Paper improves DNN accelerator robustness against bit errors with energy savings.
problem Bit errors in quantized DNN weights reduce energy efficiency.
method Combines robust fixed-point quantization, weight clipping, and random bit error training.
result Significantly improves robustness against random bit errors with high energy savings.
Improved 2-bit covariance estimator with reduced operator norm error and no tuning needed.
problem Improving 2-bit covariance estimation with reduced operator norm error and no tuning needed.
method Proposed a new 2-bit covariance matrix estimator using triangular dithering scales.
result Improved operator norm error rate that depends on effective rank of covariance matrix, closing theoretical gap.
Paper offers robust recovery for 1-bit sensing with partial Gaussian circulant matrices.
problem Accurately recovering vectors from 1-bit measurements using structured matrices.
method Correlation-based optimization with randomly signed partial Gaussian circulant matrices and generative models.
result Recovery guarantees match those for i.i.d. Gaussian matrices but with faster computation.
Bayesian Bits unifies quantization and pruning through gradient optimization.
problem Joint mixed precision quantization and pruning for efficient neural networks.
method Gradient-based optimization with a novel bit width decomposition and learnable stochastic gates.
result Bayesian Bits achieves better accuracy vs. efficiency trade-off compared to static bit width networks.
Improves matrix multiplication throughput for asymmetric bit-width operands.
problem Matrix multiplications between asymmetric bit-width operands, especially 8- and 4-bit, are not efficiently handled by existing SIMD instructions.
method Proposes a new SIMD matrix multiplication instruction that uses mixed precision on inputs (8- and 4-bit) and accumulates into 16-bit output, improving throughput.
result Offers 2x improvement in throughput compared to existing symmetric-operand-size instructions, with negligible overflow.
Implementing large-scale deep neural networks with high computational complexity on low-cost IoT devices may inevitably be constrained by limited computation resource, making the devices hard to respond in real-time. This disjunction makes the state-of-art deep learning algorithms, i.e. CNN (Convolutional Neural Networ…
Low-bit training framework reduces energy consumption in CNNs.
problem Reducing energy consumption in convolutional neural networks.
method Low-bit training framework using MLS tensor format with dynamic quantization.
result Achieves superior trade-off between accuracy and bit-width.
State of the art deep learning models have made steady progress in the fields of computer vision and natural language processing, at the expense of growing model sizes and computational complexity. Deploying these models on low power and mobile devices poses a challenge due to their limited compute capabilities and str…
This paper proposes a learning framework for n-bit quantized neural networks that improves accuracy and speed on FPGAs.
problem Efficiently implementing quantized neural networks on FPGAs to maintain accuracy and speed.
method A novel learning framework for n-bit QNNs, constrained weights, reconstructed gradient function, n-BQ-NN structure, and SVPE array.
result Quantized models achieve almost the same accuracy as full-precision models and outperform typical low-precision QNNs.
Paper optimizes KWS models using NAS and quantization for limited resources.
problem Developing efficient keyword spotting models in resource-constrained environments.
method Neural Architecture Search (NAS) for model structure optimization and quantization of weights and activations.
result Achieved high accuracy (95.55%) with minimal parameters and operations using NAS and quantization.
WrapNet optimizes inference for low-resolution neural networks by using 8-bit additions.
problem Reducing multiplication complexity in low-resolution neural networks.
method Adapting neural networks to use low-resolution (8-bit) additions in accumulators, with a cyclic activation layer and overflow penalty regularizer.
result Achieves comparable classification accuracy to 32-bit counterparts using low-resolution additions.
DJPQ optimizes neural network pruning and quantization for hardware efficiency.
problem Efficiently compress neural networks for hardware inference.
method Joint gradient-based optimization of pruning and quantization into a differentiable loss function.
result Significant reduction in Bit-Operations (BOPs) with minimal accuracy loss.
This work proposes a complete 8-bit quantization framework for large-scale deep neural networks.
problem Training large-scale deep neural networks with high performance and low memory footprint.
method WAGEUBN framework that quantizes all data paths including weights, activations, gradients, errors, updates, and batch normalization.
result Achieves competitive accuracy on the ImageNet dataset using only 8-bit integers.
DQA efficiently quantizes deep neural network activations for resource-constrained devices.
problem Efficiently quantizing deep neural network activations for resource-constrained devices.
method DQA uses simple shifting-based operations and Huffman coding for sub-6-bit quantization.
result DQA achieves significantly better accuracy than direct quantization and state-of-the-art methods.
The paper investigates how reducing memory supply voltage improves DNN accuracy under bit-cell faults.
problem Reducing energy consumption in deep neural networks by lowering memory supply voltage introduces bit-cell faults.
method The authors explore the robustness of DNN architectures to bit-cell faults and propose a regularizer to mitigate their effects.
result Operating the system in a faulty regime can save energy without significantly reducing accuracy.
A deep learning autoencoder improves error correction for one-bit quantization.
problem Improving error correction for one-bit quantization in AWGN channels.
method Proposes a novel autoencoder-based coding scheme using turbo codes as implicit regularization.
result Empirically and theoretically shows nearly optimal performance of the proposed coding scheme.
Proposes a robust neural network quantization method.
problem Training model's dependency on specific quantization methods.
method Intrinsic robustness to various quantization processes.
result Single model capable of operating at various bit-widths and policies.
Binary neural networks (BNN) have been studied extensively since they run dramatically faster at lower memory and power consumption than floating-point networks, thanks to the efficiency of bit operations. However, contemporary BNNs whose weights and activations are both single bits suffer from severe accuracy degradat…
Binary Iterative Hard Thresholding converges with optimal number of 1-bit measurements.
problem Recovering sparse signals from 1-bit compressed measurements.
method Binary Iterative Hard Thresholding (BIHT) algorithm.
result BIHT converges with only O(k/ε) measurements, optimal for recovery.
We consider the problem of estimating the arithmetic average of a finite collection of real vectors stored in a distributed fashion across several compute nodes subject to a communication budget constraint. Our analysis does not rely on any statistical assumptions about the source of the vectors. This problem arises as…
We present a theoretical analysis and empirical evaluations of a novel set of techniques for computational cost reduction of classifiers that are based on learned transform and soft-threshold. By modifying optimization procedures for dictionary and classifier training, as well as the resulting dictionary entries, our t…
High error rates improve neural network performance and reduce power consumption.
problem Training neural networks with high Bit Error Rates (BERs).
method Trained three Binarized Convolutional Neural Network architectures on various datasets with high BERs.
result High BERs do not significantly degrade test accuracy, enabling more efficient hardware.
AdaptivFloat improves deep learning inference accuracy at low precision.
problem Low precision quantization issues in deep learning inference.
method Dynamic floating-point representation with adaptive clipping.
result Consistently higher inference accuracy at low precision compared to other methods.
New techniques improve 16-bit training accuracy without 32-bit units.
problem Training deep learning models with only 16-bit floating-point units.
method Studied BFloat16 units and applied stochastic rounding and Kahan summation techniques.
result Up to 7% absolute validation accuracy gain in 16-bit-FPU training.
Quantized Neural Networks (QNNs) are often used to improve network efficiency during the inference phase, i.e. after the network has been trained. Extensive research in the field suggests many different quantization schemes. Still, the number of bits required, as well as the best quantization scheme, are yet unknown. O…
The one-bit quantization is implemented by one single comparator that operates at low power and a high rate. Hence one-bit compressive sensing (1bit-CS) becomes attractive in signal processing. When measurements are corrupted by noise during signal acquisition and transmission, 1bit-CS is usually modeled as minimizing …
Directional oil well drilling requires high precision of the wellbore positioning inside the productive area. However, due to specifics of engineering design, sensors that explicitly determine the type of the drilled rock are located farther than 15m from the drilling bit. As a result, the target area runaways can be d…
Differentiable model compression adds noise to parameters during training.
problem Model compression for efficient deployment.
method Adding pseudo quantization noise to model parameters during training.
result Compresses model by more than 8x on ImageNet with 0.3% accuracy loss.
Bit-Swap improves lossless compression for hierarchical latent variable models.
problem Efficient lossless compression for latent variable models with hierarchical structure.
method Generalizes bits-back coding to hierarchical latent variable models with Markov chain structure.
result Achieves superior lossless compression rates for hierarchical latent variable models.
A new method, REC, compresses images by encoding their latent representations efficiently.
problem Efficiently compressing single images with latent representations.
method Relative Entropy Coding (REC) that directly encodes latent representations with codelength close to relative entropy.
result REC is more efficient for single image compression compared to previous methods and is competitive for lossy compression.
One-bit measurements widely exist in the real world, and they can be used to recover sparse signals. This task is known as the problem of learning halfspaces in learning theory and one-bit compressive sensing (1bit-CS) in signal processing. In this paper, we propose novel algorithms based on both convex and nonconvex s…
This paper enables deep network inference on microcontrollers with improved accuracy and reduced memory usage.
problem Deploying deep networks on resource-constrained edge-devices with low memory and computational constraints.
method Mixed low-bitwidth compression, rule-based iterative procedure for bit precision determination, quantization-aware retraining, and integer-only model conversion.
result Improved Top1 accuracy of 68% on a 2MB FLASH memory STM32H7 microcontroller, 8% higher than 8-bit implementations.
Bit-slice sparsity improves ReRAM-based DNN acceleration.
problem Limited ADC power and area constraints in ReRAM-based DNN accelerators.
method Proposed bit-slice L1 algorithm to induce sparsity during training.
result 2x sparsity improvement compared to previous methods.
MSD removes dequantization bottleneck in LLM inference by approximating high-precision activations.
problem Dequantization bottleneck in LLM inference on modern AI accelerators.
method MSD decomposes high-precision activations into multiple low-precision components for direct multiplication with quantized weights.
result MSD avoids INT8-to-BF16 weight conversion, reducing dequantization cycles and HBM traffic.
Model compression has gained a lot of attention due to its ability to reduce hardware resource requirements significantly while maintaining accuracy of DNNs. Model compression is especially useful for memory-intensive recurrent neural networks because smaller memory footprint is crucial not only for reducing storage re…
New algorithm learns halfspaces with noise using Forster decomposition.
problem Learning halfspaces in noisy data.
method Forster decomposition and efficient mixture of distributions.
result First polynomial-time algorithm with strongly polynomial sample complexity.
The state-of-the-art hardware platforms for training Deep Neural Networks (DNNs) are moving from traditional single precision (32-bit) computations towards 16 bits of precision -- in large part due to the high energy efficiency and smaller bit storage associated with using reduced-precision representations. However, un…
Majority bit estimation in noisy random recursive DAGs.
problem Estimating the majority bit in a noisy random recursive DAG.
method Majority rule among nodes, with bit flipping and noisy channel.
result Identification of the threshold for p at which majority rule yields errors. Paper improves full-text search engines for fast exact NNS in binary codes.
problem Efficient nearest neighbor search in Hamming space for full-text search engines.
method Revisits and combines three techniques from information retrieval: bit operation, subs-code filtering, and data preprocessing with permutation.
result Significant speed-ups for NNS in binary codes over state-of-the-art term match approach.
Paper proposes a log-domain training method to reduce neural network complexity.
problem High computational complexity in training deep neural networks limits real-time training.
method End-to-end training and inference scheme using approximate logarithmic operations in the log-domain.
result 16-bit log-based training achieves within 1% accuracy of floating-point baselines.
SGQuant reduces GNN memory usage without significant accuracy loss.
problem High memory consumption in GNNs limits their applicability on memory-constrained devices.
method Proposes a specialized GNN quantization scheme (SGQuant) with a quantization algorithm, fine-tuning scheme, and multi-granularity strategy.
result SGQuant reduces GNN memory footprint from 4.25x to 31.9x with minimal accuracy loss.
Paper proposes a CNN-based method for estimating intra frame bits and quality.
problem Efficient video delivery and bit allocation in video coding.
method Deep learning approach using CNNs trained on original frames and encoded distortions.
result Accurate estimation of intra frame bits and quality for better bit allocation.
We construct an elementary, combinatorial kind of topological quantum field theory, based on curves, surfaces, and orientations. The construction derives from contact invariants in sutured Floer homology and is essentially an elaboration of a TQFT defined by Honda--Kazez--Matic. This topological field theory stores inf…
Paper proposes a hybrid model-based and data-driven approach for one-bit compressive autoencoding.
problem Designing efficient one-bit compressive autoencoding models for complex systems.
method Hybrid model-based and data-driven methodology for one-bit sparse signal recovery.
result Significant improvement in one-bit compressive autoencoding compared to state-of-the-art algorithms.
Paper proposes a hybrid model-based and data-driven method for one-bit compressive variational autoencoding.
problem Designing efficient one-bit compressive sensing systems.
method Hybrid model-based and data-driven approach for one-bit compressive variational autoencoding.
result Significant improvement in one-bit compressive sensing compared to state-of-the-art methods.
Paper proposes training deep neural networks with 8-bit floating point precision.
problem Challenges in training deep neural networks at 8-bit precision due to higher precision and dynamic range requirements.
method Proposes a method to train deep neural networks using 8-bit floating point for weights, activations, errors, and gradients. Introduces an enhanced loss scaling method and stochastic rounding technique.
result Demonstrates state-of-the-art accuracy across multiple datasets and workloads compared to full precision baseline.
New protocols show 1-bit mean estimation can be order-optimal without interaction.
problem Can 1-bit mean estimation be optimal without interaction?
method Adaptive and non-adaptive threshold and interval queries, with one adaptive transition.
result Arbitrary non-adaptive quantizers can match the adaptive rate, suggesting interaction is not necessary.
Topological data analysis classifies encrypted bits with success.
problem Classifying encrypted data with traditional machine learning methods.
method Persistent homology for generating topological features, machine learning pipeline.
result Successfully classifies encrypted data, outperforming classical models.