Research
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.

169,291 papers · 148 categories

Trend · papers per month

69137206274 · Jun 202019922001200920182026
48 results for fully-differentiable architecture

Paper discovers differential equations from data using neural networks and Bayesian methods.

problem Discovering differential equations from datasets using machine learning.
method Integrates neural network-based surrogates with Sparse Bayesian Learning (SBL).
result Proposes a robust model discovery algorithm and a Physics Informed Normalizing Flow (PINF).

A new method to simplify deep neural networks by removing unnecessary parts.

problem Overly complex deep neural networks require significant resource investment for size reduction.
method A fully differentiable sparsification method that optimizes a regularized objective function with stochastic gradient descent.
result The method can learn both the sparsified structure and weights of a network in an end-to-end manner.

New neural KB representation speeds up reasoning with large symbolic knowledge bases.

problem Efficiently reasoning with large symbolic knowledge bases.
method Sparse-matrix reified knowledge base, enabling fully differentiable, scalable neural modules.
result Competitive performance on KB completion and semantic parsing benchmarks.

ZM-Net efficiently manipulates images with unseen signals in real-time.

problem Efficiently alter images with diverse guiding signals (e.g. paintings, attributes).
method Proposes ZM-Net, a fully-differentiable architecture that jointly optimizes TNet and PNet.
result ZM-Net performs high-quality image manipulation in real-time (tens of milliseconds per image) for unseen signals.

New method recovers transportable DAG structures from different datasets.

problem Inference of DAG structures is computationally expensive and lacks transportability.
method Introduces D-Struct, a differentiable architecture that recovers transportable DAG structures.
result D-Struct recovers transportable DAG structures from different datasets.

DBQ quantizes lightweight networks efficiently for resource-constrained devices.

problem High computational and storage complexity of deep neural networks on resource-constrained devices.
method A differentiable non-uniform quantizer that can be mapped onto efficient ternary-based dot product engines.
result Achieves state-of-the-art results with minimal training overhead and best accuracy-complexity trade-off.

New method processes large images with sampling and attention.

problem Computational and memory constraints in processing large images.
method Fully differentiable end-to-end trainable model that samples and processes only a fraction of the input image.
result Sampling from attention distribution results in an unbiased estimator of the full model with minimal variance.

New model combines physics and machine learning for ocean dynamics.

problem Discovering hidden laws governing ocean dynamics.
method Develops Deep Neural Numerical Models (DNNMs) to learn hidden variables of physical laws.
result Illustrates DNNMs applied to Sea Surface Height dynamics, connecting to QG model.

Model predicts multi-agent trajectories using a differentiable simulator.

problem Predicting future positions of multiple interacting agents.
method Conditional recurrent variational neural networks (CVRNNs) with a kinematic bicycle model.
result Achieves state-of-the-art results on INTERACTION dataset.

Proposes a new layer for efficient 3D shape discrimination.

problem Irregular structure and redundancy in 3D point clouds hinder efficient inter-class discrimination.
method Integrates Blended Convolution and Synthesis layer that projects and synthesizes 3D point clouds, followed by 3D convolution in the unit ball.
result End-to-end architecture achieves compelling results on 3D shape recognition and retrieval.

AEN-SAEs address feature starvation in sparse autoencoders by stabilizing the geometric alignment of sparse coding.

problem Feature starvation in sparse autoencoders, leading to unstable and misaligned representations.
method Adaptive Elastic Net SAEs (AEN-SAEs) combine 2\ell_2 and 1\ell_1 terms to stabilize the sparse coding map and control feature interactions.
result AEN-SAEs mitigate feature starvation without heuristic resampling, maintaining competitive reconstruction abilities.

We introduce the value iteration network (VIN): a fully differentiable neural network with a `planning module' embedded within. VINs can learn to plan, and are suitable for predicting outcomes that involve planning-based reasoning, such as policies for reinforcement learning. Key to our approach is a novel differentiab…

2016-02-09abs ↗pdf ↗

CoDeQ simplifies joint model compression by integrating pruning and quantization.

problem Joint pruning and quantization methods are complex and require additional procedures.
method CoDeQ uses a dead-zone quantizer to directly induce sparsity and learn quantization parameters.
result CoDeQ achieves high sparsity and low-precision accuracy with minimal bit operations.

Gradient estimation techniques applied to programs with randomness in high energy physics.

problem Differentiating programs with discrete randomness in high energy physics.
method Several gradient estimation techniques, including Stochastic AD method, applied to simplified detector design experiments.
result Development of the first fully differentiable branching program.

PIED optimizes experimental design for inverse problems using physics-informed neural networks.

problem Optimizing experimental design for inverse problems with limited budget and constraints.
method PIED uses physics-informed neural networks (PINNs) for continuous optimization of design parameters in one-shot deployments.
result PIED significantly outperforms existing ED methods in solving inverse problems, including unknown functions.

Lat-Net compresses Lattice Boltzmann simulations using neural networks.

problem High computational and memory demands in fluid simulations.
method Convolutional autoencoders and residual connections in a fully differentiable scheme.
result Computational and memory efficient neural network that generalizes to large grid sizes and complex geometries.

GeoPhy uses geometric gradients to efficiently infer phylogenetic trees from molecular data.

problem Challenges in accurately inferring species relationships from molecular data due to combinatorially vast tree topologies.
method Introduces a novel, fully differentiable formulation of phylogenetic inference using geometric spaces and variational Bayesian methods.
result Significantly outperforms other approximate Bayesian methods in inferring phylogenetic trees.

NeuralSort optimizes sorting networks using continuous relaxations.

problem Non-differentiability of sorting operator hinders gradient-based optimization.
method Continuous relaxation of sorting operator to unimodal row-stochastic matrices, enabling gradient-based optimization.
result Gradient-based stochastic optimization over permutations is achieved.

This work combines GANs and A3C for high-resolution image compression.

problem Image compression for high-resolution images without loss of quality.
method Hybrid approach using GANs and A3C for end-to-end learning.
result Improves PSNR for high-resolution images through end-to-end learning.

Physics-informed neural networks solve PDEs using neural networks.

problem Solving nonlinear partial differential equations (PDEs) with neural networks.
method Physics-informed neural networks trained to solve PDEs while respecting physical laws.
result Physics-informed neural networks can infer solutions to PDEs and create differentiable surrogate models.

CompILE learns reusable segments from demonstrations for hierarchical task execution.

problem Learning reusable, variable-length segments of hierarchical behavior from demonstrations.
method Unsupervised, fully-differentiable sequence segmentation module for latent encoding and re-composition.
result Model generalizes to longer sequences and unseen environments, learns task boundaries and event encodings.

A new Gaussian process regression method infers implicit manifold structure from data.

problem Scaling Gaussian process regression to high-dimensional data.
method Proposes a fully differentiable Gaussian process regression technique that infers implicit manifold structure from data.
result Improves predictive performance and calibration of standard Gaussian process regression in high-dimensional settings.

A new method for differentiable structured sparsity improves neural network performance and sparsity.

problem Non-differentiability of structured sparsity penalties in neural networks.
method Introducing DD-Gating, a differentiable approach to structured overparameterization.
result The DD-Gating objective converges to the L2,2/DL_{2,2/D}-regularized loss and induces sparse learning dynamics.

Researchers develop a method to interpret GNNs by identifying unnecessary edges in NLP models.

problem Understanding which parts of graphs contribute to NLP model predictions.
method A post-hoc method using differentiable edge masking to identify and drop unnecessary edges.
result Large proportions of edges can be dropped without affecting model performance, providing insights into model predictions.

A new clustering method using deep neural networks with size constraints.

problem Clustering high-dimensional data like images, especially when similarity is not well captured by Euclidean distance.
method Rewriting kk-means as an optimal transport task, adding entropic regularization, and introducing constraints on cluster sizes.
result The proposed method outperforms state-of-the-art clustering methods in unsupervised accuracy.

Topology-enhanced loss improves 3D object reconstruction from 2D images.

problem Challenges in reconstructing 3D objects from 2D images, especially capturing shape information.
method Integrates multi-scale topological features into the reconstruction loss using cubical complexes and optimal transport distance.
result Topology-aware loss substantially improves 3D reconstruction quality.

A new method speeds up computation of Sinkhorn divergences to linear time.

problem Expensive computation of Sinkhorn divergences for comparing probability distributions.
method Using positive features to approximate ground costs, reducing computation time to linear.
result Sinkhorn divergences can be computed in linear time, scaling as O(nr).

GDM models time series with smoother transitions and interpretable states.

problem Capturing smooth, variable-speed transitions and stochastic mixtures of states.
method Introduces a continuous relaxation of discrete states and a Gumbel noise model.
result Models real-world datasets more faithfully with smoother dynamics and interpretable states.

Proposes differentiable and sparse top-k operators for neural networks.

problem Discontinuity of top-k operator makes it unsuitable for end-to-end training with backpropagation.
method Formulates top-k as a linear program over permutahedron, introduces p-norm regularization, and uses isotonic optimization.
result Successfully applied to neural network pruning, fine-tuning, and routing.

New aggregation method improves GNN robustness to structural perturbations.

problem Graph Neural Networks (GNNs) are vulnerable to adversarial attacks that manipulate graph structure.
method Proposes a robust aggregation function with a breakdown point of 0.5, inspired by robust statistics.
result Improves GNN robustness by a factor of 3 on Cora ML and 5.5 on Citeseer, and 8 for low-degree nodes.

New algorithm trains policies using a model-based adversarial approach.

problem Training policies in environments with limited interactions and high variance gradients.
method Introduces Model based Adversarial Imitation Learning (MAIL) algorithm using a forward model.
result Initial results surpass current state-of-the-art in MuJoCo physics simulator.