New method uses fewer parameters to match state-of-the-art performance on multiple natural language tasks.
problem Efficiently adapting BERT for multiple tasks with fewer parameters.
method PALs (projected attention layers) for shared BERT model with task-specific parameters.
result Matches state-of-the-art performance on GLUE benchmark with 7 times fewer parameters.
Syllable-aware models perform similarly to character-based ones but use fewer parameters and train faster.
problem Improving word-level language modeling performance with syllable-based models.
method Used a syllable-aware neural language model with fewer parameters and faster training.
result Achieved comparable performance to character-based models but with 18%-33% fewer parameters and 1.2-2.2 times faster training.
Hierarchical CNNs improve image recognition with fewer parameters.
problem Difficulty in analyzing deep neural networks.
method Structured deep convolutional networks with progressively higher dimensional attributes learned from data.
result Hierarchical networks achieve comparable precision to state-of-the-art networks with fewer parameters.
GCRNNs improve graph problem solving with fewer parameters.
problem Graph process problems like earthquake epicenter identification and weather prediction.
method GCRNNs use convolutional filter banks and time-gated variations of GCRNNs (Gated GCRNNs) to improve performance.
result GCRNNs significantly improve performance over GNNs and another graph recurrent architecture.
Deep neural networks can approximate invariant/equivariant functions with fewer parameters.
problem Approximating functions that respect group symmetries with neural networks.
method Constructing deep neural networks with G-actions and G-equivariant/invariant affine transformations. result Deep neural networks can approximate G-invariant/equivariant functions with exponentially fewer parameters. Capsule networks improve performance on image classification tasks with fewer parameters.
problem Improving performance of capsule networks with fewer parameters.
method Inverted dot-product attention routing, Layer Normalization, concurrent iterative routing.
result Improves performance on benchmark datasets CIFAR-10 and CIFAR-100, and performs at-par with ResNet-18.
Quantum model generates financial data with fewer parameters.
problem Generating financial data with fewer parameters.
method Applied time-series quantum generative model to financial data.
result Fewer parameters required compared to classical methods.
DP-SGD can update fewer coordinates while maintaining privacy.
problem How to update fewer coordinates in DP-SGD without losing optimization signal.
method TP-TopK (Two-Phase TopK DP-SGD), a two-phase method for coordinate-sparse private training.
result Private training can update fewer coordinates without losing optimization signal, scaling noise with active dimension \(k\) instead of full dimension \(d\).
Improved NLP performance with fewer parameters and less data using conditional multi-task learning.
problem Challenges in transferring knowledge across different NLP tasks, including overfitting, forgetting, and negative transfer.
method Proposes a novel Transformer architecture with conditional attention and task-conditioned modules for efficient parameter sharing and mitigating forgetting.
result Achieves state-of-the-art performance on 26 NLP tasks with 66% less data and 50% fewer parameters compared to existing methods.
We describe a simple neural language model that relies only on character-level inputs. Predictions are still made at the word-level. Our model employs a convolutional neural network (CNN) and a highway network over characters, whose output is given to a long short-term memory (LSTM) recurrent neural network language mo…
Compact RNNs reduce parameters and improve efficiency.
problem High computational cost of RNNs with large inputs.
method Block-Term Tensor Decomposition (BT-TD) to reduce RNN parameters.
result BT-RNN achieves better accuracy and faster convergence than standard RNNs.
RDL-Net improves speech enhancement with fewer parameters and better performance.
problem Improving speech enhancement with fewer parameters and better performance.
method Proposes RDL-Net, a CNN combining residual and dense aggregations without over-allocating parameters.
result RDL-Net achieves higher speech enhancement performance with fewer parameters and lower computational requirements.
This work proves neural networks can learn certain function classes with fewer parameters than usual.
problem Understanding what neural networks can learn and why they don't overfit when overparameterized.
method Proved learning of concept classes in overparameterized neural networks using SGD.
result Overparameterized neural networks can learn with fewer parameters than usual.
Non-local U-Net improves biomedical image segmentation with fewer parameters and faster computation.
problem Efficiency and effectiveness in biomedical image segmentation using local operators.
method Proposed non-local U-Nets with flexible global aggregation blocks.
result Top performances achieved with fewer parameters and faster computation.
InSphereNet uses infilling spheres for 3D object classification, improving accuracy with fewer parameters.
problem 3D object classification using points, voxels, or images.
method Constructs infilling spheres from signed distance field (SDF) for classification.
result InSphereNet achieves superior accuracy with fewer inputs and parameters.
Bonsai-Net efficiently discovers state-of-the-art models with fewer parameters.
problem Efficiently discovering state-of-the-art neural architectures with minimal computational expense.
method Bonsai-Net uses a modified differential pruner to explore a relaxed search space.
result Bonsai-Net consistently discovers better architectures than random search with fewer parameters.
Neural networks fit fewer samples than their parameters suggest in practice.
problem Understanding the practical limitations of neural network flexibility.
method Examination of neural network optimization, parameter efficiency, and loss surfaces.
result Neural networks can only fit training sets with significantly fewer samples than their parameters suggest.
BT-Nets simplify DNNs by reducing parameter count.
problem Redundancy in DNN weights limits deployment to resource-constrained environments.
method Replaces FC-layers with BT-layers using block term decomposition.
result BT-layers achieve high compression ratios while maintaining performance.
Neural circuit model re-purposed for robotic control tasks.
problem Learning simple robotic control tasks.
method Re-purposing a biological neural circuit model to control robotic tasks using a search-based optimization algorithm.
result Neuronal Circuit Policies (NCPs) perform on par and in some cases surpass contemporary deep learning models with fewer parameters and interpretable dynamics.
New method learns compact policies with fewer parameters for faster inference.
problem Training efficient policies for real hardware with limited resources.
method Structured random orthogonal matrices for gradient approximation.
result Compact policies with up to 13x fewer parameters than ES.
Quaternion CNNs improve speech recognition with fewer parameters.
problem Efficient end-to-end speech recognition with minimal parameters.
method Integrating quaternion algebra into CNNs for speech feature processing.
result Quaternion CNNs achieve lower phoneme error rates with fewer parameters.
Pixel-space diffusion models outperform latent models on high-resolution image synthesis.
problem Efficiency and quality trade-off in high-resolution image synthesis.
method Sigmoid loss-weighting, simplified architecture, and resolution scaling.
result Achieved 1.5 FID on ImageNet512, new SOTA results on other datasets.
Compact model uses RBMs for sequence classification with fewer parameters.
problem Sequence classification with dynamic models and complex neural networks.
method Rolling RBMs over time for representation learning and temporal inference.
result Outperforms state-of-the-art models in melody and character recognition.
Study shows DNNs can recover functions with fewer samples than model parameters at overparameterization.
problem Determining reliable function recovery in overparameterized deep neural networks.
method Introducing 'local linear recovery' (LLR) and proving upper bounds on sample sizes for recovery.
result Upper bounds on optimistic sample sizes for function recovery in overparameterized DNNs are achieved.
dAUTOMAP scales AUTOMAP by decomposing domain transformation.
problem Inadequate scalability limits AUTOMAP's practicality.
method Decomposes AUTOMAP's domain transformation for linear scalability.
result dAUTOMAP outperforms AUTOMAP with fewer parameters.
Study proposes Local Linear Encoding for better feature discretization.
problem Improving feature discretization for numeric data.
method Theoretical analysis and Local Linear Encoding (LLE) method.
result LLE outperforms conventional methods with fewer parameters.
New framework uses symmetry-based matrices for efficient, flexible NNs.
problem Designing neural networks with relaxed equivariance.
method Symmetry-based structured matrices, Group Matrices (GMs).
result GMs enable competitive performance with fewer parameters.
Lean 2-layer RBMs achieve similar representational power as single-layer RBMs with fewer parameters.
problem Understanding and quantifying the representational power of multi-layer RBMs.
method Inherent Structure Capacity (ISC) and Lean RBMs.
result 2-layer RBMs can achieve the same representational power as single-layer RBMs with fewer parameters.
BayesSum improves Bayesian quadrature for discrete domains, requiring fewer samples.
problem Estimating intractable expectations over discrete domains.
method BayesSum is a Bayesian quadrature extension for discrete domains, leveraging prior information through Gaussian processes.
result BayesSum requires fewer samples than Monte Carlo, achieving faster convergence rates.
Higher granularity in MoE models boosts expressivity exponentially.
problem Expressivity of Mixture-of-Experts models with varying granularity.
method Comparing models with different numbers of active experts (granularity).
result Exponential separation in network expressivity based on granularity.
Study improves fractional posterior for 1-bit matrix completion.
problem Estimating a binary matrix from observed entries.
method Fractional posterior approach with low-rank factorization and spectral scaled Student priors.
result Concentration results for fractional posterior, demonstrating effectiveness in matrix recovery.
SHVC improves image compression with fewer parameters.
problem Challenges in VAE compression, especially with bits-back coding.
method Introduces autoregressive sub-pixel convolution and autoregressive initial bits.
result Achieves state-of-the-art compression performance with fewer model parameters.
Estimates mode of discrete distributions with fewer samples.
problem Identifying the mode of a discrete distribution with high probability.
method Generalizes PPR martingale confidence sequences to handle multiple modes.
result PPR-1v1 stopping rule is asymptotically optimal and significantly more efficient.
BaM improves BBVI by optimizing a score-based divergence, leading to faster convergence.
problem Slow convergence of black-box variational inference methods.
method Batch and match (BaM) approach based on a score-based divergence.
result BaM converges exponentially quickly to the target mean and covariance.
Paper proposes using backpropagation for probabilistic program learning.
problem Difficult to learn probabilistic models from data.
method Learning parameters of a probabilistic program using backpropagation.
result Trains probabilistic models similar to neural networks.
New method finds smaller networks with similar performance to large models in fewer epochs.
problem Training large neural networks is computationally expensive and energy-intensive.
method Use PCA to identify a basis of hidden layer activations and reduce network parameters.
result Principal Component Networks (PCNs) can train faster and use less energy than overparameterized models without accuracy loss.
Internal node bagging uses dropout-like training to improve model fitting with fewer parameters.
problem Improving model fitting with fewer parameters.
method Explicitly forces a group of nodes to learn a certain feature, combining them in inference time.
result Internal node bagging performs significantly better than dropout on small models.
Quaternion Conformer GAN (QC-GAN) is a parameter-efficient speech enhancement framework that combines a Quaternion Conformer generator with MetricGAN-based training.
problem Speech Enhancement
method Quaternion Conformer GAN
result Achieved a PESQ score of 3.48 with 0.89M parameters, comparable to state-of-the-art models at less than half their size.
Express Wavenet reduces neural network parameters to 1% of standard networks.
problem Optical neural networks with high parameter count.
method Wavelet modulation, random shift wavelets, expressway structure.
result Express Wavenet achieves high accuracy with significantly fewer parameters.
A low-rank tensor model simplifies multi-dimensional Markov chains.
problem Simplifying the dynamics of multi-dimensional Markov chains.
method Low-rank tensor decomposition for multi-dimensional state spaces.
result Our tensor model requires fewer parameters and samples than conventional methods.
Efficiently learns 3D convolutions with less data.
problem High parameter and data costs in 3D convolutions.
method Temporal factorization of 3D kernels.
result Significantly reduces training data requirement and parameter count.
Fewer data weight updates lead to faster convergence in machine learning models.
problem Improving robustness of machine learning models through data mixing.
method Analyzing convergence behavior of data mixing with a finite number of inner steps.
result The optimal number of inner steps scales with the budget and type of gradients used.
Kernelised flows improve density estimation and generation with fewer parameters.
problem Limited expressiveness of flow-based models due to invertibility constraints.
method Integrates kernels into normalising flows to enhance expressiveness and efficiency.
result Kernelised flows outperform neural network-based flows in parameter efficiency and low-data scenarios.
We show how the success of deep learning could depend not only on mathematics but also on physics: although well-known mathematical theorems guarantee that neural networks can approximate arbitrary functions well, the class of functions of practical interest can frequently be approximated through "cheap learning" with …
Proposes a neural network for efficient imbalance electricity price forecasting.
problem Accurate and efficient imbalance electricity price forecasting in industrial energy trading systems.
method Market-rule-informed neural network framework.
result The proposed model achieves competitive forecasting performance with fewer parameters and shorter training time.
Paper improves natural language understanding with less data using a new training method.
problem Limited data hinders performance of small models in natural language tasks.
method Generation-Distillation: uses large finetuned models to generate new training data and distill knowledge into smaller models.
result Achieves comparable performance to BERT with 300x fewer parameters and outperforms prior distillation methods.
Adjoined Networks trains both base and compressed networks together for efficient model compression.
problem Efficiently compressing deep neural networks while maintaining accuracy.
method Adjoined Networks (AN) trains both a base network and a smaller compressed network simultaneously, sharing parameters.
result AN achieves 71.8% top-1 accuracy with 1.8M parameters and 1.6 GFLOPs on ImageNet.
Wavelets help compress neural networks efficiently.
problem Efficiently compressing linear layers in neural networks.
method Learnable wavelet transforms to compress RNNs.
result Wavelet compressed RNNs have fewer parameters and perform competitively.