Proposes a method to improve neural architectures reproducibly.
problem Lack of reproducibility in Neural Architecture Transformer (NAT).
method Differentiable Neural Architecture Transformation (DNAT).
result DNAT outperforms NAT and is applicable to various models and datasets.
GIT-Net uses neural networks to approximate PDE operators efficiently.
problem Approximating PDE operators for complex geometries.
method Parametrizes adaptive generalized integral transforms with deep neural networks.
result GIT-Net outperforms existing neural network operators in multiple areas.
Differentiable Window improves attention modules by enabling more focused attentions.
problem Improving attention mechanisms in neural networks.
method Proposes Differentiable Window, a neural module for dynamic window selection.
result Consistent and sizable improvements across various NLP tasks.
Transformer-based multi-scale model outperforms traditional methods in solving PDEs on irregular domains.
problem Solving partial differential equations on irregular domains using deep learning.
method Introduces Multi-Scale Attention Transformer (\msat{}) for solving PDEs.
result Achieves state-of-the-art generalization on complex geometry problems with significant speedup.
CViT learns complex physical systems using vision transformer techniques.
problem Learning maps between infinite-dimensional function spaces in scientific machine learning.
method Combines vision transformer encoder, grid-based coordinate embedding, and cross-attention mechanism.
result Achieves state-of-the-art performance on multiple benchmarks, often surpassing larger models.
NDM incorporates geometric structure into neural networks for better optimization and interpretability.
problem Efficient and interpretable deep learning architectures.
method NDM is a neural network architecture that explicitly incorporates geometric structure into its design, using a Coordinate Layer, Geometric Layer, and Evolution Layer.
result NDM provides intrinsic regularization, enhancing generalization and robustness.
Differentially-private FNAS protects privacy while collaboratively searching for neural architectures.
problem Collaborative neural architecture search with privacy concerns.
method Federated Neural Architecture Search (FNAS) with differential privacy (DP-FNAS).
result DP-FNAS can search for highly-performant neural architectures while protecting individual parties' privacy.
GATES improves neural architecture search by modeling operations as information transformation.
problem Improving predictor-based neural architecture search efficiency.
method GATES models operations as information transformation, covering both node and edge cell search spaces.
result GATES boosts sample efficiency and improves predictor performance.
A new method for discrete data normalizing flows using latent transformations.
problem Challenges in parameterizing bijective transformations for discrete data.
method Predict a distribution over latent transformations to make the marginal likelihood differentiable.
result Discrete-data normalizing flows can be trained using gradient-based learning with unbiased score function estimation.
Novel NAS method balances performance and hardware metrics efficiently.
problem Challenging multi-objective optimization in neural architecture search.
method Parameterizes joint architectural distribution via hypernetwork conditioned on hardware features and preferences.
result Zero-shot transferability to new devices with representative and diverse architectures.
Deep equilibrium models estimate latent variables from data.
problem Estimating latent variables from data.
method Generalized exponential family models, deep equilibrium networks.
result Deep equilibrium models solve MAP estimates for latent and transformation parameters.
Differentiable resampling improves particle filter performance.
problem Non-differentiability of traditional resampling in particle filters.
method Introduced a neural network resampler (particle transformer) trained with a likelihood-based loss function.
result Learned resampling outperforms traditional methods on synthetic and real-world tasks.
New model predicts neural network performance from early training epochs, incorporating architecture impact.
problem Predicting neural network performance from early training epochs, neglecting architecture impact.
method Architecture-aware graph ordinary differential equation model.
result Model outperforms state-of-the-art methods for MLP and CNN learning curves.
New findings show modern neural networks have finite sample complexity in o-minimal structures.
problem Understanding the learnability of modern neural networks in a broad context.
method Analyzing feedforward neural networks definable in o-minimal structures.
result Modern neural networks, including MLPs, CNNs, GNNs, and transformers, have finite sample complexity in the agnostic PAC setting.
How can prior knowledge on the transformation invariances of a domain be incorporated into the architecture of a neural network? We propose Equivariant Transformers (ETs), a family of differentiable image-to-image mappings that improve the robustness of models towards pre-defined continuous transformation groups. Throu…
We present Neural Random Forest Imitation - a novel approach for transforming random forests into neural networks. Existing methods propose a direct mapping and produce very inefficient architectures. In this work, we introduce an imitation learning approach by generating training data from a random forest and learning…
DNArch learns CNN architectures by backpropagation.
problem Discovering optimal CNN architectures.
method Differentiable Neural Architectures (DNArch) learns CNN architectures by backpropagation, controlling kernel sizes, channels, downsampling positions, and depth.
result DNArch finds performant CNN architectures across various tasks.
Framework predicts Navier-Stokes solutions on 2D domains using graph neural networks.
problem Predicting stationary Navier-Stokes solutions in non-parametrized 2D geometries.
method Graph-based multi-fidelity learning framework combining reduced-order models, Transformers, and Mamba architectures.
result Mamba architecture reduces computational cost while maintaining performance.
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.
einspace expands NAS search space to include diverse neural architectures.
problem NAS results are often limited to existing structures; new designs are rare.
method einspace uses a probabilistic context-free grammar to create a versatile search space.
result einspace discovers novel and improved architectures, including convolutions and attention.
Designing effective architectures is one of the key factors behind the success of deep neural networks. Existing deep architectures are either manually designed or automatically searched by some Neural Architecture Search (NAS) methods. However, even a well-searched architecture may still contain many non-significant o…
This work integrates differentiation and integration in Physics-Informed Neural Networks.
problem Solving integro-differential equations and computing integral transforms.
method Augmenting Physics-Informed Neural Networks with automatic integration.
result Solving complex integral transforms and integro-differential equations.
Transformer models perform slower than convolutional networks in learning hierarchical language structures.
problem Understanding how neural networks learn hierarchical language structures.
method Theoretical scaling laws and empirical validation of neural network performance.
result Convolutional networks outperform transformers in learning hierarchical language structures.
Efficient algorithm finds fast Transformer models.
problem Slow inference time of Transformer models.
method Decompose Transformer architecture into components, use sampling-based one-shot search.
result Achieved 10% to 30% speedup on pre-trained BERT and 70% on top of a previous state-of-the-art model.
New method adapts neural networks without losing prior knowledge.
problem Understanding and enabling flexible adaptation of neural networks.
method Differential geometry framework, functionally invariant paths (FIP).
result Achieves comparable state-of-the-art performance on continual learning and sparsification tasks.
HardCoRe-NAS finds fitting neural networks adhering to hard resource constraints.
problem Finding fitting neural networks that adhere to hard resource constraints.
method Accurate formulation of resource requirement and scalable search method.
result HardCoRe-NAS generates state-of-the-art architectures strictly satisfying hard resource constraints.
Gradients of neural networks can be computed efficiently for any architecture, but some applications require differential operators with higher time complexity. We describe a family of restricted neural network architectures that allow efficient computation of a family of differential operators involving dimension-wise…
A promising class of generative models maps points from a simple distribution to a complex distribution through an invertible neural network. Likelihood-based training of these models requires restricting their architectures to allow cheap computation of Jacobian determinants. Alternatively, the Jacobian trace can be u…
Graded Transformers embed algebraic structure in neural networks through graded transformations.
problem Efficiently modeling hierarchical and structured data in neural networks.
method Introduces Linearly Graded Transformer (LGT) and Exponentially Graded Transformer (EGT) with graded scaling operators.
result Establishes rigorous guarantees and improved efficiency for structured data.
Paper introduces a neural framework for accurate energy forecasting.
problem Challenges of forecasting energy demand and supply due to variability of renewable sources and dynamic consumption patterns.
method Integrates Neural ODEs, graph attention, multi-resolution wavelet transformations, and adaptive learning of frequencies.
result Consistently outperforms state-of-the-art baselines in various forecasting metrics across diverse datasets.
Differentiable NAS frameworks grow networks wider and deeper, revealing biases in wiring evolution.
problem Understanding the evolution of neural architecture wiring in differentiable NAS methods.
method Unified view on searching algorithms, local cost minimization, empirical and theoretical analyses.
result Implicit inductive biases cause observed searching patterns in differentiable NAS methods.
The scattering transform is a multilayered wavelet-based deep learning architecture that acts as a model of convolutional neural networks. Recently, several works have introduced generalizations of the scattering transform for non-Euclidean settings such as graphs. Our work builds upon these constructions by introducin…
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.
The paper explores deep learning through algebra and geometry, highlighting geometric structures and differential processes.
problem Understanding the geometric and algebraic foundations of deep learning.
method Investigates neural networks from perceptron to transformer, emphasizing geometric structures and differential processes.
result A coordinate-free formulation of backpropagation equations using canonical scalar products on matrix spaces.
Recurrent neural network (RNN)'s architecture is a key factor influencing its performance. We propose algorithms to optimize hidden sizes under running time constraint. We convert the discrete optimization into a subset selection problem. By novel transformations, the objective function becomes submodular and constrain…
A new method prunes neural networks efficiently without losing effectiveness.
problem Efficient pruning of neural networks without sacrificing performance.
method Deterministic approximation of binary gates and L0 regularization. result Pruning neural networks significantly without loss in effectiveness.
New neural stack and Turing Machine architectures prove stability and computational power.
problem Designing stable neural network architectures for Turing Machine simulation.
method Introducing neural stack and Turing Machine architectures, proving stability and computational equivalence.
result Differentiable nnTM with bounded neurons can simulate Turing Machine in real-time and is equivalent to UTM.
Fourier methods have a long and proven track record as an excellent tool in data processing. As memory and computational constraints gain importance in embedded and mobile applications, we propose to combine Fourier methods and recurrent neural network architectures. The short-time Fourier transform allows us to effici…
ISR creates analytical relationships from data via invertible maps.
problem Creating analytical relationships from datasets.
method Combines INNs and EQL, using invertible maps and sparsity promoting regularization.
result ISR can serve as a normalizing flow for density estimation and solve inverse problems.
New deep learning architecture learns martingales efficiently.
problem Efficiently learning martingales in financial derivatives pricing.
method High-order weak approximation algorithms of Runge-Kutta type.
result Deep neural networks based on this architecture learn martingales effectively.
Paper develops a new model for predicting volatility surface.
problem Predicting volatility in financial markets is challenging due to its non-observable nature and complex dynamics.
method Physics-informed convolutional transformer architecture.
result The new model outperforms other deep-learning architectures in predicting volatility surface.
Equivariance is a nice property to have as it produces much more parameter efficient neural architectures and preserves the structure of the input through the feature mapping. Even though some combinations of transformations might never appear (e.g. an upright face with a horizontal nose), current equivariant architect…
DANCE optimizes neural network and accelerator design for faster, more efficient DNN execution.
problem Challenges in optimizing neural network and accelerator design for efficient DNN execution.
method Differentiable approach to co-exploration of accelerator and network architecture design.
result Significantly shorter time to achieve superior accuracy and hardware cost metrics.
Alternatives to recurrent neural networks, in particular, architectures based on attention or convolutions, have been gaining momentum for processing input sequences. In spite of their relevance, the computational properties of these alternatives have not yet been fully explored. We study the computational power of two…
In this paper, we present a new comparative study on automatic essay scoring (AES). The current state-of-the-art natural language processing (NLP) neural network architectures are used in this work to achieve above human-level accuracy on the publicly available Kaggle AES dataset. We compare two powerful language model…
Unified method for deriving ridgelet transforms for various neural network architectures.
problem Deriving closed-form expressions for ridgelet transforms in modern neural network architectures.
method Unified Fourier slice method to derive ridgelet transforms for diverse neural network types.
result Systematic method to derive ridgelet transforms for various neural network architectures.
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.
PINNs solve differential geometry problems in complex shapes.
problem Solving differential geometry problems in complex shapes.
method Training neural networks with loss functions inspired by differential conditions.
result PINNs are effective for differential geometry problems.