Deep neural networks as image priors have been recently introduced for problems such as denoising, super-resolution and inpainting with promising performance gains over hand-crafted image priors such as sparsity and low-rank. Unlike learned generative priors they do not require any training over large datasets. However…
On-device research index
arXiv research
A locally-built, LLM-digested index of recent arXiv papers in quant finance, geometry/topology, and statistical ML — keyword search served straight from SQLite on this machine.
168,742 papers · 148 categories
Trend · papers per month
48 results for “untrained neural networks”
Study how untrained policies explore in RL environments.
problem Challenges in reinforcement learning, especially sparse or adversarial reward structures.
method Theoretical and empirical analysis of untrained deep neural policies in a toy model.
result Untrained policies generate correlated actions and non-trivial state-visitation distributions.
A new neural network model MDRBM improves noise-robustness in classification.
problem Improving noise-robustness in classification problems.
method Stacking a probabilistic-ELM layer on a discriminative restricted Boltzmann machine (DRBM).
result MDRBM outperforms other models, especially in noise-robustness.
Employing deep neural networks as natural image priors to solve inverse problems either requires large amounts of data to sufficiently train expressive generative models or can succeed with no data via untrained neural networks. However, very few works have considered how to interpolate between these no- to high-data r…
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.
Untrained neural networks can unfairly assign predictions to the same class.
problem Biasing effects in neural networks during initial training phases.
method Theoretical analysis of deep neural networks, focusing on Initial Guessing Bias (IGB).
result Model structure and preprocessing methods influence IGB.
Survey of deep learning methods for inverse problems, highlighting theoretical challenges.
problem Addressing theoretical gaps in deep learning for inverse problems.
method Survey of existing theoretical developments and open problems.
result Highlighting ongoing challenges in deep learning for inverse problems.
Automates neural network design without training, speeding up search by seconds.
problem Time and effort in hand-designing deep neural networks.
method Predict trained accuracy from untrained network state using activation overlap.
result Search for powerful networks in seconds, verified on various benchmarks.
RWTNs improve NTN performance in SRL tasks.
problem Improving NTN performance in SRL tasks.
method Randomly weighted, untrained tensors in NTN encoder network.
result RWTNs outperform LTNs in SII tasks.
Deep neural networks, in particular convolutional neural networks, have become highly effective tools for compressing images and solving inverse problems including denoising, inpainting, and reconstruction from few and noisy measurements. This success can be attributed in part to their ability to represent and generate…
This research uncovers high-performing subnetworks in deep GNNs without training.
problem Challenges in applying SLTH to deeper GNNs with high memory requirements.
method Introduces Multicoated Supermasks (M-Sup) and Multi-Stage Folding for GNNs.
result Uncovered untrained recurrent networks with performance similar to trained models.
Explains neural network properties with a simple operation.
problem Understanding neural network performance and generalization.
method Analyze a single operation: expand-and-sparsify.
result Clarifies phenomena like Lottery Ticket Hypothesis and overparameterized models.
Ring-reservoir networks simplify graph embeddings efficiently.
problem Efficient graph embeddings using deep neural networks.
method Progressive simplification of Reservoir Computing models to ring topology.
result Ring-reservoir networks show consistent advantages in predictive performance.
Extends branch and bound for probabilistic neural network verification.
problem Probabilistic verification of neural networks.
method Split preactivations, use linear bounds, divide problems iteratively.
result Sound and complete for feedforward-ReLU networks.
Study on MC dropout in wide neural networks and its convergence to Gaussian processes.
problem Understanding the behavior of Monte Carlo dropout in wide neural networks.
method Rigorously studied the limiting distribution of wide untrained NNs under dropout, proving convergence to Gaussian processes. Investigated correlations and non-Gaussian behavior in finite width NNs.
result Wide untrained neural networks under dropout converge to Gaussian processes for fixed sets of weights and biases.
Deep neural networks progressively transform their inputs across multiple processing layers. What are the geometrical properties of the representations learned by these networks? Here we study the intrinsic dimensionality (ID) of data-representations, i.e. the minimal number of parameters needed to describe a represent…
Efficient deep GNNs achieve state-of-the-art performance without training.
problem Efficiency issue in deep graph neural networks.
method Representing graphs as fixed points of dynamical systems and using small, sparse, untrained recurrent networks.
result Small deep GNNs without training can achieve or improve state-of-the-art performance.
The weight initialization and the activation function of deep neural networks have a crucial impact on the performance of the training procedure. An inappropriate selection can lead to the loss of information of the input during forward propagation and the exponential vanishing/exploding of gradients during back-propag…
The paper examines when NTK theory applies to real finite-width neural networks.
problem Understanding when NTK theory accurately predicts the behavior of finite-width neural networks.
method Empirical study of fully-connected ReLU and sigmoid DNNs with various hyperparameters and depths.
result NTK theory does not always apply to sufficiently deep networks with exploding gradients, and the kernel changes significantly during training.
Deep learning methods improve overlapping speaker separation across languages and noise.
problem Overlapping speaker separation in realistic scenarios.
method Deep clustering and deep attractor networks.
result Deep learning methods are effective for a broad range of languages and can handle untrained languages with common features.
Learning three data points can generate all types of periodic orbits in a neural network.
problem Can learning three data points generate all types of periodic orbits in a neural network?
method Investigated a continuous one-dimensional map with period three in a random neural network in its thermodynamic limit.
result Almost all learned periods are unstable, and each network has its own characteristic attractors.
Modified RV-coefficient reveals how training affects neural network representations.
problem Understanding how training affects intermediate representations in convolutional neural networks.
method Experimented with modified RV-coefficient (RV2) to compare activation patterns in deep networks trained on varying amounts of data and layers.
result RV2 successfully recovered expected similarity patterns and provided interpretable similarity matrices.
Paper proposes efficient network pruning method for deep neural networks.
problem High computational and memory cost of deep neural networks.
method Annealing and direct sparsity control for channel-level pruning.
result Proposed method achieves better or competitive performance compared to other methods.
Network pruning is a promising avenue for compressing deep neural networks. A typical approach to pruning starts by training a model and then removing redundant parameters while minimizing the impact on what is learned. Alternatively, a recent approach shows that pruning can be done at initialization prior to training,…
Observing how deep neural networks understand physics through the energy spectrum of one-dimensional quantum mechanicsphysics.comp-ph
NNs accurately predict energy eigenvalues and other physical phenomena in 1D quantum mechanics.
problem Understanding how neural networks interpret physics.
method Training NNs to predict energy eigenvalues from potentials and testing their ability to generalize.
result NNs can predict physical phenomena not learned during training, indicating a new way of understanding physics.
A new method reduces memory usage for deep neural networks ensembles.
problem Reducing memory footprint of ensemble methods for deep learning.
method Extract multiple sub-networks from a single neural network through end-to-end optimization.
result Achieves higher or comparable accuracy with significantly less memory usage.
Deep Information Propagationstat.ML
We study the behavior of untrained neural networks whose weights and biases are randomly distributed using mean field theory. We show the existence of depth scales that naturally limit the maximum depth of signal propagation through these random networks. Our main practical result is to show that random networks may be…
Randomized neural networks use fixed connections for efficiency.
problem Efficiency in deep learning models.
method Fixed connections in neural networks.
result Deep randomized neural networks achieve state-of-the-art results.
A new method trains deep networks by separating weight locations from values.
problem Training deep networks efficiently and effectively.
method Lookahead Permutation (LaPerm) to train DNNs by reconnecting weights.
result LaPerm can train DNNs with random and dense, sparse, or single-valued initial weights.
Proposes DMOC for more nuanced neural network robustness.
problem Lipschitz continuity is too coarse for nuanced data-dependent behavior.
method Data-driven, architecture-agnostic framework based on DMOC.
result DMOC provides a finer notion of robustness relative to data distribution.
AGOP mechanism explains deep neural collapse in neural networks.
problem Explaining the rigid structure of data representations in deep neural networks.
method Introducing AGOP and Deep RFM to demonstrate DNC.
result AGOP mechanism causes deep neural collapse in neural networks.
Eigenvalue Distribution of Large Random Matrices Arising in Deep Neural Networks: Orthogonal Casestat.ML
The paper proves a distribution claim for neural network Jacobians.
problem Distribution of singular values in deep neural networks.
method Free probability and random matrix theory techniques.
result Singular value distribution matches for specific cases.
The weight initialization and the activation function of deep neural networks have a crucial impact on the performance of the training procedure. An inappropriate selection can lead to the loss of information of the input during forward propagation and the exponential vanishing/exploding of gradients during back-propag…
Feed-forward networks are widely used in cross-modal applications to bridge modalities by mapping distributed vectors of one modality to the other, or to a shared space. The predicted vectors are then used to perform e.g., retrieval or labeling. Thus, the success of the whole system relies on the ability of the mapping…
New proof links initial class bias to DNN trainability, challenging traditional understanding.
problem Understanding the initial class bias in DNNs and its impact on trainability.
method Theoretical proof linking initial class bias to mean field theories of DNNs.
result Efficient learning is connected to a network's prejudice towards a specific class, contradicting traditional understanding.
Researchers develop PAIN to improve self-driving safety through adversarial training.
problem Overfitting and poor generalizability of neural networks in self-driving vehicles.
method PAIN combines adversarial training in CARLA simulation to generate edge cases.
result Trained self-driving vehicles are more resilient to environmental uncertainty and less prone to collisions.
Bayesian deep ensembles improve prediction accuracy in various settings.
problem Improving prediction accuracy of deep ensembles in out-of-distribution settings.
method Introducing a randomised, untrainable function to each ensemble member, enabling a posterior predictive distribution interpretation.
result Bayesian deep ensembles make more conservative predictions and outperform standard ensembles in various tasks.
Deep reinforcement learning has been successful in a variety of tasks, such as game playing and robotic manipulation. However, attempting to learn \textit{tabula rasa} disregards the logical structure of many domains as well as the wealth of readily available knowledge from domain experts that could help "warm start" t…
New algorithm finds important synapses without training data.
problem Finding important synapses in neural networks without data.
method Iterative Synaptic Flow Pruning (SynFlow) based on conservation law.
result Algorithm consistently outperforms existing pruning methods.
Reconstructing signature features from randomized vector fields in differential equations.
problem Reconstructing signature features from controlled differential equations with random vector fields.
method Using controlled ordinary differential equations driven by continuous bounded variation curves, the study explores the extent to which signature features can be reconstructed from the non-linear flow of these equations.
result The number of signature features that can be reconstructed from the non-linear flow of controlled ordinary differential equations with random vector fields is exponential in the hidden dimension, under certain conditions.
One pixel modification can make deep models unlearnable.
problem Protecting data from unauthorized training of deep neural networks.
method Perturbing only one pixel in each image to degrade model accuracy.
result Generated One-Pixel Shortcut (OPS) cannot be erased by adversarial training and strong augmentations.
CSTs improve stability in covariance spectrum analysis without training.
problem Stability and expressiveness in covariance spectrum analysis.
method Sequential application of covariance wavelet filters to input data.
result Stable and expressive hierarchical representations in low-data settings.
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…
Deep learning tackles low-photon nanoscale holographic phase retrieval.
problem Low-photon imaging challenges at nanoscale.
method Dataset-free deep learning framework with physical model integration.
result Significantly improves signal recovery from higher noise levels.
Dealing with uncertainty is essential for efficient reinforcement learning. There is a growing literature on uncertainty estimation for deep learning from fixed datasets, but many of the most popular approaches are poorly-suited to sequential decision problems. Other methods, such as bootstrap sampling, have no mechani…
The recent "Lottery Ticket Hypothesis" paper by Frankle & Carbin showed that a simple approach to creating sparse networks (keeping the large weights) results in models that are trainable from scratch, but only when starting from the same initial weights. The performance of these networks often exceeds the performance …
We present a method for translating music across musical instruments, genres, and styles. This method is based on a multi-domain wavenet autoencoder, with a shared encoder and a disentangled latent space that is trained end-to-end on waveforms. Employing a diverse training dataset and large net capacity, the domain-ind…
Reservoir Computing (RC) provides an efficient way for designing dynamical recurrent neural models. While training is restricted to a simple output component, the recurrent connections are left untrained after initialization, subject to stability constraints specified by the Echo State Property (ESP). Literature condit…