Instantaneous and on demand accuracy-efficiency trade-off has been recently explored in the context of neural networks slimming. In this paper, we propose a flexible quantization strategy, termed Switchable Precision neural Networks (SP-Nets), to train a shared network capable of operating at multiple quantization levels. At runtime, the network can adjust its precision on the fly according to instant memory, latency, power consumption and accuracy demands. For example, by constraining the network weights to 1-bit with switchable precision activations, our shared network spans from BinaryConnect to Binarized Neural Network, allowing to perform dot-products using only summations or bit operations. In addition, a self-distillation scheme is proposed to increase the performance of the quantized switches. We tested our approach with three different quantizers and demonstrate the performance of SP-Nets against independently trained quantized models in classification accuracy for Tiny ImageNet and ImageNet datasets using ResNet-18 and MobileNet architectures.
Attention mechanism has gained great success in vision recognition. Many works are devoted to improving the effectiveness of attention mechanism, which finely design the structure of the attention operator. These works need lots of experiments to pick out the optimal settings when scenarios change, which consumes a lot of time and computational resources. In addition, a neural network often contains many network layers, and most studies often use the same attention module to enhance different network layers, which hinders the further improvement of the performance of the self-attention mechani…
Tsinghua University（清华大学）；Beijing National Research Center for Information Science and Technology (BNRist)（北京信息科学与技术国家研究中心（BNRist））；Intel Labs China（英特尔中国研究院）
We propose a general method to train a single convolutional neural network which is capable of switching image resolutions at inference. Thus the running speed can be selected to meet various computational resource limits. Networks trained with the proposed method are named Resolution Switchable Networks (RS-Nets). The basic training framework shares network parameters for handling images which differ in resolution, yet keeps separate batch normalization layers. Though it is parameter-efficient in design, it leads to inconsistent accuracy variations at different resolutions, for which we provi…
The Chinese University of Hong Kong（香港中文大学）；The University of Hong Kong（香港大学）；SenseTime Research（商汤科技研究院）
We address a learning-to-normalize problem by proposing Switchable Normalization (SN), which learns to select different normalizers for different normalization layers of a deep neural network. SN employs three distinct scopes to compute statistics (means and variances) including a channel, a layer, and a minibatch. SN switches between them by learning their importance weights in an end-to-end manner. It has several good properties. First, it adapts to various network architectures and tasks (see Fig.1). Second, it is robust to a wide range of batch sizes, maintaining high performance even when…
Korea Advanced Institute of Science and Technology（韩国科学技术院）
Recent proposals of deep beamformers using deep neural networks have attracted significant attention as computational efficient alternatives to adaptive and compressive beamformers. Moreover, deep beamformers are versatile in that image post-processing algorithms can be combined with the beamforming. Unfortunately, in the current technology, a separate beamformer should be trained and stored for each application, demanding significant scanner resources. To address this problem, here we propose a {\em switchable} deep beamformer that can produce various types of output such as DAS, speckle remo…
The University of Hong Kong（香港大学）；SenseTime Research（商汤科技研究院）；The Chinese University of Hong Kong（香港中文大学）
We address a learning-to-normalize problem by proposing Switchable Normalization (SN), which learns to select different normalizers for different normalization layers of a deep neural network. SN employs three distinct scopes to compute statistics (means and variances) including a channel, a layer, and a minibatch. SN switches between them by learning their importance weights in an end-to-end manner. It has several good properties. First, it adapts to various network architectures and tasks. Second, it is robust to a wide range of batch sizes, maintaining high performance even when small minib…
Recent work in network quantization has substantially reduced the time and space complexity of neural network inference, enabling their deployment on embedded and mobile devices with limited computational and memory resources. However, existing quantization methods often represent all weights and activations with the same precision (bit-width). In this paper, we explore a new dimension of the design space: quantizing different layers with different bit-widths. We formulate this problem as a neural architecture search problem and propose a novel differentiable neural architecture search (DNAS) …
The University of Adelaide（阿德莱德大学）；South China University of Technology（华南理工大学）
In this paper, we seek to tackle a challenge in training low-precision networks: the notorious difficulty in propagating gradient through a low-precision network due to the non-differentiable quantization function. We propose a solution by training the low-precision network with a fullprecision auxiliary module. Specifically, during training, we construct a mix-precision network by augmenting the original low-precision network with the full precision auxiliary module. Then the augmented mix-precision network and the low-precision network are jointly optimized. This strategy creates additional …
In this paper, we propose to train a network with binary weights and low-bitwidth activations, designed especially for mobile devices with limited power consumption. Most previous works on quantizing CNNs uncritically assume the same architecture, though with reduced precision. However, we take the view that for best performance it is possible (and even likely) that a different architecture may be better suited to dealing with low precision weights and activations. Specifically, we propose a "network expansion" strategy in which we aggregate a set of homogeneous low-precision branches to impli…
This paper introduces two architectures for the inference of convolutional neural networks (CNNs). Both architectures exploit weight sparsity and compression to reduce computational complexity and bandwidth. The first architecture uses multiply-accumulators (MACs) but avoids unnecessary multiplications by skipping zero weights. The second architecture exploits weight sparsity at the level of their bit representation by substituting resource-intensive MACs with much smaller Bit Layer Multiply Accumulators (BLMACs). The use of BLMACs also allows variable precision weights as variable size intege…
School of Electrical and Information Engineering, The University of Sydney（悉尼大学电气与信息工程学院）；Xilinx Research Labs（赛灵思研究实验室）
A low precision deep neural network training technique for producing sparse, ternary neural networks is presented. The technique incorporates hard- ware implementation costs during training to achieve significant model compression for inference. Training involves three stages: network training using L2 regularization and a quantization threshold regularizer, quantization pruning, and finally retraining. Resulting networks achieve improved accuracy, reduced memory footprint and reduced computational complexity compared with conventional methods, on MNIST and CIFAR10 datasets. Our networks are u…
Institute of Automation, Chinese Academy of Sciences（中国科学院自动化研究所）；School of Artificial Intelligence, University of Chinese Academy of Sciences（中国科学院大学人工智能学院）
Quantization is a widely used technique to compress and accelerate deep neural networks. However, conventional quantization methods use the same bit-width for all (or most of) the layers, which often suffer significant accuracy degradation in the ultra-low precision regime and ignore the fact that emergent hardware accelerators begin to support mixed-precision computation. Consequently, we present a novel and principled framework to solve the mixed-precision quantization problem in this paper. Briefly speaking, we first formulate the mixed-precision quantization as a discrete constrained optim…
Efficient model inference is an important and practical issue in the deployment of deep neural network on resource constraint platforms. Network quantization addresses this problem effectively by leveraging low-bit representation and arithmetic that could be conducted on dedicated embedded systems. In the previous works, the parameter bitwidth is set homogeneously and there is a trade-off between superior performance and aggressive compression. Actually the stacked network layers, which are generally regarded as hierarchical feature extractors, contribute diversely to the overall performance. …
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 multi-branch 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 me…
The University of Adelaide（阿德莱德大学）；South China University of Technology（华南理工大学）
This paper tackles the problem of training a deep convolutional neural network with both low-precision weights and low-bitwidth activations. Optimizing a low-precision network is very challenging since the training process can easily get trapped in a poor local minima, which results in substantial accuracy loss. To mitigate this problem, we propose three simple-yet-effective approaches to improve the network training. First, we propose to use a two-stage optimization strategy to progressively find good local minima. Specifically, we propose to first optimize a net with quantized weights and th…
Monash University（莫纳什大学）；The University of Adelaide（阿德莱德大学）；South China University of Technology（华南理工大学）
This paper tackles the problem of training a deep convolutional neural network of both low-bitwidth weights and activations. Optimizing a low-precision network is very challenging due to the non-differentiability of the quantizer, which may result in substantial accuracy loss. To address this, we propose three practical approaches, including (i) progressive quantization; (ii) stochastic precision; and (iii) joint knowledge distillation to improve the network training. First, for progressive quantization, we propose two schemes to progressively find good local minima. Specifically, we propose t…
High throughput and low latency inference of deep neural networks are critical for the deployment of deep learning applications. This paper presents the efficient inference techniques of IntelCaffe, the first Intel optimized deep learning framework that supports efficient 8-bit low precision inference and model optimization techniques of convolutional neural networks on Intel Xeon Scalable Processors. The 8-bit optimized model is automatically generated with a calibration process from FP32 model without the need of fine-tuning or retraining. We show that the inference throughput and latency wi…
University of Science and Technology of China（中国科学技术大学）；Intellifusion Inc.（云天励飞公司）
Recent years have witnessed the great advance of deep learning in a variety of vision tasks. Many state-of-the-art deep neural networks suffer from large size and high complexity, which makes it difficult to deploy in resource-limited platforms such as mobile devices. To this end, low-precision neural networks are widely studied which quantize weights or activations into the low-bit format. Though being efficient, low-precision networks are usually hard to train and encounter severe accuracy degradation. In this paper, we propose a new training strategy through expanding low-precision networks…
University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）；Snap Inc.（斯纳普公司）；ByteDance Inc.（字节跳动公司）
We present a simple and general method to train a single neural network executable at different widths (number of channels in a layer), permitting instant and adaptive accuracy-efficiency trade-offs at runtime. Instead of training individual networks with different width configurations, we train a shared network with switchable batch normalization. At runtime, the network can adjust its width on the fly according to on-device benchmarks and resource constraints, rather than downloading and offloading different models. Our trained networks, named slimmable neural networks, achieve similar (and …
Model size and inference speed/power have become a major challenge in the deployment of Neural Networks for many applications. A promising approach to address these problems is quantization. However, uniformly quantizing a model to ultra low precision leads to significant accuracy degradation. A novel solution for this is to use mixed-precision quantization, as some parts of the network may allow lower precision as compared to other layers. However, there is no systematic way to determine the precision of different layers. A brute force approach is not feasible for deep networks, as the search…
Neural network quantization has an inherent problem called accumulated quantization error, which is the key obstacle towards ultra-low precision, e.g., 2- or 3-bit precision. To resolve this problem, we propose precision highway, which forms an end-to-end high-precision information flow while performing the ultra low-precision computation. First, we describe how the precision highway reduce the accumulated quantization error in both convolutional and recurrent neural networks. We also provide the quantitative analysis of the benefit of precision highway and evaluate the overhead on the state-o…
Xilinx Research Labs（赛灵思研究院）；Imperial College London（帝国理工学院）；University of Sydney（悉尼大学）
Modern CNN are typically based on floating point linear algebra based implementations. Recently, reduced precision NN have been gaining popularity as they require significantly less memory and computational resources compared to floating point. This is particularly important in power constrained compute environments. However, in many cases a reduction in precision comes at a small cost to the accuracy of the resultant network. In this work, we investigate the accuracy-throughput trade-off for various parameter precision applied to different types of NN models. We firstly propose a quantization…
In this paper, we derive a neural network architecture based on an analytical formulation of the parallel-to-fan beam conversion problem following the concept of precision learning. The network allows to learn the unknown operators in this conversion in a data-driven manner avoiding interpolation and potential loss of resolution. Integration of known operators results in a small number of trainable parameters that can be estimated from synthetic data only. The concept is evaluated in the context of Hybrid MRI/X-ray imaging where transformation of the parallel-beam MRI projections to fan-beam X…
Adaptive Precision Training: Quantify Back Propagation in Neural Networks with Fixed-point Numbers. Recent emerged quantization technique has been applied to inference of deep neural networks for fast and efficient execution. However, directly applying quantization in training can cause significant accuracy loss, thus remaining an open challenge.
Deep neural networks with adaptive configurations have gained increasing attention due to the instant and flexible deployment of these models on platforms with different resource budgets. In this paper, we investigate a novel option to achieve this goal by enabling adaptive bit-widths of weights and activations in the model. We first examine the benefits and challenges of training quantized model with adaptive bit-widths, and then experiment with several approaches including direct adaptation, progressive training and joint training. We discover that joint training is able to produce comparabl…
We propose precision gating (PG), an end-to-end trainable dynamic dual-precision quantization technique for deep neural networks. PG computes most features in a low precision and only a small proportion of important features in a higher precision to preserve accuracy. The proposed approach is applicable to a variety of DNN architectures and significantly reduces the computational cost of DNN execution with almost no accuracy loss. Our experiments indicate that PG achieves excellent results on CNNs, including statically compressed mobile-friendly networks such as ShuffleNet. Compared to the sta…
Brown University（布朗大学）；Ben-Gurion University（本-古里安大学）
Quantized neural networks (QNNs) are among the main approaches for deploying deep neural networks on low resource edge devices. Training QNNs using different levels of precision throughout the network (dynamic quantization) typically achieves superior trade-offs between performance and computational load. However, optimizing the different precision levels of QNNs can be complicated, as the values of the bit allocations are discrete and difficult to differentiate for. Also, adequately accounting for the dependencies between the bit allocation of different layers is not straight-forward. To meet…
Eindhoven University of Technology（埃因霍温理工大学）；NXP Semiconductors（恩智浦半导体）
Bayesian Optimization Mixed-Precision Neural Architecture Search (BOMP-NAS) is an approach to quantization-aware neural architecture search (QA-NAS) that leverages both Bayesian optimization (BO) and mixed-precision quantization (MP) to efficiently search for compact, high performance deep neural networks. The results show that integrating quantization-aware fine-tuning (QAFT) into the NAS loop is a necessary step to find networks that perform well under low-precision quantization: integrating it allows a model size reduction of nearly 50\% on the CIFAR-10 dataset. BOMP-NAS is able to find neu…
We present PrecisionBatching, a quantized inference algorithm for speeding up neural network execution on traditional hardware platforms at low bitwidths without the need for retraining or recalibration. PrecisionBatching decomposes a neural network into individual bitlayers and accumulates them using fast 1-bit operations while maintaining activations in full precision. PrecisionBatching not only facilitates quantized inference at low bitwidths (< 8 bits) without the need for retraining/recalibration, but also 1) enables traditional hardware platforms the ability to realize inference speedups…
Mixed-precision quantization can potentially achieve the optimal tradeoff between performance and compression rate of deep neural networks, and thus, have been widely investigated. However, it lacks a systematic method to determine the exact quantization scheme. Previous methods either examine only a small manually-designed search space or utilize a cumbersome neural architecture search to explore the vast search space. These approaches cannot lead to an optimal quantization scheme efficiently. This work proposes bit-level sparsity quantization (BSQ) to tackle the mixed-precision quantization …
Institute of Computing Technology, Chinese Academy of Sciences（中国科学院计算技术研究所）；University of Chinese Academy of Sciences（中国科学院大学）；Jilin University（吉林大学）
Accelerating deep convolutional neural networks has become an active topic and sparked an interest in academia and industry. In this paper, we propose an efficient low-precision quantized Winograd convolution algorithm, called LANCE, which combines the advantages of fast convolution and quantization techniques. By embedding linear quantization operations into the Winograd-domain, the fast convolution can be performed efficiently under low-precision computation on graphics processing units. We test neural network models with LANCE on representative image classification datasets, including SVHN,…
École Polytechnique de Montréal（蒙特利尔综合理工学院）；Université de Montréal（蒙特利尔大学）
Multipliers are the most space and power-hungry arithmetic operators of the digital implementation of deep neural networks. We train a set of state-of-the-art neural networks (Maxout networks) on three benchmark datasets: MNIST, CIFAR-10 and SVHN. They are trained with three distinct formats: floating point, fixed point and dynamic fixed point. For each of those datasets and for each of those formats, we assess the impact of the precision of the multiplications on the final error after training. We find that very low precision is sufficient not just for running trained networks but also for tr…
We consider learning deep neural networks (DNNs) that consist of low-precision weights and activations for efficient inference of fixed-point operations. In training low-precision networks, gradient descent in the backward pass is performed with high-precision weights while quantized low-precision weights and activations are used in the forward pass to calculate the loss function for training. Thus, the gradient descent becomes suboptimal, and accuracy loss follows. In order to reduce the mismatch in the forward and backward passes, we utilize mean squared quantization error (MSQE) regularizat…
Department of Electrical Engineering, Technion（以色列理工学院电气工程系）；Habana Labs, Intel（英特尔哈瓦那实验室）；Technion（以色列理工学院）；Intel（英特尔）；Habana Labs（哈瓦那实验室）
Power consumption is a major obstacle in the deployment of deep neural networks (DNNs) on end devices. Existing approaches for reducing power consumption rely on quite general principles, including avoidance of multiplication operations and aggressive quantization of weights and activations. However, these methods do not take into account the precise power consumed by each module in the network, and are therefore not optimal. In this paper we develop accurate power consumption models for all arithmetic operations in the DNN, under various working conditions. We reveal several important factors…
Deep neural networks have achieved state-of-the-art accuracies in a wide range of computer vision, speech recognition, and machine translation tasks. However the limits of memory bandwidth and computational power constrain the range of devices capable of deploying these modern networks. To address this problem, we propose SQuantizer, a new training method that jointly optimizes for both sparse and low-precision neural networks while maintaining high accuracy and providing a high compression rate. This approach brings sparsification and low-bit quantization into a single training pass, employin…
University of California, Santa Barbara（加州大学圣巴巴拉分校）；Pacific Northwest National Laboratory（太平洋西北国家实验室）
Over the years, accelerating neural networks with quantization has been widely studied. Unfortunately, prior efforts with diverse precisions (e.g., 1-bit weights and 2-bit activations) are usually restricted by limited precision support on GPUs (e.g., int1 and int4). To break such restrictions, we introduce the first Arbitrary Precision Neural Network framework (APNN-TC) to fully exploit quantization benefits on Ampere GPU Tensor Cores. Specifically, APNN-TC first incorporates a novel emulation algorithm to support arbitrary short bit-width computation with int1 compute primitives and XOR/AND …
In this paper we introduce ShiftCNN, a generalized low-precision architecture for inference of multiplierless convolutional neural networks (CNNs). ShiftCNN is based on a power-of-two weight representation and, as a result, performs only shift and addition operations. Furthermore, ShiftCNN substantially reduces computational cost of convolutional layers by precomputing convolution terms. Such an optimization can be applied to any CNN architecture with a relatively small codebook of weights and allows to decrease the number of product operations by at least two orders of magnitude. The proposed…
Artificial neural networks can be trained with relatively low-precision floating-point and fixed-point arithmetic, using between one and 16 bits. Previous works have focused on relatively wide-but-shallow, feed-forward networks. We introduce a quantization scheme that is compatible with training very deep neural networks. Quantizing the network activations in the middle of each batch-normalization module can greatly reduce the amount of memory and computational power needed, with little loss in accuracy.
A key contributing factor to incredible success of deep neural networks has been the significant rise on massively parallel computing devices allowing researchers to greatly increase the size and depth of deep neural networks, leading to significant improvements in modeling accuracy. Although deeper, larger, or complex deep neural networks have shown considerable promise, the computational complexity of such networks is a major barrier to utilization in resource-starved scenarios. We explore the synaptogenesis of deep neural networks in the formation of efficient deep neural network architectu…
Low-precision networks, with weights and activations quantized to low bit-width, are widely used to accelerate inference on edge devices. However, current solutions are uniform, using identical bit-width for all filters. This fails to account for the different sensitivities of different filters and is suboptimal. Mixed-precision networks address this problem, by tuning the bit-width to individual filter requirements. In this work, the problem of optimal mixed-precision network search (MPS) is considered. To circumvent its difficulties of discrete search space and combinatorial optimization, a …
The `Internet of Things' has brought increased demand for AI-based edge computing in applications ranging from healthcare monitoring systems to autonomous vehicles. Quantization is a powerful tool to address the growing computational cost of such applications, and yields significant compression over full-precision networks. However, quantization can result in substantial loss of performance for complex image classification tasks. To address this, we propose a Principal Component Analysis (PCA) driven methodology to identify the important layers of a binary network, and design mixed-precision n…
Birla Institute of Technology and Science, Pilani（贝拉理工学院（皮拉尼））
On-board processing elements on UAVs are currently inadequate for training and inference of Deep Neural Networks. This is largely due to the energy consumption of memory accesses in such a network. HadaNets introduce a flexible train-from-scratch tensor quantization scheme by pairing a full precision tensor to a binary tensor in the form of a Hadamard product. Unlike wider reduced precision neural network models, we preserve the train-time parameter count, thus out-performing XNOR-Nets without a train-time memory penalty. Such training routines could see great utility in semi-supervised online…
UIUC（伊利诺伊大学厄巴纳-香槟分校）；Wormpex AI Research；University of Oregon（俄勒冈大学）
We present any-precision deep neural networks (DNNs), which are trained with a new method that allows the learned DNNs to be flexible in numerical precision during inference. The same model in runtime can be flexibly and directly set to different bit-widths, by truncating the least significant bits, to support dynamic speed and accuracy trade-off. When all layers are set to low-bits, we show that the model achieved accuracy comparable to dedicated models trained at the same precision. This nice property facilitates flexible deployment of deep learning models in real-world applications, where i…
Purdue University（普渡大学）；IBM T.J. Watson Research Center（IBM T.J. 沃森研究中心）
Precision scaling has emerged as a popular technique to optimize the compute and storage requirements of Deep Neural Networks (DNNs). Efforts toward creating ultra-low-precision (sub-8-bit) DNNs suggest that the minimum precision required to achieve a given network-level accuracy varies considerably across networks, and even across layers within a network, requiring support for variable precision in DNN hardware. Previous proposals such as bit-serial hardware incur high overheads, significantly diminishing the benefits of lower precision. To efficiently support precision re-configurability in …
Noninvasive precision modulation of high-level neural population activity via natural vision perturbations
McGovern Institute for Brain Research, Dept. of Brain and Cognitive Sciences, Massachusetts Institute of Technology（麦戈文脑科学研究所，脑科学与认知科学系，麻省理工学院）；Center for Brains, Minds, and Machines, MIT（脑、心智与机器中心，麻省理工学院）；MIT Quest for Intelligence（麻省理工学院智能探索计划）
We introduce a quantization-aware training algorithm that guarantees avoiding numerical overflow when reducing the precision of accumulators during inference. We leverage weight normalization as a means of constraining parameters during training using accumulator bit width bounds that we derive. We evaluate our algorithm across multiple quantized models that we train for different tasks, showing that our approach can reduce the precision of accumulators while maintaining model accuracy with respect to a floating-point baseline. We then show that this reduction translates to increased design ef…
Beihang University（北京航空航天大学）；SenseTime Group Limited（商汤集团有限公司）；Beijing Institute of Technology（北京理工大学）
Hardware-friendly network quantization (e.g., binary/uniform quantization) can efficiently accelerate the inference and meanwhile reduce memory consumption of the deep neural networks, which is crucial for model deployment on resource-limited devices like mobile phones. However, due to the discreteness of low-bit quantization, existing quantization methods often face the unstable training process and severe performance degradation. To address this problem, in this paper we propose Differentiable Soft Quantization (DSQ) to bridge the gap between the full-precision and low-bit networks. DSQ can …
