The paper uses information geometry to analyze model compression techniques, focusing on operator factorization.
problem Efficiently compressing deep learning models for resource-constrained devices.
method Information geometry applied to model compression, focusing on operator factorization.
result Iterative methods are crucial for fine-tuning models, especially when compression ratios are fixed.
Sign-based algorithms (e.g. signSGD) have been proposed as a biased gradient compression technique to alleviate the communication bottleneck in training large neural networks across multiple workers. We show simple convex counter-examples where signSGD does not converge to the optimum. Further, even when it does conver…
Data compression is a popular technique for improving the efficiency of data processing workloads such as SQL queries and more recently, machine learning (ML) with classical batch gradient methods. But the efficacy of such ideas for mini-batch stochastic gradient descent (MGD), arguably the workhorse algorithm of moder…
In order to mitigate the high communication cost in distributed and federated learning, various vector compression schemes, such as quantization, sparsification and dithering, have become very popular. In designing a compression method, one aims to communicate as few bits as possible, which minimizes the cost per commu…
Real-world data typically contain repeated and periodic patterns. This suggests that they can be effectively represented and compressed using only a few coefficients of an appropriate basis (e.g., Fourier, Wavelets, etc.). However, distance estimation when the data are represented using different sets of coefficients i…
New method alters Seifert surfaces without compression.
problem Understanding alterations of Seifert surfaces.
method Introducing alteration as a new surface operation.
result Seifert surfaces can be altered without compression.
Modern deep learning models are often trained in parallel over a collection of distributed machines to reduce training time. In such settings, communication of model updates among machines becomes a significant performance bottleneck and various lossy update compression techniques have been proposed to alleviate this p…
The low displacement rank (LDR) framework for structured matrices represents a matrix through two displacement operators and a low-rank residual. Existing use of LDR matrices in deep learning has applied fixed displacement operators encoding forms of shift invariance akin to convolutions. We introduce a class of LDR ma…
Asynchronous computation and gradient compression have emerged as two key techniques for achieving scalability in distributed optimization for large-scale machine learning. This paper presents a unified analysis framework for distributed gradient methods operating with staled and compressed gradients. Non-asymptotic bo…
For a real or complex semisimple Lie group G and two nested parabolic subgroups Q⊂P⊂G, we study parabolic geometries of type (G,Q). Associated to the group P, we introduce a class of relative natural bundles and relative tractor bundles and construct some basic invariant differential operators on …
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.
Paper establishes convergence rates for learning elliptic pseudo-differential operators.
problem Learning elliptic pseudo-differential operators in partial differential equations.
method Wavelet-Galerkin framework, structured infinite-dimensional regression problem, sparse estimator, matrix compression, nested-support strategy.
result Obtained convergence rates for the estimator and efficient Galerkin solver.
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.
New video compression method outperforms traditional approaches.
problem Efficient video compression in low latency mode.
method Feedback Recurrent Autoencoder network architecture.
result State of the art MS-SSIM/rate performance on UVG dataset.
GeneCAI optimizes DNN compression hyper-parameters for mobile devices.
problem Efficient deployment of complex DNNs on resource-limited devices.
method GeneCAI uses genetic algorithm to learn optimal hyper-parameters.
result GeneCAI finds models with better accuracy-complexity trade-off.
Turbo-DDCM speeds up zero-shot image compression.
problem Slowness and high computational demand in zero-shot diffusion-based compression.
method Modified DDCM framework with Turbo-DDCM, combining noise vectors and improved encoding.
result Turbo-DDCM achieves faster compression than state-of-the-art methods.
We study compressing empirical measures in finite RKHSs using convex optimization.
problem Efficiently approximating empirical measures in high-dimensional spaces.
method Convex optimization and lower bounds on ball size.
result High probability lower bounds on ball size under various conditions.
New algorithm reduces communication in distributed learning by sharing compressed beliefs.
problem Efficiently learning from private data in a distributed setting with large hypothesis sets.
method Proposes a belief update rule for distributed cooperative learning with compressed (sparse or quantized) beliefs.
result Beliefs converge almost surely to optimal hypotheses with a linear concentration rate.
Compressing convolutional neural networks (CNNs) is essential for transferring the success of CNNs to a wide variety of applications to mobile devices. In contrast to directly recognizing subtle weights or filters as redundant in a given CNN, this paper presents an evolutionary method to automatically eliminate redunda…
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.
IntSGD compresses SGD gradients without floats, converging as SGD.
problem Efficiently compressing stochastic gradients in distributed SGD.
method Adaptive integer compression of gradients, estimating scaling adaptively.
result IntSGD matches SGD's iteration complexity for convex and non-convex functions.
We present reconstruction algorithms for smooth signals with block sparsity from their compressed measurements. We tackle the issue of varying group size via group-sparse least absolute shrinkage selection operator (LASSO) as well as via latent group LASSO regularizations. We achieve smoothness in the signal via fusion…
This paper introduces ASCAI, a novel adaptive sampling methodology that can learn how to effectively compress Deep Neural Networks (DNNs) for accelerated inference on resource-constrained platforms. Modern DNN compression techniques comprise various hyperparameters that require per-layer customization to ensure high ac…
This paper proposes a simple adaptive sensing and group testing algorithm for sparse signal recovery. The algorithm, termed Compressive Adaptive Sense and Search (CASS), is shown to be near-optimal in that it succeeds at the lowest possible signal-to-noise-ratio (SNR) levels, improving on previous work in adaptive comp…
New method compresses LSTM networks using MPS tensor trains.
problem Challenges in maintaining performance of compressed RNNs.
method Use of MPS tensor trains for LSTM network compression.
result MPS tensor trains outperform MPOs in storage and inference time.
Compressed sensing (CS) is a sampling theory that allows reconstruction of sparse (or compressible) signals from an incomplete number of measurements, using of a sensing mechanism implemented by an appropriate projection matrix. The CS theory is based on random Gaussian projection matrices, which satisfy recovery guara…
This work proposes ACTC for adaptive distributed learning under communication constraints.
problem Adaptive distributed learning in networks with communication constraints.
method ACTC (Adapt-Compress-Then-Combine) strategy with diffusion exchange of compressed updates.
result ACTC iterates converge to the optimizer with significant bit savings.
We solve image inverse problems using a flow-based noise model.
problem Image inverse problems with complex noise patterns.
method Normalizing flow prior for maximum a posteriori estimation.
result Empirical validation on various inverse problems.
A new method prunes neural network channels based on operation characteristics.
problem Compressing deep neural networks efficiently and maintaining accuracy.
method Differentiable masks for channel pruning considering BN and ReLU.
result Outstanding performance in accuracy with less resources compared to state-of-the-art methods.
Machine learning at the edge offers great benefits such as increased privacy and security, low latency, and more autonomy. However, a major challenge is that many devices, in particular edge devices, have very limited memory, weak processors, and scarce energy supply. We propose a hybrid hardware-software framework tha…
Artemis framework improves distributed learning with bidirectional compression and partial participation.
problem Learning in distributed or federated settings with communication constraints and device partial participation.
method Artemis framework using bidirectional compression, memory mechanism, and Polyak-Ruppert averaging.
result Fast rates of convergence (linear up to a threshold) under weak assumptions on stochastic gradients.
Video analytics requires operating with large amounts of data. Compressive sensing allows to reduce the number of measurements required to represent the video using the prior knowledge of sparsity of the original signal, but it imposes certain conditions on the design matrix. The Bayesian compressive sensing approach r…
Estimates rate-distortion function for large datasets using neural networks.
problem Designing lossy data compression schemes and comparing them with theoretical limits.
method Re-formulate rate-distortion objective and solve using neural networks.
result NERD accurately estimates the rate-distortion function for real-world datasets.
New methods improve tree ensemble models by compressing them while maintaining accuracy.
problem Theoretical understanding and practical compression of tree ensembles like random forests and gradient boosting machines.
method Spectral perspective on tree ensembles, deriving minimax rates and developing compression schemes.
result Leading eigenfunctions/singular vectors capture dominant predictive directions, leading to smaller, competitive models.
Designs CNNs for better image reconstruction.
problem Image reconstruction from limited data.
method Parseval convolution operators and chaining of elementary modules.
result CNN-based algorithm yields better results than sparsity-based methods.
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.
Deep neural networks generalize well on unseen data though the number of parameters often far exceeds the number of training examples. Recently proposed complexity measures have provided insights to understanding the generalizability in neural networks from perspectives of PAC-Bayes, robustness, overparametrization, co…
Deep neural networks (DNNs) frequently contain far more weights, represented at a higher precision, than are required for the specific task which they are trained to perform. Consequently, they can often be compressed using techniques such as weight pruning and quantization that reduce both the model size and inference…
Deep neural networks (DNNs) have become the state-of-the-art technique for machine learning tasks in various applications. However, due to their size and the computational complexity, large DNNs are not readily deployable on edge devices in real-time. To manage complexity and accelerate computation, network compression…
Vectors of data are at the heart of machine learning and data mining. Recently, vector quantization methods have shown great promise in reducing both the time and space costs of operating on vectors. We introduce a vector quantization algorithm that can compress vectors over 12x faster than existing techniques while al…
Network agents solve adaptive regression problems with compressed signals.
problem Solving regression problems in networks with communication constraints.
method ACTC diffusion strategy with randomized differential compression.
result Optimized resource allocation improves performance.
EControl improves fast distributed optimization with compression and error control.
problem Stable convergence issues in distributed training with compression.
method Proposes EControl to regulate error compensation and prove fast convergence.
result Proves fast convergence for EControl in various convex settings without additional assumptions.
The implementation of Deep Convolutional Neural Networks (ConvNets) on tiny end-nodes with limited non-volatile memory space calls for smart compression strategies capable of shrinking the footprint yet preserving predictive accuracy. There exist a number of strategies for this purpose, from those that play with the to…
ALF reduces network parameters and operations by 70% and 61%, respectively, on embedded hardware.
problem Efficient deployment of deep learning models on resource-constrained hardware.
method Autoencoder-based low-rank filter-sharing technique.
result ALF achieves significant compression with minimal accuracy loss.
New method identifies latent treatment effects from proxy models.
problem Identifying heterogeneous treatment effects under unobserved confounding.
method Compressed observable operator and spectral analysis of treatment effects.
result Eigenvalues of the operator represent latent treatment effects.
We propose tensorial neural networks (TNNs), a generalization of existing neural networks by extending tensor operations on low order operands to those on high order ones. The problem of parameter learning is challenging, as it corresponds to hierarchical nonlinear tensor decomposition. We propose to solve the learning…
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…
Paper analyzes convergence rates of compressed LSR algorithms in federated learning.
problem Impact of compression on convergence rates in distributed learning.
method Analyzes a general stochastic approximation algorithm for LSR with weak assumptions.
result Convergence rates depend on the covariance of additive noise and compression strategy.