Continual lifelong learning is essential to many applications. In this paper, we propose a simple but effective approach to continual deep learning. Our approach leverages the principles of deep model compression, critical weights selection, and progressive networks expansion. By enforcing their integration in an iterative manner, we introduce an incremental learning method that is scalable to the number of sequential tasks in a continual learning process. Our approach is easy to implement and owns several favorable characteristics. First, it can avoid forgetting (i.e., learn new tasks while remembering all previous tasks). Second, it allows model expansion but can maintain the model compactness when handling sequential tasks. Besides, through our compaction and selection/expansion mechanism, we show that the knowledge accumulated through learning previous tasks is helpful to build a better model for the new tasks compared to training the models independently with tasks. Experimental results show that our approach can incrementally learn a deep model tackling multiple tasks without forgetting, while the model compactness is maintained with the performance more satisfiable than individual task training.
Combining different models is a widely used paradigm in machine learning applications. While the most common approach is to form an ensemble of models and average their individual predictions, this approach is often rendered infeasible by given resource constraints in terms of memory and computation, which grow linearly with the number of models. We present a layer-wise model fusion algorithm for neural networks that utilizes optimal transport to (soft-) align neurons across the models before averaging their associated parameters. We show that this can successfully yield "one-shot" knowledge transfer (i.e, without requiring any retraining) between neural networks trained on heterogeneous non-i.i.d. data. In both i.i.d. and non-i.i.d. settings , we illustrate that our approach significantly outperforms vanilla averaging, as well as how it can serve as an efficient replacement for the ensemble with moderate fine-tuning, for standard convolutional networks (like VGG11), residual networks (like ResNet18), and multi-layer perceptrons on CIFAR10, CIFAR100, and MNIST. Finally, our approach also provides a principled way to combine the parameters of neural networks with different widths, and we explore its application for model compression. The code is available at the following link, https://github.com/sidak/otfusion.
Large language models have recently achieved state of the art performance across a wide variety of natural language tasks. Meanwhile, the size of these models and their latency have significantly increased, which makes their usage costly, and raises an interesting question: do language models need to be large? We study this question through the lens of model compression. We present a generic, structured pruning approach by parameterizing each weight matrix using its low-rank factorization, and adaptively removing rank-1 components during training. On language modeling tasks, our structured approach outperforms other unstructured and block-structured pruning baselines at various compression levels, while achieving significant speedups during both training and inference. We also demonstrate that our method can be applied to pruning adaptive word embeddings in large language models, and to pruning the BERT model on several downstream fine-tuning classification benchmarks.
Deep neural networks have significantly alleviated the burden of feature engineering, but comparable efforts are now required to determine effective architectures for these networks. Furthermore, as network sizes have become excessively large, a substantial amount of resources is invested in reducing their sizes. These challenges can be effectively addressed through the sparsification of over-complete models. In this study, we propose a fully differentiable sparsification method for deep neural networks, which can zero out unimportant parameters by directly optimizing a regularized objective function with stochastic gradient descent. Consequently, the proposed method can learn both the sparsified structure and weights of a network in an end-to-end manner. It can be directly applied to various modern deep neural networks and requires minimal modification to the training process. To the best of our knowledge, this is the first fully differentiable sparsification method.
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.
DenseNets introduce concatenation-type skip connections that achieve state-of-the-art accuracy in several computer vision tasks. In this paper, we reveal that the topology of the concatenation-type skip connections is closely related to the gradient propagation which, in turn, enables a predictable behavior of DNNs' test performance. To this end, we introduce a new metric called NN-Mass to quantify how effectively information flows through DNNs. Moreover, we empirically show that NN-Mass also works for other types of skip connections, e.g., for ResNets, Wide-ResNets (WRNs), and MobileNets, which contain addition-type skip connections (i.e., residuals or inverted residuals). As such, for both DenseNet-like CNNs and ResNets/WRNs/MobileNets, our theoretically grounded NN-Mass can identify models with similar accuracy, despite having significantly different size/compute requirements. Detailed experiments on both synthetic and real datasets (e.g., MNIST, CIFAR-10, CIFAR-100, ImageNet) provide extensive evidence for our insights. Finally, the closed-form equation of our NN-Mass enables us to design significantly compressed DenseNets (for CIFAR-10) and MobileNets (for ImageNet) directly at initialization without time-consuming training and/or searching.
Machine learning as a service (MLaaS), and algorithm marketplaces are on a rise. Data holders can easily train complex models on their data using third party provided learning codes. Training accurate ML models requires massive labeled data and advanced learning algorithms. The resulting models are considered as intellectual property of the model owners and their copyright should be protected. Also, MLaaS needs to be trusted not to embed secret information about the training data into the model, such that it could be later retrieved when the model is deployed. In this paper, we present \emph{membership encoding} for training deep neural networks and encoding the membership information, i.e. whether a data point is used for training, for a subset of training data. Membership encoding has several applications in different scenarios, including robust watermarking for model copyright protection, and also the risk analysis of stealthy data embedding privacy attacks. Our encoding algorithm can determine the membership of significantly redacted data points, and is also robust to model compression and fine-tuning. It also enables encoding a significant fraction of the training set, with negligible drop in the model's prediction accuracy.
Deep neural networks (DNNs), as the basis of object detection, will play a key role in the development of future autonomous systems with full autonomy. The autonomous systems have special requirements of real-time, energy-efficient implementations of DNNs on a power-constrained system. Two research thrusts are dedicated to performance and energy efficiency enhancement of the inference phase of DNNs. The first one is model compression techniques while the second is efficient hardware implementation. Recent works on extremely-low-bit CNNs such as the binary neural network (BNN) and XNOR-Net replace the traditional floating-point operations with binary bit operations which significantly reduces the memory bandwidth and storage requirement. However, it suffers from non-negligible accuracy loss and underutilized digital signal processing (DSP) blocks of FPGAs. To overcome these limitations, this paper proposes REQ-YOLO, a resource-aware, systematic weight quantization framework for object detection, considering both algorithm and hardware resource aspects in object detection. We adopt the block-circulant matrix method and propose a heterogeneous weight quantization using the Alternating Direction Method of Multipliers (ADMM), an effective optimization technique for general, non-convex optimization problems. To achieve real-time, highly-efficient implementations on FPGA, we present the detailed hardware implementation of block circulant matrices on CONV layers and develop an efficient processing element (PE) structure supporting the heterogeneous weight quantization, CONV dataflow and pipelining techniques, design optimization, and a template-based automatic synthesis framework to optimally exploit hardware resource. Experimental results show that our proposed REQ-YOLO framework can significantly compress the YOLO model while introducing very small accuracy degradation.
Deep Neural Network (DNN) is powerful but computationally expensive and memory intensive, thus impeding its practical usage on resource-constrained front-end devices. DNN pruning is an approach for deep model compression, which aims at eliminating some parameters with tolerable performance degradation. In this paper, we propose a novel momentum-SGD-based optimization method to reduce the network complexity by on-the-fly pruning. Concretely, given a global compression ratio, we categorize all the parameters into two parts at each training iteration which are updated using different rules. In this way, we gradually zero out the redundant parameters, as we update them using only the ordinary weight decay but no gradients derived from the objective function. As a departure from prior methods that require heavy human works to tune the layer-wise sparsity ratios, prune by solving complicated non-differentiable problems or finetune the model after pruning, our method is characterized by 1) global compression that automatically finds the appropriate per-layer sparsity ratios; 2) end-to-end training; 3) no need for a time-consuming re-training process after pruning; and 4) superior capability to find better winning tickets which have won the initialization lottery.
Pretrained language models like BERT have achieved good results on NLP tasks, but are impractical on resource-limited devices due to memory footprint. A large fraction of this footprint comes from the input embeddings with large input vocabulary and embedding dimensions. Existing knowledge distillation methods used for model compression cannot be directly applied to train student models with reduced vocabulary sizes. To this end, we propose a distillation method to align the teacher and student embeddings via mixed-vocabulary training. Our method compresses BERT-LARGE to a task-agnostic model with smaller vocabulary and hidden dimensions, which is an order of magnitude smaller than other distilled BERT models and offers a better size-accuracy trade-off on language understanding benchmarks as well as a practical dialogue task.
We propose a method for specializing deep detectors and trackers to restricted settings. Our approach is designed with the following goals in mind: (a) Improving accuracy in restricted domains; (b) preventing overfitting to new domains and forgetting of generalized capabilities; (c) aggressive model compression and acceleration. To this end, we propose a novel loss that balances compression and acceleration of a deep learning model vs. loss of generalization capabilities. We apply our method to the existing tracker and detector models. We report detection results on the VIRAT and CAVIAR data sets. These results show our method to offer unprecedented compression rates along with improved detection. We apply our loss for tracker compression at test time, as it processes each video. Our tests on the OTB2015 benchmark show that applying compression during test time actually improves tracking performance.
Today's deep neural networks require substantial computation resources for their training, storage, and inference, which limits their effective use on resource-constrained devices. Many recent research activities explore different options for compressing and optimizing deep models. On the one hand, in many real-world applications, we face the data imbalance challenge, i.e. when the number of labeled instances of one class considerably outweighs the number of labeled instances of the other class. On the other hand, applications may pose a class imbalance problem, i.e. higher number of false positives produced when training a model and optimizing its performance may be tolerable, yet the number of false negatives must stay low. The problem originates from the fact that some classes are more important for the application than others, e.g. detection problems in medical and surveillance domains. Motivated by the success of the lottery ticket hypothesis, in this paper we propose an iterative deep model compression technique, which keeps the number of false negatives of the compressed model close to the one of the original model at the price of increasing the number of false positives if necessary. Our experimental evaluation using two benchmark data sets shows that the resulting compressed sub-networks 1) achieve up to 35% lower number of false negatives than the compressed model without class optimization, 2) provide an overall higher AUC_ROC measure, and 3) use up to 99% fewer parameters compared to the original network.
Model compression techniques on Deep Neural Network (DNN) have been widely acknowledged as an effective way to achieve acceleration on a variety of platforms, and DNN weight pruning is a straightforward and effective method. There are currently two mainstreams of pruning methods representing two extremes of pruning regularity: non-structured, fine-grained pruning can achieve high sparsity and accuracy, but is not hardware friendly; structured, coarse-grained pruning exploits hardware-efficient structures in pruning, but suffers from accuracy drop when the pruning rate is high. In this paper, we introduce PCONV, comprising a new sparsity dimension, -- fine-grained pruning patterns inside the coarse-grained structures. PCONV comprises two types of sparsities, Sparse Convolution Patterns (SCP) which is generated from intra-convolution kernel pruning and connectivity sparsity generated from inter-convolution kernel pruning. Essentially, SCP enhances accuracy due to its special vision properties, and connectivity sparsity increases pruning rate while maintaining balanced workload on filter computation. To deploy PCONV, we develop a novel compiler-assisted DNN inference framework and execute PCONV models in real-time without accuracy compromise, which cannot be achieved in prior work. Our experimental results show that, PCONV outperforms three state-of-art end-to-end DNN frameworks, TensorFlow-Lite, TVM, and Alibaba Mobile Neural Network with speedup up to 39.2x, 11.4x, and 6.3x, respectively, with no accuracy loss. Mobile devices can achieve real-time inference on large-scale DNNs.
Pruning is one of the most effective model reduction techniques. Deep networks require massive computation and such models need to be compressed to bring them on edge devices. Most existing pruning techniques are focused on vision-based models like convolutional networks, while text-based models are still evolving. The emergence of multi-modal multi-task learning calls for a general method that works on vision and text architectures simultaneously. We introduce a \emph{differentiable mask}, that induces sparsity on various granularity to fill this gap. We apply our method successfully to prune weights, filters, subnetwork of a convolutional architecture, as well as nodes of a recurrent network.
We introduce knowledge distillation for end-to-end person search. End-to-End methods are the current state-of-the-art for person search that solve both detection and re-identification jointly. These approaches for joint optimization show their largest drop in performance due to a sub-optimal detector. We propose two distinct approaches for extra supervision of end-to-end person search methods in a teacher-student setting. The first is adopted from state-of-the-art knowledge distillation in object detection. We employ this to supervise the detector of our person search model at various levels using a specialized detector. The second approach is new, simple and yet considerably more effective. This distills knowledge from a teacher re-identification technique via a pre-computed look-up table of ID features. It relaxes the learning of identification features and allows the student to focus on the detection task. This procedure not only helps fixing the sub-optimal detector training in the joint optimization and simultaneously improving the person search, but also closes the performance gap between the teacher and the student for model compression in this case. Overall, we demonstrate significant improvements for two recent state-of-the-art methods using our proposed knowledge distillation approach on two benchmark datasets. Moreover, on the model compression task our approach brings the performance of smaller models on par with the larger models.
Recent developments in natural language representations have been accompanied by large and expensive models that leverage vast amounts of general-domain text through self-supervised pre-training. Due to the cost of applying such models to down-stream tasks, several model compression techniques on pre-trained language representations have been proposed (Sun et al., 2019; Sanh, 2019). However, surprisingly, the simple baseline of just pre-training and fine-tuning compact models has been overlooked. In this paper, we first show that pre-training remains important in the context of smaller architectures, and fine-tuning pre-trained compact models can be competitive to more elaborate methods proposed in concurrent work. Starting with pre-trained compact models, we then explore transferring task knowledge from large fine-tuned models through standard knowledge distillation. The resulting simple, yet effective and general algorithm, Pre-trained Distillation, brings further improvements. Through extensive experiments, we more generally explore the interaction between pre-training and distillation under two variables that have been under-studied: model size and properties of unlabeled task data. One surprising observation is that they have a compound effect even when sequentially applied on the same data. To accelerate future research, we will make our 24 pre-trained miniature BERT models publicly available.
Despite their ubiquity in NLP tasks, Long Short-Term Memory (LSTM) networks suffer from computational inefficiencies caused by inherent unparallelizable recurrences, which further aggravates as LSTMs require more parameters for larger memory capacity. In this paper, we propose to apply low-rank matrix factorization (MF) algorithms to different recurrences in LSTMs, and explore the effectiveness on different NLP tasks and model components. We discover that additive recurrence is more important than multiplicative recurrence, and explain this by identifying meaningful correlations between matrix norms and compression performance. We compare our approach across two settings: 1) compressing core LSTM recurrences in language models, 2) compressing biLSTM layers of ELMo evaluated in three downstream NLP tasks.
Pre-trained language models such as BERT have proven to be highly effective for natural language processing (NLP) tasks. However, the high demand for computing resources in training such models hinders their application in practice. In order to alleviate this resource hunger in large-scale model training, we propose a Patient Knowledge Distillation approach to compress an original large model (teacher) into an equally-effective lightweight shallow network (student). Different from previous knowledge distillation methods, which only use the output from the last layer of the teacher network for distillation, our student model patiently learns from multiple intermediate layers of the teacher model for incremental knowledge extraction, following two strategies: ($i$) PKD-Last: learning from the last $k$ layers; and ($ii$) PKD-Skip: learning from every $k$ layers. These two patient distillation schemes enable the exploitation of rich information in the teacher's hidden layers, and encourage the student model to patiently learn from and imitate the teacher through a multi-layer distillation process. Empirically, this translates into improved results on multiple NLP tasks with significant gain in training efficiency, without sacrificing model accuracy.
The state-of-art DNN structures involve intensive computation and high memory storage. To mitigate the challenges, the memristor crossbar array has emerged as an intrinsically suitable matrix computation and low-power acceleration framework for DNN applications. However, the high accuracy solution for extreme model compression on memristor crossbar array architecture is still waiting for unraveling. In this paper, we propose a memristor-based DNN framework which combines both structured weight pruning and quantization by incorporating alternating direction method of multipliers (ADMM) algorithm for better pruning and quantization performance. We also discover the non-optimality of the ADMM solution in weight pruning and the unused data path in a structured pruned model. Motivated by these discoveries, we design a software-hardware co-optimization framework which contains the first proposed Network Purification and Unused Path Removal algorithms targeting on post-processing a structured pruned model after ADMM steps. By taking memristor hardware constraints into our whole framework, we achieve extreme high compression ratio on the state-of-art neural network structures with minimum accuracy loss. For quantizing structured pruned model, our framework achieves nearly no accuracy loss after quantizing weights to 8-bit memristor weight representation. We share our models at anonymous link https://bit.ly/2VnMUy0.
In recent years, convolutional networks have demonstrated unprecedented performance in the image restoration task of super-resolution (SR). SR entails the upscaling of a single low-resolution image in order to meet application-specific image quality demands and plays a key role in mobile devices. To comply with privacy regulations and reduce the overhead of cloud computing, executing SR models locally on-device constitutes a key alternative approach. Nevertheless, the excessive compute and memory requirements of SR workloads pose a challenge in mapping SR networks on resource-constrained mobile platforms. This work presents MobiSR, a novel framework for performing efficient super-resolution on-device. Given a target mobile platform, the proposed framework considers popular model compression techniques and traverses the design space to reach the highest performing trade-off between image quality and processing speed. At run time, a novel scheduler dispatches incoming image patches to the appropriate model-engine pair based on the patch's estimated upscaling difficulty in order to meet the required image quality with minimum processing latency. Quantitative evaluation shows that the proposed framework yields on-device SR designs that achieve an average speedup of 2.13x over highly-optimized parallel difficulty-unaware mappings and 4.79x over highly-optimized single compute engine implementations.