Three dimensional convolutional neural networks (3DCNNs) have been applied in many tasks, e.g., video and 3D point cloud recognition. However, due to the higher dimension of convolutional kernels, the space complexity of 3DCNNs is generally larger than that of traditional two dimensional convolutional neural networks (2DCNNs). To miniaturize 3DCNNs for the deployment in confining environments such as embedded devices, neural network compression is a promising approach. In this work, we adopt the tensor train (TT) decomposition, a straightforward and simple in situ training compression method, to shrink the 3DCNN models. Through proposing tensorizing 3D convolutional kernels in TT format, we investigate how to select appropriate TT ranks for achieving higher compression ratio. We have also discussed the redundancy of 3D convolutional kernels for compression, core significance and future directions of this work, as well as the theoretical computation complexity versus practical executing time of convolution in TT. In the light of multiple contrast experiments based on VIVA challenge, UCF11, and UCF101 datasets, we conclude that TT decomposition can compress 3DCNNs by around one hundred times without significant accuracy loss, which will enable its applications in extensive real world scenarios.
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 time without appreciable loss in accuracy. However, finding the best compression strategy and corresponding target sparsity for a given DNN, hardware platform, and optimization objective currently requires expensive, frequently manual, trial-and-error experimentation. In this paper, we introduce a programmable system for model compression called Condensa. Users programmatically compose simple operators, in Python, to build more complex and practically interesting compression strategies. Given a strategy and user-provided objective (such as minimization of running time), Condensa uses a novel Bayesian optimization-based algorithm to automatically infer desirable sparsities. Our experiments on four real-world DNNs demonstrate memory footprint and hardware runtime throughput improvements of 188x and 2.59x, respectively, using at most ten samples per search. We have released a reference implementation of Condensa at https://github.com/NVlabs/condensa.
Among the neural network compression techniques, knowledge distillation is an effective one which forces a simpler student network to mimic the output of a larger teacher network. However, most of such model distillation methods focus on the image-level classification task. Directly adapting these methods to the task of semantic segmentation only brings marginal improvements. In this paper, we propose a simple, yet effective knowledge representation referred to as pixel-wise feature similarities (PFS) to tackle the challenging distillation problem of semantic segmentation. The developed PFS encodes spatial structural information for each pixel location of the high-level convolutional features, which helps guide the distillation process in an easier way. Furthermore, a novel weighted pixel-level soft prediction imitation approach is proposed to enable the student network to selectively mimic the teacher network's output, according to their pixel-wise knowledge-gaps. Extensive experiments are conducted on the challenging datasets of Pascal VOC 2012, ADE20K and Pascal Context. Our approach brings significant performance improvements compared to several strong baselines and achieves new state-of-the-art results.
There has recently been an increasing desire to evaluate neural networks locally on computationally-limited devices in order to exploit their recent effectiveness for several applications; such effectiveness has nevertheless come together with a considerable increase in the size of modern neural networks, which constitute a major downside in several of the aforementioned computationally-limited settings. There has thus been a demand of compression techniques for neural networks. Several proposal in this direction have been made, which famously include hashing-based methods and pruning-based ones. However, the evaluation of the efficacy of these techniques has so far been heterogeneous, with no clear evidence in favor of any of them over the others. The goal of this work is to address this latter issue by providing a comparative study. While most previous studies test the capability of a technique in reducing the number of parameters of state-of-the-art networks , we follow [CWT + 15] in evaluating their performance on basic ar-chitectures on the MNIST dataset and variants of it, which allows for a clearer analysis of some aspects of their behavior. To the best of our knowledge, we are the first to directly compare famous approaches such as HashedNet, Optimal Brain Damage (OBD), and magnitude-based pruning with L1 and L2 regularization among them and against equivalent-size feed-forward neural networks with simple (fully-connected) and structural (convolutional) neural networks. Rather surprisingly, our experiments show that (iterative) pruning-based methods are substantially better than the HashedNet architecture, whose compression doesn't appear advantageous to a carefully chosen convolutional network. We also show that, when the compression level is high, the famous OBD pruning heuristics deteriorates to the point of being less efficient than simple magnitude-based techniques.
Deep Neural Networks (DNNs) are applied in a wide range of usecases. There is an increased demand for deploying DNNs on devices that do not have abundant resources such as memory and computation units. Recently, network compression through a variety of techniques such as pruning and quantization have been proposed to reduce the resource requirement. A key parameter that all existing compression techniques are sensitive to is the compression ratio (e.g., pruning sparsity, quantization bitwidth) of each layer. Traditional solutions treat the compression ratios of each layer as hyper-parameters, and tune them using human heuristic. Recent researchers start using black-box hyper-parameter optimizations, but they will introduce new hyper-parameters and have efficiency issue. In this paper, we propose a framework to jointly prune and quantize the DNNs automatically according to a target model size without using any hyper-parameters to manually set the compression ratio for each layer. In the experiments, we show that our framework can compress the weights data of ResNet-50 to be 836$\times$ smaller without accuracy loss on CIFAR-10, and compress AlexNet to be 205$\times$ smaller without accuracy loss on ImageNet classification.
Neural networks have been notorious for being computationally expensive. This is mainly because neural networks are often over-parametrized and most likely have redundant nodes or layers as they are getting deeper and wider. Their demand for hardware resources prohibits their extensive use in embedded devices and puts restrictions on tasks like real-time image classification or object detection. In this work, we propose a network-agnostic model compression method infused with a novel dynamical clustering approach to reduce the computational cost and memory footprint of deep neural networks. We evaluated our new compression method on five different state-of-the-art image classification and object detection networks. In classification networks, we pruned about 95% of network parameters. In advanced detection networks such as YOLOv3, our proposed compression method managed to reduce the model parameters up to 59.70% which yielded 110X less memory without sacrificing much in accuracy.
This paper proposes an improved training algorithm for binary neural networks in which both weights and activations are binary numbers. A key but fairly overlooked feature of the current state-of-the-art method of XNOR-Net is the use of analytically calculated real-valued scaling factors for re-weighting the output of binary convolutions. We argue that analytic calculation of these factors is sub-optimal. Instead, in this work, we make the following contributions: (a) we propose to fuse the activation and weight scaling factors into a single one that is learned discriminatively via backpropagation. (b) More importantly, we explore several ways of constructing the shape of the scale factors while keeping the computational budget fixed. (c) We empirically measure the accuracy of our approximations and show that they are significantly more accurate than the analytically calculated one. (d) We show that our approach significantly outperforms XNOR-Net within the same computational budget when tested on the challenging task of ImageNet classification, offering up to 6\% accuracy gain.
Weight and activation binarization is an effective approach to deep neural network compression and can accelerate the inference by leveraging bitwise operations. Although many binarization methods have improved the accuracy of the model by minimizing the quantization error in forward propagation, there remains a noticeable performance gap between the binarized model and the full-precision one. Our empirical study indicates that the quantization brings information loss in both forward and backward propagation, which is the bottleneck of training accurate binary neural networks. To address these issues, we propose an Information Retention Network (IR-Net) to retain the information that consists in the forward activations and backward gradients. IR-Net mainly relies on two technical contributions: (1) Libra Parameter Binarization (Libra-PB): simultaneously minimizing both quantization error and information loss of parameters by balanced and standardized weights in forward propagation; (2) Error Decay Estimator (EDE): minimizing the information loss of gradients by gradually approximating the sign function in backward propagation, jointly considering the updating ability and accurate gradients. We are the first to investigate both forward and backward processes of binary networks from the unified information perspective, which provides new insight into the mechanism of network binarization. Comprehensive experiments with various network structures on CIFAR-10 and ImageNet datasets manifest that the proposed IR-Net can consistently outperform state-of-the-art quantization methods.
Convolutional neural networks (CNNs) based solutions have achieved state-of-the-art performances for many computer vision tasks, including classification and super-resolution of images. Usually the success of these methods comes with a cost of millions of parameters due to stacking deep convolutional layers. Moreover, quite a large number of filters are also used for a single convolutional layer, which exaggerates the parameter burden of current methods. Thus, in this paper, we try to reduce the number of parameters of CNNs by learning a basis of the filters in convolutional layers. For the forward pass, the learned basis is used to approximate the original filters and then used as parameters for the convolutional layers. We validate our proposed solution for multiple CNN architectures on image classification and image super-resolution benchmarks and compare favorably to the existing state-of-the-art in terms of reduction of parameters and preservation of accuracy.
The field of video compression has developed some of the most sophisticated and efficient compression algorithms known in the literature, enabling very high compressibility for little loss of information. Whilst some of these techniques are domain specific, many of their underlying principles are universal in that they can be adapted and applied for compressing different types of data. In this work we present DeepCABAC, a compression algorithm for deep neural networks that is based on one of the state-of-the-art video coding techniques. Concretely, it applies a Context-based Adaptive Binary Arithmetic Coder (CABAC) to the network's parameters, which was originally designed for the H.264/AVC video coding standard and became the state-of-the-art for lossless compression. Moreover, DeepCABAC employs a novel quantization scheme that minimizes the rate-distortion function while simultaneously taking the impact of quantization onto the accuracy of the network into account. Experimental results show that DeepCABAC consistently attains higher compression rates than previously proposed coding techniques for neural network compression. For instance, it is able to compress the VGG16 ImageNet model by x63.6 with no loss of accuracy, thus being able to represent the entire network with merely 8.7MB. The source code for encoding and decoding can be found at https://github.com/fraunhoferhhi/DeepCABAC.
Knowledge distillation is a widely applicable technique for training a student neural network under the guidance of a trained teacher network. For example, in neural network compression, a high-capacity teacher is distilled to train a compact student; in privileged learning, a teacher trained with privileged data is distilled to train a student without access to that data. The distillation loss determines how a teacher's knowledge is captured and transferred to the student. In this paper, we propose a new form of knowledge distillation loss that is inspired by the observation that semantically similar inputs tend to elicit similar activation patterns in a trained network. Similarity-preserving knowledge distillation guides the training of a student network such that input pairs that produce similar (dissimilar) activations in the teacher network produce similar (dissimilar) activations in the student network. In contrast to previous distillation methods, the student is not required to mimic the representation space of the teacher, but rather to preserve the pairwise similarities in its own representation space. Experiments on three public datasets demonstrate the potential of our approach.
With the development of deep neural networks, the size of network models becomes larger and larger. Model compression has become an urgent need for deploying these network models to mobile or embedded devices. Model quantization is a representative model compression technique. Although a lot of quantization methods have been proposed, many of them suffer from a high quantization error caused by a long-tail distribution of network weights. In this paper, we propose a novel quantization method, called weight normalization based quantization (WNQ), for model compression. WNQ adopts weight normalization to avoid the long-tail distribution of network weights and subsequently reduces the quantization error. Experiments on CIFAR-100 and ImageNet show that WNQ can outperform other baselines to achieve state-of-the-art performance.
Neural network compression empowers the effective yet unwieldy deep convolutional neural networks (CNN) to be deployed in resource-constrained scenarios. Most state-of-the-art approaches prune the model in filter-level according to the "importance" of filters. Despite their success, we notice they suffer from at least two of the following problems: 1) The redundancy among filters is not considered because the importance is evaluated independently. 2) Cross-layer filter comparison is unachievable since the importance is defined locally within each layer. Consequently, we must manually specify layer-wise pruning ratios. 3) They are prone to generate sub-optimal solutions because they neglect the inequality between reducing parameters and reducing computational cost. Reducing the same number of parameters in different positions in the network may reduce different computational cost. To address the above problems, we develop a novel algorithm named as COP (correlation-based pruning), which can detect the redundant filters efficiently. We enable the cross-layer filter comparison through global normalization. We add parameter-quantity and computational-cost regularization terms to the importance, which enables the users to customize the compression according to their preference (smaller or faster). Extensive experiments have shown COP outperforms the others significantly. The code is released at https://github.com/ZJULearning/COP.
Neural networks are easier to optimise when they have many more weights than are required for modelling the mapping from inputs to outputs. This suggests a two-stage learning procedure that first learns a large net and then prunes away connections or hidden units. But standard training does not necessarily encourage nets to be amenable to pruning. We introduce targeted dropout, a method for training a neural network so that it is robust to subsequent pruning. Before computing the gradients for each weight update, targeted dropout stochastically selects a set of units or weights to be dropped using a simple self-reinforcing sparsity criterion and then computes the gradients for the remaining weights. The resulting network is robust to post hoc pruning of weights or units that frequently occur in the dropped sets. The method improves upon more complicated sparsifying regularisers while being simple to implement and easy to tune.
We present DeepCABAC, a novel context-adaptive binary arithmetic coder for compressing deep neural networks. It quantizes each weight parameter by minimizing a weighted rate-distortion function, which implicitly takes the impact of quantization on to the accuracy of the network into account. Subsequently, it compresses the quantized values into a bitstream representation with minimal redundancies. We show that DeepCABAC is able to reach very high compression ratios across a wide set of different network architectures and datasets. For instance, we are able to compress by x63.6 the VGG16 ImageNet model with no loss of accuracy, thus being able to represent the entire network with merely 8.7MB.
Deep neural networks (DNNs) have become the driving force behind recent artificial intelligence (AI) research. An important problem with implementing a neural network is the design of its architecture. Typically, such an architecture is obtained manually by exploring its hyperparameter space and kept fixed during training. This approach is time-consuming and inefficient. Another issue is that modern neural networks often contain millions of parameters, whereas many applications and devices require small inference models. However, efforts to migrate DNNs to such devices typically entail a significant loss of classification accuracy. To address these challenges, we propose a two-step neural network synthesis methodology, called DR+SCANN, that combines two complementary approaches to design compact and accurate DNNs. At the core of our framework is the SCANN methodology that uses three basic architecture-changing operations, namely connection growth, neuron growth, and connection pruning, to synthesize feed-forward architectures with arbitrary structure. SCANN encapsulates three synthesis methodologies that apply a repeated grow-and-prune paradigm to three architectural starting points. DR+SCANN combines the SCANN methodology with dataset dimensionality reduction to alleviate the curse of dimensionality. We demonstrate the efficacy of SCANN and DR+SCANN on various image and non-image datasets. We evaluate SCANN on MNIST and ImageNet benchmarks. In addition, we also evaluate the efficacy of using dimensionality reduction alongside SCANN (DR+SCANN) on nine small to medium-size datasets. We also show that our synthesis methodology yields neural networks that are much better at navigating the accuracy vs. energy efficiency space. This would enable neural network-based inference even on Internet-of-Things sensors.
Neural Networks accomplish amazing things, but they suffer from computational and memory bottlenecks that restrict their usage. Nowhere can this be better seen than in the mobile space, where specialized hardware is being created just to satisfy the demand for neural networks. Previous studies have shown that neural networks have vastly more connections than they actually need to do their work. This thesis develops a method that can compress networks to less than 10% of memory and less than 25% of computational power, without loss of accuracy, and without creating sparse networks that require special code to run.
Learning portable neural networks is very essential for computer vision for the purpose that pre-trained heavy deep models can be well applied on edge devices such as mobile phones and micro sensors. Most existing deep neural network compression and speed-up methods are very effective for training compact deep models, when we can directly access the training dataset. However, training data for the given deep network are often unavailable due to some practice problems (e.g. privacy, legal issue, and transmission), and the architecture of the given network are also unknown except some interfaces. To this end, we propose a novel framework for training efficient deep neural networks by exploiting generative adversarial networks (GANs). To be specific, the pre-trained teacher networks are regarded as a fixed discriminator and the generator is utilized for derivating training samples which can obtain the maximum response on the discriminator. Then, an efficient network with smaller model size and computational complexity is trained using the generated data and the teacher network, simultaneously. Efficient student networks learned using the proposed Data-Free Learning (DAFL) method achieve 92.22% and 74.47% accuracies using ResNet-18 without any training data on the CIFAR-10 and CIFAR-100 datasets, respectively. Meanwhile, our student network obtains an 80.56% accuracy on the CelebA benchmark.
One of the current challenges in machine learning is how to deal with data coming at increasing rates in data streams. New predictive learning strategies are needed to cope with the high throughput data and concept drift. One of the data stream mining tasks where new learning strategies are needed is multi-target regression, due to its applicability in a high number of real world problems. While reliable and effective learning strategies have been proposed for batch multi-target regression, few have been proposed for multi-target online learning in data streams. Besides, most of the existing solutions do not consider the occurrence of inter-target correlations when making predictions. In this work, we propose a novel online learning strategy for multi-target regression in data streams. The proposed strategy extends existing online decision tree learning algorithm to explore inter-target dependencies while making predictions. For such, the proposed strategy, called Stacked Single-target Hoeffding Tree (SST-HT), uses the inter-target dependencies as an additional information source to enhance predictive accuracy. Throughout an extensive experimental setup, we evaluate our proposal against state-of-the-art decision tree-based algorithms for online multi-target regression. According to the experimental results, SST-HT presents superior predictive accuracy, with a small increase in the processing time and memory requirements.
The low-rank tensor approximation is very promising for the compression of deep neural networks. We propose a new simple and efficient iterative approach, which alternates low-rank factorization with a smart rank selection and fine-tuning. We demonstrate the efficiency of our method comparing to non-iterative ones. Our approach improves the compression rate while maintaining the accuracy for a variety of tasks.