Neural network pruning with suitable retraining can yield networks with considerably fewer parameters than the original with comparable degrees of accuracy. Typical pruning methods require large, fully trained networks as a starting point from which they perform a time-intensive iterative pruning and retraining procedure to regain the original accuracy. We propose a novel pruning method that prunes a network real-time during training, reducing the overall training time to achieve an efficient compressed network. We introduce an activation density based analysis to identify the optimal relative sizing or compression for each layer of the network. Our method is architecture agnostic, allowing it to be employed on a wide variety of systems. For VGG-19 and ResNet18 on CIFAR-10, CIFAR-100, and TinyImageNet, we obtain exceedingly sparse networks (up to $200 \times$ reduction in parameters and over $60 \times$ reduction in inference compute operations in the best case) with accuracy comparable to the baseline network. By reducing the network size periodically during training, we achieve total training times that are shorter than those of previously proposed pruning methods. Furthermore, training compressed networks at different epochs with our proposed method yields considerable reduction in training compute complexity ($1.6\times$ to $3.2\times$ lower) at near iso-accuracy as compared to a baseline network trained entirely from scratch.
Sparsification is an efficient approach to accelerate CNN inference, but it is challenging to take advantage of sparsity in training procedure because the involved gradients are dynamically changed. Actually, an important observation shows that most of the activation gradients in back-propagation are very close to zero and only have a tiny impact on weight-updating. Hence, we consider pruning these very small gradients randomly to accelerate CNN training according to the statistical distribution of activation gradients. Meanwhile, we theoretically analyze the impact of pruning algorithm on the…
Quantization and pruning are core techniques used to reduce the inference costs of deep neural networks. State-of-the-art quantization techniques are currently applied to both the weights and activations; however, pruning is most often applied to only the weights of the network. In this work, we jointly apply novel uniform quantization and unstructured pruning methods to both the weights and activations of deep neural networks during training. Using our methods, we empirically evaluate the currently accepted prune-then-quantize paradigm across a wide range of computer vision tasks and observe …
Pruning is a promising approach to compress complex deep learning models in order to deploy them on resource-constrained edge devices. However, many existing pruning solutions are based on unstructured pruning, which yields models that cannot efficiently run on commodity hardware and require users to manually explore and tune the pruning process, which is time-consuming and often leads to sub-optimal results. To address these limitations, this paper presents an adaptive, activation-based, structured pruning approach to automatically and efficiently generate small, accurate, and hardware-effici…
Amazon.com Inc.（亚马逊公司）；AWS Arizona State University（亚马逊云服务亚利桑那州立大学）；Arizona State University（亚利桑那州立大学）
Deploying complex deep learning models on edge devices is challenging because they have substantial compute and memory resource requirements, whereas edge devices' resource budget is limited. To solve this problem, extensive pruning techniques have been proposed for compressing networks. Recent advances based on the Lottery Ticket Hypothesis (LTH) show that iterative model pruning tends to produce smaller and more accurate models. However, LTH research focuses on unstructured pruning, which is hardware-inefficient and difficult to accelerate on hardware platforms. In this paper, we investigate…
Duke University（杜克大学）；Black Sesame Technologies（黑芝麻科技）
With the rapid scaling up of deep neural networks (DNNs), extensive research studies on network model compression such as weight pruning have been performed for improving deployment efficiency. This work aims to advance the compression beyond the weights to neuron activations. We propose the joint regularization technique which simultaneously regulates the distribution of weights and activations. By distinguishing and leveraging the significance difference among neuron responses and connections during learning, the jointly pruned network, namely \textit{JPnet}, optimizes the sparsity of activa…
State-of-the-art neural networks are getting deeper and wider. While their performance increases with the increasing number of layers and neurons, it is crucial to design an efficient deep architecture in order to reduce computational and memory costs. Designing an efficient neural network, however, is labor intensive requiring many experiments, and fine-tunings. In this paper, we introduce network trimming which iteratively optimizes the network by pruning unimportant neurons based on analysis of their outputs on a large dataset. Our algorithm is inspired by an observation that the outputs of…
Samsung India Research Bangalore（三星印度班加罗尔研究所）；International Institute of Information Technology（国际信息技术学院）
Success of deep neural networks in diverse tasks across domains of computer vision, speech recognition and natural language processing, has necessitated understanding the dynamics of training process and also working of trained models. Two independent contributions of this paper are 1) Novel activation function for faster training convergence 2) Systematic pruning of filters of models trained irrespective of activation function. We analyze the topological transformation of the space of training samples as it gets transformed by each successive layer during training, by changing the activation …
Indian Institute of Information Technology, Sri City（斯里城印度信息技术学院）；Indian Institute of Information Technology, Allahabad（阿拉哈巴德印度信息技术学院）；DryvAmigo（DryvAmigo公司）
This paper proposes an Information Bottleneck theory based filter pruning method that uses a statistical measure called Mutual Information (MI). The MI between filters and class labels, also called \textit{Relevance}, is computed using the filter's activation maps and the annotations. The filters having High Relevance (HRel) are considered to be more important. Consequently, the least important filters, which have lower Mutual Information with the class labels, are pruned. Unlike the existing MI based pruning methods, the proposed method determines the significance of the filters purely based …
Department of Machine Learning and Neural Computing（机器学习与神经计算系）；Donders Institute for Brain, Cognition and Behaviour（大脑、认知与行为研究所）；Radboud University（拉德堡德大学）
ETH Zurich（苏黎世联邦理工学院）；University of Cambridge（剑桥大学）
Structured pruning is a well-known technique to reduce the storage size and inference cost of neural networks. The usual pruning pipeline consists of ranking the network internal filters and activations with respect to their contributions to the network performance, removing the units with the lowest contribution, and fine-tuning the network to reduce the harm induced by pruning. Recent results showed that random pruning performs on par with other metrics, given enough fine-tuning resources. In this work, we show that this is not true on a low-data regime when fine-tuning is either not possibl…
Robert Bosch GmbH（罗伯特·博世公司）；University of Freiburg（弗莱堡大学）；Institute for Autonomous Intelligent Systems（自主智能系统研究所）
Deep neural networks (DNNs) are usually over-parameterized to increase the likelihood of getting adequate initial weights by random initialization. Consequently, trained DNNs have many redundancies which can be pruned from the model to reduce complexity and improve the ability to generalize. Structural sparsity, as achieved by filter pruning, directly reduces the tensor sizes of weights and activations and is thus particularly effective for reducing complexity. We propose "Holistic Filter Pruning" (HFP), a novel approach for common DNN training that is easy to implement and enables to specify …
We propose a new formulation for pruning convolutional kernels in neural networks to enable efficient inference. We interleave greedy criteria-based pruning with fine-tuning by backpropagation - a computationally efficient procedure that maintains good generalization in the pruned network. We propose a new criterion based on Taylor expansion that approximates the change in the cost function induced by pruning network parameters. We focus on transfer learning, where large pretrained networks are adapted to specialized tasks. The proposed criterion demonstrates superior performance compared to o…
Pruning is a promising approach to compress deep learning models in order to deploy them on resource-constrained edge devices. However, many existing pruning solutions are based on unstructured pruning, which yields models that cannot efficiently run on commodity hardware; and they often require users to manually explore and tune the pruning process, which is time-consuming and often leads to sub-optimal results. To address these limitations, this paper presents Automatic Attention Pruning (AAP), an adaptive, attention-based, structured pruning approach to automatically generate small, accurat…
One-Cycle Structured Pruning via Stability-Driven Subnetwork Search
IT Application Research Center, Korea Electronics Technology Institute（韩国电子技术研究所应用信息技术中心）；Department of Intelligent Semiconductors, Soongsil University（顺慈大学智能半导体系）；Department of Computer Science & Engineering, Seoul National University（首尔国立大学计算机科学与工程系）
In recent years, most of the deep learning solutions are targeted to be deployed in mobile devices. This makes the need for development of lightweight models all the more imminent. Another solution is to optimize and prune regular deep learning models. In this paper, we tackle the problem of CNN model pruning with the help of Self-Similarity Matrix (SSM) computed from the 2D CNN filters. We propose two novel algorithms to rank and prune redundant filters which contribute similar activation maps to the output. One of the key features of our method is that there is no need of finetuning after tr…
Cranberry-Lemon University（蔓越莓柠檬大学）；University of the Witwatersrand（金山大学）
Deep Neural Networks(DNNs) have many parameters and activation data, and these both are expensive to implement. One method to reduce the size of the DNN is to quantize the pre-trained model by using a low-bit expression for weights and activations, using fine-tuning to recover the drop in accuracy. However, it is generally difficult to train neural networks which use low-bit expressions. One reason is that the weights in the middle layer of the DNN have a wide dynamic range and so when quantizing the wide dynamic range into a few bits, the step size becomes large, which leads to a large quanti…
As deep neural networks are growing in size and being increasingly deployed to more resource-limited devices, there has been a recent surge of interest in network pruning methods, which aim to remove less important weights or activations of a given network. A common limitation of most existing pruning techniques, is that they require pre-training of the network at least once before pruning, and thus we can benefit from reduction in memory and computation only at the inference time. However, reducing the training cost of neural networks with rapid structural pruning may be beneficial either to …
High bandwidth requirements are an obstacle for accelerating the training and inference of deep neural networks. Most previous research focuses on reducing the size of kernel maps for inference. We analyze parameter sparsity of six popular convolutional neural networks - AlexNet, MobileNet, ResNet-50, SqueezeNet, TinyNet, and VGG16. Of the networks considered, those using ReLU (AlexNet, SqueezeNet, VGG16) contain a high percentage of 0-valued parameters and can be statically pruned. Networks with Non-ReLU activation functions in some cases may not contain any 0-valued parameters (ResNet-50, Ti…
This paper presents an efficient and robust approach for reducing the size of deep neural networks by pruning entire neurons. It exploits maxout units for combining neurons into more complex convex functions and it makes use of a local relevance measurement that ranks neurons according to their activation on the training set for pruning them. Additionally, a parameter reduction comparison between neuron and weight pruning is shown. It will be empirically shown that the proposed neuron pruning reduces the number of parameters dramatically. The evaluation is performed on two tasks, the MNIST han…
Korea University of Technology and Education（韩国技术教育大学）
Convolutional neural networks have shown tremendous performance capabilities in computer vision tasks, but their excessive amounts of weight storage and arithmetic operations prevent them from being adopted in embedded environments. One of the solutions involves pruning, where certain unimportant weights are forced to have a value of zero. Many pruning schemes have been proposed, but these have mainly focused on the number of pruned weights. Previous pruning schemes scarcely considered ASIC or FPGA accelerator architectures. When these pruned networks are run on accelerators, the lack of consi…
National Key Laboratory for Novel Software Technology（新型软件技术国家重点实验室）；Nanjing University（南京大学）；Nanjing（南京）；China（中国）
Channel pruning is an important family of methods to speed up deep model's inference. Previous filter pruning algorithms regard channel pruning and model fine-tuning as two independent steps. This paper argues that combining them into a single end-to-end trainable system will lead to better results. We propose an efficient channel selection layer, namely AutoPruner, to find less important filters automatically in a joint training manner. Our AutoPruner takes previous activation responses as an input and generates a true binary index code for pruning. Hence, all the filters corresponding to zer…
Modern neural network architectures typically have many millions of parameters and can be pruned significantly without substantial loss in effectiveness which demonstrates they are over-parameterized. The contribution of this work is two-fold. The first is a method for approximating a multivariate Bernoulli random variable by means of a deterministic and differentiable transformation of any real-valued multivariate random variable. The second is a method for model selection by element-wise multiplication of parameters with approximate binary gates that may be computed deterministically or stoc…
National Key Laboratory for Novel Software Technology（计算机软件新技术国家重点实验室）；Nanjing University（南京大学）
This paper aims to simultaneously accelerate and compress off-the-shelf CNN models via filter pruning strategy. The importance of each filter is evaluated by the proposed entropy-based method first. Then several unimportant filters are discarded to get a smaller CNN model. Finally, fine-tuning is adopted to recover its generalization ability which is damaged during filter pruning. Our method can reduce the size of intermediate activations, which would dominate most memory footprint during model training stage but is less concerned in previous compression methods. Experiments on the ILSVRC-12 b…
Spectral Complex Autoencoder Pruning: A Fidelity-Guided Criterion for Extreme Structured Channel Compression
School of Computer, Jiangsu University of Science and Technology（江苏科技大学计算机学院）；Department of Electrical and Computer Engineering, University of Nevada, Las Vegas（内华达大学拉斯维加斯分校电子与计算机工程系）
University of Science and Technology Beijing（北京科技大学）；Hua Xia General Processor Technologies（华夏通用处理器技术公司）；General Processor Technologies（通用处理器技术公司）
Deep neural networks have been applied in many applications exhibiting extraordinary abilities in the field of computer vision. However, complex network architectures challenge efficient real-time deployment and require significant computation resources and energy costs. These challenges can be overcome through optimizations such as network compression. Network compression can often be realized with little loss of accuracy. In some cases accuracy may even improve. This paper provides a survey on two types of network compression: pruning and quantization. Pruning can be categorized as static if…
Depthwise separable convolutions and frequency-domain convolutions are two recent ideas for building efficient convolutional neural networks. They are seemingly incompatible: the vast majority of operations in depthwise separable CNNs are in pointwise convolutional layers, but pointwise layers use 1x1 kernels, which do not benefit from frequency transformation. This paper unifies these two ideas by transforming the activations, not the kernels. Our key insights are that 1) pointwise convolutions commute with frequency transformation and thus can be computed in the frequency domain without modi…
Stochastic Activation Pruning (SAP) (Dhillon et al., 2018) is a defense to adversarial examples that was attacked and found to be broken by the "Obfuscated Gradients" paper (Athalye et al., 2018). We discover a flaw in the re-implementation that artificially weakens SAP. When SAP is applied properly, the proposed attack is not effective. However, we show that a new use of the BPDA attack technique can still reduce the accuracy of SAP to 0.1%.
Sparse ActionGen: Accelerating Diffusion Policy with Real-time Pruning
Tsinghua Shenzhen International Graduate School, Tsinghua University（清华大学深圳国际研究生院，清华大学）；Department of Computer Science and Technology, Tsinghua University（清华大学计算机科学与技术系）；Department of Computer Science（计算机科学系）
Filter Pruning for Efficient CNNs via Knowledge-driven Differential Filter Sampler
East China Normal University（华东师范大学）；Xiamen University（厦门大学）；Beihang University（北京航空航天大学）；Zhongguancun Laboratory（中关村实验室）；Tencent（腾讯）；University of Sheffield（谢菲尔德大学）；University at Buffalo（布法罗大学）
Amazon AI（亚马逊人工智能）；UT Austin（得克萨斯大学奥斯汀分校）；UC Irvine（加利福尼亚大学欧文分校）；CMU（卡内基梅隆大学）；Caltech（加州理工学院）；Imperial College London（伦敦帝国理工学院）
Neural networks are known to be vulnerable to adversarial examples. Carefully chosen perturbations to real images, while imperceptible to humans, induce misclassification and threaten the reliability of deep learning systems in the wild. To guard against adversarial examples, we take inspiration from game theory and cast the problem as a minimax zero-sum game between the adversary and the model. In general, for such games, the optimal strategy for both players requires a stochastic policy, also known as a mixed strategy. In this light, we propose Stochastic Activation Pruning (SAP), a mixed st…
