In this paper, we propose a simple variant of the original SVRG, called variance reduced stochastic gradient descent (VR-SGD). Unlike the choices of snapshot and starting points in SVRG and its proximal variant, Prox-SVRG, the two vectors of VR-SGD are set to the average and last iterate of the previous epoch, respectively. The settings allow us to use much larger learning rates, and also make our convergence analysis more challenging. We also design two different update rules for smooth and non-smooth objective functions, respectively, which means that VR-SGD can tackle non-smooth and/or non-strongly convex problems directly without any reduction techniques. Moreover, we analyze the convergence properties of VR-SGD for strongly convex problems, which show that VR-SGD attains linear convergence. Different from most algorithms that have no convergence guarantees for non-strongly convex problems, we also provide the convergence guarantees of VR-SGD for this case, and empirically verify that VR-SGD with varying learning rates achieves similar performance to its momentum accelerated variant that has the optimal convergence rate $\mathcal {O}(1/T^2)$ O ( 1 / T 2 ) . Finally, we apply VR-SGD to solve various machine learning problems, such as convex and non-convex empirical risk minimization, and leading eigenvalue computation. Experimental results show that VR-SGD converges significantly faster than SVRG and Prox-SVRG, and usually outperforms state-of-the-art accelerated methods, e.g., Katyusha.

/pdf/vr-sgd-a-simple-stochastic-variance-reduction-method-for-2yswi4a5th.pdf

VR-SGD: A Simple Stochastic Variance Reduction Method for Machine Learning

SARAH and SPIDER are two recently developed stochastic variance-reduced algorithms, and SPIDER has been shown to achieve a near-optimal first-order oracle complexity in smooth nonconvex optimization. However, SPIDER uses an accuracy-dependent stepsize that slows down the convergence in practice, and cannot handle objective functions that involve nonsmooth regularizers. In this paper, we propose SpiderBoost as an improved scheme, which allows to use a much larger constant-level stepsize while maintaining the same near-optimal oracle complexity, and can be extended with proximal mapping to handle composite optimization (which is nonsmooth and nonconvex) with provable convergence guarantee. In particular, we show that proximal SpiderBoost achieves an oracle complexity of $\mathcal{O}(\min\{n^{1/2}\epsilon^{-2},\epsilon^{-3}\})$ in composite nonconvex optimization, improving the state-of-the-art result by a factor of $\mathcal{O}(\min\{n^{1/6},\epsilon^{-1/3}\})$. We further develop a novel momentum scheme to accelerate SpiderBoost for composite optimization, which achieves the near-optimal oracle complexity in theory and substantial improvement in experiments.

SpiderBoost and Momentum: Faster Stochastic Variance Reduction Algorithms

In this paper, we propose a simple variant of the original SVRG, called variance reduced stochastic gradient descent (VR-SGD). Unlike the choices of snapshot and starting points in SVRG and its proximal variant, Prox-SVRG, the two vectors of VR-SGD are set to the average and last iterate of the previous epoch, respectively. The settings allow us to use much larger learning rates, and also make our convergence analysis more challenging. We also design two different update rules for smooth and non-smooth objective functions, respectively, which means that VR-SGD can tackle non-smooth and/or non-strongly convex problems directly without any reduction techniques. Moreover, we analyze the convergence properties of VR-SGD for strongly convex problems, which show that VR-SGD attains linear convergence. Different from its counterparts that have no convergence guarantees for non-strongly convex problems, we also provide the convergence guarantees of VR-SGD for this case, and empirically verify that VR-SGD with varying learning rates achieves similar performance to its momentum accelerated variant that has the optimal convergence rate $\mathcal{O}(1/T^2)$. Finally, we apply VR-SGD to solve various machine learning problems, such as convex and non-convex empirical risk minimization, and leading eigenvalue computation. Experimental results show that VR-SGD converges significantly faster than SVRG and Prox-SVRG, and usually outperforms state-of-the-art accelerated methods, e.g., Katyusha.

In this paper, we introduce a simplified and unified method for finite-sum convex optimization, named \emph{Variance Reduction via Accelerated Dual Averaging (VRADA)}. In both general convex and strongly convex settings, VRADA can attain an $O\big(\frac{1}{n}\big)$-accurate solution in $O(n\log\log n)$ number of stochastic gradient evaluations which improves the best-known result $O(n\log n)$, where $n$ is the number of samples. Meanwhile, VRADA matches the lower bound of the general convex setting up to a $\log\log n$ factor and matches the lower bounds in both regimes $n\le \Theta(\kappa)$ and $n\gg \kappa$ of the strongly convex setting, where $\kappa$ denotes the condition number. Besides improving the best-known results and matching all the above lower bounds simultaneously, VRADA has more unified and simplified algorithmic implementation and convergence analysis for both the general convex and strongly convex settings. The underlying novel approaches such as the novel initialization strategy in VRADA may be of independent interest. Through experiments on real datasets, we show the good performance of VRADA over existing methods for large-scale machine learning problems.

/pdf/variance-reduction-via-accelerated-dual-averaging-for-finite-4wt2q3h3rv.pdf

Variance Reduction via Accelerated Dual Averaging for Finite-Sum Optimization

Quantization has become a popular technique to compress neural networks and reduce compute cost, but most prior work focuses on studying quantization without changing the network size. Many real-world applications of neural networks have compute cost and memory budgets, which can be traded off with model quality by changing the number of parameters. In this work, we use ResNet as a case study to systematically investigate the effects of quantization on inference compute cost-quality tradeoff curves. Our results suggest that for each bfloat16 ResNet model, there are quantized models with lower cost and higher ac-curacy; in other words, the bfloat16 compute cost-quality tradeoff curve is Pareto-dominated by the 4-bit and 8-bit curves, with models primarily quantized to 4-bit yielding the best Pareto curve. Furthermore, we achieve state-of-the-art results on ImageNet for 4-bit ResNet-50 with quantization-aware training, obtaining a top-1 eval accuracy of 77.09%. We demonstrate the regularizing effect of quantization by measuring the generalization gap. The quantization method we used is optimized for practicality: It requires little tuning and is designed with hardware capabilities in mind. Our work motivates further research into optimal numeric formats for quantization, as well as the development of machine learning accelerators supporting these formats. As part of this work, we contribute a quantization library written in JAX, which is open-sourced at https://github.com/google-research/google-research/tree/master/aqt.

/pdf/pareto-optimal-quantized-resnet-is-mostly-4-bit-1m6qcqam6r.pdf

Pareto-Optimal Quantized ResNet Is Mostly 4-bit

This paper proposes an accelerated proximal stochastic variance reduced gradient (ASVRG) method, in which we design a simple and effective momentum acceleration trick. Unlike most existing accelerated stochastic variance reduction methods such as Katyusha, ASVRG has only one additional variable and one momentum parameter. Thus, ASVRG is much simpler than those methods, and has much lower per-iteration complexity. We prove that ASVRG achieves the best known oracle complexities for both strongly convex and non-strongly convex objectives. In addition, we extend ASVRG to mini-batch and non-smooth settings. We also empirically verify our theoretical results and show that the performance of ASVRG is comparable with, and sometimes even better than that of the state-of-the-art stochastic methods.

/pdf/asvrg-accelerated-proximal-svrg-11nk2vqr8x.pdf

ASVRG: Accelerated Proximal SVRG

The training of deep neural networks (DNNs) requires intensive resources both for computation and for storage performance. Thus, DNNs cannot be efficiently applied to mobile phones and embedded devices, which seriously limits their applicability in industry applications. To address this issue, we propose a novel encoding scheme of using {−1, +1} to decompose quantized neural networks (QNNs) into multibranch binary networks, which can be efficiently implemented by bitwise operations (xnor and bitcount) to achieve model compression, computational acceleration and resource saving. Based on our method, users can easily achieve different encoding precisions arbitrarily according to their requirements and hardware resources. The proposed mechanism is very suitable for the use of FPGA and ASIC in terms of data storage and computation, which provides a feasible idea for smart chips. We validate the effectiveness of our method on both large-scale image classification tasks (e.g., ImageNet) and object detection tasks. In particular, our method with lowbit encoding can still achieve almost the same performance as its full-precision counterparts.

/pdf/multi-precision-quantized-neural-networks-via-encoding-doidqzj6py.pdf

Multi-Precision Quantized Neural Networks via Encoding Decomposition of {-1,+1}

Since model quantization helps to reduce the model size and computation latency, it has been successfully applied in many applications of mobile phones, embedded devices and smart chips. The mixed-precision quantization model can match different quantization bit-precisions according to the sensitivity of different layers to achieve great performance. However, it is a difficult problem to quickly determine the quantization bit-precision of each layer in deep neural networks according to some constraints (e.g., hardware resources, energy consumption, model size and computation latency). To address this issue, we propose a novel sequential single path search (SSPS) method for mixed-precision quantization,in which the given constraints are introduced into its loss function to guide searching process. A single path search cell is used to combine a fully differentiable supernet, which can be optimized by gradient-based algorithms. Moreover, we sequentially determine the candidate precisions according to the selection certainties to exponentially reduce the search space and speed up the convergence of searching process. Experiments show that our method can efficiently search the mixed-precision models for different architectures (e.g., ResNet-20, 18, 34, 50 and MobileNet-V2) and datasets (e.g., CIFAR-10, ImageNet and COCO) under given constraints, and our experimental results verify that SSPS significantly outperforms their uniform counterparts.

Effective and Fast: A Novel Sequential Single Path Search for Mixed-Precision Quantization.

Model quantization can reduce the model size and computational latency, it has become an essential technique for the deployment of deep neural networks on resourceconstrained hardware (e.g., mobile phones and embedded devices). The existing quantization methods mainly consider the numerical elements of the weights and activation values, ignoring the relationship between elements. The decline of representation ability and information loss usually lead to the performance degradation. Inspired by the characteristics of images in the frequency domain, we propose a novel multiscale wavelet quantization (MWQ) method. This method decomposes original data into multiscale frequency components by wavelet transform, and then quantizes the components of different scales, respectively. It exploits the multiscale frequency and spatial information to alleviate the information loss caused by quantization in the spatial domain. Because of the flexibility of MWQ, we demonstrate three applications (e.g., model compression, quantized network optimization, and information enhancement) on the ImageNet and COCO datasets. Experimental results show that our method has stronger representation ability and can play an effective role in quantized neural networks.

MWQ: Multiscale Wavelet Quantized Neural Networks.

As an effective technique to achieve the implementation of deep neural networks in edge devices, model quantization has been successfully applied in many practical applications. No matter the methods of quantization aware training (QAT) or post-training quantization (PTQ), they all depend on the target bit-widths. When the precision of quantization is adjusted, it is necessary to fine-tune the quantized model or minimize the quantization noise, which brings inconvenience in practical applications. In this work, we propose a method to train a model for all quantization that supports diverse bit-widths (e.g., form 8-bit to 1-bit) to satisfy the online quantization bit-width adjustment. It is hot-swappable that can provide specific quantization strategies for different candidates through multiscale quantization. We use wavelet decomposition and reconstruction to increase the diversity of weights, thus significantly improving the performance of each quantization candidate, especially at ultra-low bit-widths (e.g., 3-bit, 2-bit, and 1-bit). Experimental results on ImageNet and COCO show that our method can achieve accuracy comparable performance to dedicated models trained at the same precision.

Yan Ren

Papers

ASVRG: Accelerated Proximal SVRG

Multi-Precision Quantized Neural Networks via Encoding Decomposition of {-1,+1}

Effective and Fast: A Novel Sequential Single Path Search for Mixed-Precision Quantization.

MWQ: Multiscale Wavelet Quantized Neural Networks.

One Model for All Quantization: A Quantized Network Supporting Hot-Swap Bit-Width Adjustment.