predictive distribution and expected entropy.
New deep architecture models uncertainty by sharing neural connectivity patterns.
problem Uncertainty modeling in deep neural networks remains challenging.
method Proposes a new deep architecture that shares neural connectivity patterns between generative and discriminative networks to model a confounder.
result Demonstrates significant improvement in uncertainty estimation compared to state-of-the-art methods.
New neural network models extreme value distributions with preserved shape constraints.
problem Modeling multivariate extreme value distributions with preserved shape constraints.
method d-max-decreasing neural network architecture for non-parametric calibration and generation of MEVs.
result The proposed architecture approximates the dependence structure of MEVs at parametric rate and preserves essential shape constraints.
Theoretical study on how model architecture affects contrastive learning performance.
problem Understanding the role of model architecture in self-supervised learning.
method Theoretical analysis of contrastive learning, focusing on model capacity and clustering structures.
result Contrastive representations have lower dimensionality than the number of clusters in the data distribution.
This study analyzes NAS benchmarks and finds that only a subset of operations is crucial for generating high-performing architectures.
problem NAS benchmarks lack generability and provide skewed performance distributions, leading to unreliable comparisons.
method Empirical analysis of widely used NAS benchmarks (101, 201, TransNAS-Bench-101) focusing on operation importance and generability.
result Only a subset of operations is necessary to generate architectures close to the upper-bound performance range, and convolution layers have the highest impact.
FedNAS automates federated learning by searching for better architectures.
problem Non-I.I.D. data makes predefined model architectures suboptimal.
method Federated Neural Architecture Search (FedNAS) for collaborative architecture optimization.
result FedNAS searches for better architectures that outperform predefined models.
This research proves guarantees on sequence models' generalization to longer and novel sequences.
problem Generalization to longer sequences and novel token combinations in sequence models.
method Provable guarantees on length and compositional generalization for various sequence models.
result Limited capacity models achieve both length and compositional generalization with diverse training distributions.
The paper optimizes dynamic scheduling for ring architectures in deep learning training.
problem Optimizing deep learning training times with ring architectures.
method Formulated a non-convex, non-linear, NP-hard integer programming problem and developed a doubling heuristic.
result Dynamic scheduling can significantly reduce job completion times in ring architectures.
Efficient neural architecture search by sampling structure and operations.
problem Efficiently searching for optimal neural architectures.
method Decouples structure and operation search, using reinforcement learning with policy vectors.
result Significantly improved efficiency compared to traditional methods.
Paper outlines a system for ML models to learn continuously from evolving data.
problem Managing ML models in environments where data evolves.
method Describes a reference architecture for self-maintaining systems.
result Proposes a reference architecture for continual AutoML.
FiGS searches over a larger space of architectures for efficient mobile models.
problem Designing small, efficient deep networks for mobile devices.
method Differentiable search method using sparse regularization and Logistic-Sigmoid distribution.
result FiGS produces state-of-the-art parameter-efficient models on ImageNet and improves object detection performance.
New method finds minimal neural networks without weights.
problem Importance of weight parameters in neural networks.
method Search for minimal neural network architectures without explicit weight training.
result Minimal neural networks can perform tasks without weight training.
In-network learning outperforms Federated and Split learning in wireless networks.
problem Efficiently using distributed features for inference in wireless networks.
method Proposes 'in-network learning' architecture, uses neural networks for optimization, compares with Federated and Split learning.
result In-network learning offers better accuracy and bandwidth savings.
LeJEPA provides a scalable, theory-driven approach to self-supervised learning.
problem Lack of practical guidance and theory in JEPAs.
method Identified optimal Gaussian distribution and introduced SIGReg objective.
result LeJEPA achieves state-of-the-art performance with minimal hyperparameters and heuristics.
Dida learns meta-features invariant to feature permutations.
problem Learning invariant meta-features from dataset samples.
method Distribution-based deep architecture inheriting universal approximation and robustness.
result Dida outperforms state-of-the-art methods on dataset-level tasks.
Deep learning models predict generalization gaps without specific task or architecture.
problem Predicting when deep learning works across different tasks and architectures.
method Created a dataset of 13,500 neural networks trained on various spiral datasets and parameters. Used this dataset to train predictors for generalization gaps.
result DNNs and RNNs outperform linear models in predicting generalization gaps, with RNNs achieving R2=0.584. Evidential Softmax preserves multimodality in sparse probability distributions for generative models.
problem Sparse probability distributions in deep generative models make exact marginalization computationally intractable.
method Introduce ev-softmax, a sparse normalization function that preserves multimodality and can be trained with probabilistic loss functions.
result ev-softmax outperforms existing techniques in distributional accuracy and dimensionality reduction.
Unified framework denoises data and abstains from uncertain predictions.
problem Data quality and predictive uncertainty in deep neural networks.
method Unified filtering framework leveraging data density.
result Framework outperforms state-of-the-art techniques in denoising and abstaining.
We study the problem of learning conditional generators from noisy labeled samples, where the labels are corrupted by random noise. A standard training of conditional GANs will not only produce samples with wrong labels, but also generate poor quality samples. We consider two scenarios, depending on whether the noise m…
IMPACT improves RL training speed without sacrificing sample efficiency.
problem Limited sample efficiency in scalable RL architectures.
method Proposes IMPACT, extending IMPALA with target networks, circular buffers, and truncated importance sampling.
result IMPACT achieves higher rewards and significantly reduces training time compared to IMPALA.
Investigates deep hedging under rough volatility models.
problem Performance of deep hedging framework under non-Markovian conditions.
method Analysis of rough volatility models, use of parsimonious network architectures.
result Parsimonious network architectures can capture non-Markovian time-series.
This paper develops a new neural network architecture for modeling spatial distributions (i.e., distributions on R^d) which is computationally efficient and specifically designed to take advantage of the spatial structure of limit order books. The new architecture yields a low-dimensional model of price movements deep …
Transformers show better in-context learning resilience under distribution shifts than simple MLPs.
problem Understanding in-context learning under varying distribution shifts.
method Comparing transformers and set-based MLPs on linear regression tasks.
result Transformers better emulate OLS performance and exhibit better resilience to mild distribution shifts.
When using reinforcement learning (RL) algorithms to evaluate a policy it is common, given a large state space, to introduce some form of approximation architecture for the value function (VF). The exact form of this architecture can have a significant effect on the accuracy of the VF estimate, however, and determining…