The size of deep learning models in artificial intelligence (AI) software is increasing rapidly, which hinders the large-scale deployment on resource-restricted devices (e.g., smartphones). To mitigate this issue, AI software compression plays a crucial role, which aims to compress model size while keeping high performance. However, the intrinsic defects in the big model may be inherited by the compressed one. Such defects may be easily leveraged by attackers, since the compressed models are usually deployed in a large number of devices without adequate protection. In this paper, we try to address the safe model compression problem from a safety-performance co-optimization perspective. Specifically, inspired by the test-driven development (TDD) paradigm in software engineering, we propose a test-driven sparse training framework called SafeCompress. By simulating the attack mechanism as the safety test, SafeCompress can automatically compress a big model to a small one following the dynamic sparse training paradigm. Further, considering a representative attack, i.e., membership inference attack (MIA), we develop a concrete safe model compression mechanism, called MIA-SafeCompress. Extensive experiments are conducted to evaluate MIA-SafeCompress on five datasets for both computer vision and natural language processing tasks. The results verify the effectiveness and generalization of our method. We also discuss how to adapt SafeCompress to other attacks besides MIA, demonstrating the flexibility of SafeCompress.
Graph Neural Networks (GNNs) are widely used to perform different machine learning tasks on graphs. As the size of the graphs grows, and the GNNs get deeper, training and inference time become costly in addition to the memory requirement. Thus, without sacrificing accuracy, graph sparsification, or model compression becomes a viable approach for graph learning tasks. A few existing techniques only study the sparsification of graphs and GNN models. In this paper, we develop a SparseGCN pipeline to study all possible sparsification in GNN. We provide a theoretical analysis and empirically show that it can add up to 11.6\% additional sparsity to the embedding matrix without sacrificing the accuracy of the commonly used benchmark graph datasets.
Deep neural networks (DNNs) have proven successful in a wide variety of applications such as speech recognition and synthesis, computer vision, machine translation, and game playing, to name but a few. However, existing deep neural network models are computationally expensive and memory intensive, hindering their deployment in devices with low memory resources or in applications with strict latency requirements. Therefore, a natural thought is to perform model compression and acceleration in deep networks without significantly decreasing the model performance, which is what we call reducing the complexity. In the following work, we try reducing the complexity of state of the art LSTM models for natural language tasks such as text classification, by distilling their knowledge to CNN based models, thus reducing the inference time(or latency) during testing.
We develop a uniform theoretical approach towards the analysis of various neural network connectivity architectures by introducing the notion of a quiver neural network. Inspired by quiver representation theory in mathematics, this approach gives a compact way to capture elaborate data flows in complex network architectures. As an application, we use parameter space symmetries to prove a lossless model compression algorithm for quiver neural networks with certain non-pointwise activations known as rescaling activations. In the case of radial rescaling activations, we prove that training the compressed model with gradient descent is equivalent to training the original model with projected gradient descent.
Advancements in deep learning are often associated with increasing model sizes. The model size dramatically affects the deployment cost and latency of deep models. For instance, models like BERT cannot be deployed on edge devices and mobiles due to their sheer size. As a result, most advances in Deep Learning are yet to reach the edge. Model compression has sought much-deserved attention in literature across natural language processing, vision, and recommendation domains. This paper proposes a model-agnostic, cache-friendly model compression approach: Random Operation Access Specific Tile (ROAST) hashing. ROAST collapses the parameters by clubbing them through a lightweight mapping. Notably, while clubbing these parameters, ROAST utilizes cache hierarchies by aligning the memory access pattern with the parameter access pattern. ROAST is up to $\sim 25 \times$ faster to train and $\sim 50 \times$ faster to infer than the popular parameter sharing method HashedNet. Additionally, ROAST introduces global weight sharing, which is empirically and theoretically superior to local weight sharing in HashedNet, and can be of independent interest in itself. With ROAST, we present the first compressed BERT, which is $100\times - 1000\times$ smaller but does not result in quality degradation. These compression levels on universal architecture like transformers are promising for the future of SOTA model deployment on resource-constrained devices like mobile and edge devices
The number of mobile robots with constrained computing resources that need to execute complex machine learning models has been increasing during the past decade. Commonly, these robots rely on edge infrastructure accessible over wireless communication to execute heavy computational complex tasks. However, the edge might become unavailable and, consequently, oblige the execution of the tasks on the robot. This work focuses on making it possible to execute the tasks on the robots by reducing the complexity and the total number of parameters of pre-trained computer vision models. This is achieved by using model compression techniques such as Pruning and Knowledge Distillation. These compression techniques have strong theoretical and practical foundations, but their combined usage has not been widely explored in the literature. Therefore, this work especially focuses on investigating the effects of combining these two compression techniques. The results of this work reveal that up to 90% of the total number of parameters of a computer vision model can be removed without any considerable reduction in the model's accuracy.
Deep neural networks have exhibited remarkable performance in image super-resolution (SR) tasks by learning a mapping from low-resolution (LR) images to high-resolution (HR) images. However, the SR problem is typically an ill-posed problem and existing methods would come with several limitations. First, the possible mapping space of SR can be extremely large since there may exist many different HR images that can be super-resolved from the same LR image. As a result, it is hard to directly learn a promising SR mapping from such a large space. Second, it is often inevitable to develop very large models with extremely high computational cost to yield promising SR performance. In practice, one can use model compression techniques to obtain compact models by reducing model redundancy. Nevertheless, it is hard for existing model compression methods to accurately identify the redundant components due to the extremely large SR mapping space. To alleviate the first challenge, we propose a dual regression learning scheme to reduce the space of possible SR mappings. Specifically, in addition to the mapping from LR to HR images, we learn an additional dual regression mapping to estimate the downsampling kernel and reconstruct LR images. In this way, the dual mapping acts as a constraint to reduce the space of possible mappings. To address the second challenge, we propose a dual regression compression (DRC) method to reduce model redundancy in both layer-level and channel-level based on channel pruning. Specifically, we first develop a channel number search method that minimizes the dual regression loss to determine the redundancy of each layer. Given the searched channel numbers, we further exploit the dual regression manner to evaluate the importance of channels and prune the redundant ones. Extensive experiments show the effectiveness of our method in obtaining accurate and efficient SR models.
In 6G mobile communication systems, various AI-based network functions and applications have been standardized. Federated learning (FL) is adopted as the core learning architecture for 6G systems to avoid privacy leakage from mobile user data. However, in FL, users with non-independent and identically distributed (non-IID) datasets can deteriorate the performance of the global model because the convergence direction of the gradient for each dataset is different, thereby inducing a weight divergence problem. To address this problem, we propose a novel diffusion strategy for machine learning (ML) models (FedDif) to maximize the performance of the global model with non-IID data. FedDif enables the local model to learn different distributions before parameter aggregation by passing the local models through users via device-to-device communication. Furthermore, we theoretically demonstrate that FedDif can circumvent the weight-divergence problem. Based on this theory, we propose a communication-efficient diffusion strategy for ML models that can determine the trade-off between learning performance and communication cost using auction theory. The experimental results show that FedDif improves the top-1 test accuracy by up to 34.89\% and reduces communication costs by 14.6% to a maximum of 63.49%.
Deploying Machine learning (ML) on milliwatt-scale edge devices (tinyML) is gaining popularity due to recent breakthroughs in ML and Internet of Things (IoT). Most tinyML research focuses on model compression techniques that trade accuracy (and model capacity) for compact models to fit into the KB-sized tiny-edge devices. In this paper, we show how such models can be enhanced by the addition of an early exit intermediate classifier. If the intermediate classifier exhibits sufficient confidence in its prediction, the network exits early thereby, resulting in considerable savings in time. Although early exit classifiers have been proposed in previous work, these previous proposals focus on large networks, making their techniques suboptimal/impractical for tinyML applications. Our technique is optimized specifically for tiny-CNN sized models. In addition, we present a method to alleviate the effect of network overthinking by leveraging the representations learned by the early exit. We evaluate T-RecX on three CNNs from the MLPerf tiny benchmark suite for image classification, keyword spotting and visual wake word detection tasks. Our results show that T-RecX 1) improves the accuracy of baseline network, 2) achieves 31.58% average reduction in FLOPS in exchange for one percent accuracy across all evaluated models. Furthermore, we show that our methods consistently outperform popular prior works on the tiny-CNNs we evaluate.
As a promising approach in model compression, knowledge distillation improves the performance of a compact model by transferring the knowledge from a cumbersome one. The kind of knowledge used to guide the training of the student is important. Previous distillation methods in semantic segmentation strive to extract various forms of knowledge from the features, which involve elaborate manual design relying on prior information and have limited performance gains. In this paper, we propose a simple yet effective feature distillation method called normalized feature distillation (NFD), aiming to enable effective distillation with the original features without the need to manually design new forms of knowledge. The key idea is to prevent the student from focusing on imitating the magnitude of the teacher's feature response by normalization. Our method achieves state-of-the-art distillation results for semantic segmentation on Cityscapes, VOC 2012, and ADE20K datasets. Code will be available.
We tackle the essential task of finding dense visual correspondences between a pair of images. This is a challenging problem due to various factors such as poor texture, repetitive patterns, illumination variation, and motion blur in practical scenarios. In contrast to methods that use dense correspondence ground-truths as direct supervision for local feature matching training, we train 3DG-STFM: a multi-modal matching model (Teacher) to enforce the depth consistency under 3D dense correspondence supervision and transfer the knowledge to 2D unimodal matching model (Student). Both teacher and student models consist of two transformer-based matching modules that obtain dense correspondences in a coarse-to-fine manner. The teacher model guides the student model to learn RGB-induced depth information for the matching purpose on both coarse and fine branches. We also evaluate 3DG-STFM on a model compression task. To the best of our knowledge, 3DG-STFM is the first student-teacher learning method for the local feature matching task. The experiments show that our method outperforms state-of-the-art methods on indoor and outdoor camera pose estimations, and homography estimation problems. Code is available at: https://github.com/Ryan-prime/3DG-STFM.
Unmanned aerial vehicle (UAV) tracking has wide potential applications in such as agriculture, navigation, and public security. However, the limitations of computing resources, battery capacity, and maximum load of UAV hinder the deployment of deep learning-based tracking algorithms on UAV. Consequently, discriminative correlation filters (DCF) trackers stand out in the UAV tracking community because of their high efficiency. However, their precision is usually much lower than trackers based on deep learning. Model compression is a promising way to narrow the gap (i.e., effciency, precision) between DCF- and deep learning- based trackers, which has not caught much attention in UAV tracking. In this paper, we propose the P-SiamFC++ tracker, which is the first to use rank-based filter pruning to compress the SiamFC++ model, achieving a remarkable balance between efficiency and precision. Our method is general and may encourage further studies on UAV tracking with model compression. Extensive experiments on four UAV benchmarks, including UAV123@10fps, DTB70, UAVDT and Vistrone2018, show that P-SiamFC++ tracker significantly outperforms state-of-the-art UAV tracking methods.
Mobile devices run deep learning models for various purposes, such as image classification and speech recognition. Due to the resource constraints of mobile devices, researchers have focused on either making a lightweight deep neural network (DNN) model using model pruning or generating an efficient code using compiler optimization. Surprisingly, we found that the straightforward integration between model compression and compiler auto-tuning often does not produce the most efficient model for a target device. We propose CPrune, a compiler-informed model pruning for efficient target-aware DNN execution to support an application with a required target accuracy. CPrune makes a lightweight DNN model through informed pruning based on the structural information of subgraphs built during the compiler tuning process. Our experimental results show that CPrune increases the DNN execution speed up to 2.73x compared to the state-of-the-art TVM auto-tune while satisfying the accuracy requirement.
Model compression, such as pruning and quantization, has been widely applied to optimize neural networks on resource-limited classical devices. Recently, there are growing interest in variational quantum circuits (VQC), that is, a type of neural network on quantum computers (a.k.a., quantum neural networks). It is well known that the near-term quantum devices have high noise and limited resources (i.e., quantum bits, qubits); yet, how to compress quantum neural networks has not been thoroughly studied. One might think it is straightforward to apply the classical compression techniques to quantum scenarios. However, this paper reveals that there exist differences between the compression of quantum and classical neural networks. Based on our observations, we claim that the compilation/traspilation has to be involved in the compression process. On top of this, we propose the very first systematical framework, namely CompVQC, to compress quantum neural networks (QNNs).In CompVQC, the key component is a novel compression algorithm, which is based on the alternating direction method of multipliers (ADMM) approach. Experiments demonstrate the advantage of the CompVQC, reducing the circuit depth (almost over 2.5 %) with a negligible accuracy drop (<1%), which outperforms other competitors. Another promising truth is our CompVQC can indeed promote the robustness of the QNN on the near-term noisy quantum devices.
Factorizing a large matrix into small matrices is a popular strategy for model compression. Singular value decomposition (SVD) plays a vital role in this compression strategy, approximating a learned matrix with fewer parameters. However, SVD minimizes the squared error toward reconstructing the original matrix without gauging the importance of the parameters, potentially giving a larger reconstruction error for those who affect the task accuracy more. In other words, the optimization objective of SVD is not aligned with the trained model's task accuracy. We analyze this previously unexplored problem, make observations, and address it by introducing Fisher information to weigh the importance of parameters affecting the model prediction. This idea leads to our method: Fisher-Weighted SVD (FWSVD). Although the factorized matrices from our approach do not result in smaller reconstruction errors, we find that our resulting task accuracy is much closer to the original model's performance. We perform analysis with the transformer-based language models, showing our weighted SVD largely alleviates the mismatched optimization objectives and can maintain model performance with a higher compression rate. Our method can directly compress a task-specific model while achieving better performance than other compact model strategies requiring expensive model pre-training. Moreover, the evaluation of compressing an already compact model shows our method can further reduce 9% to 30% parameters with an insignificant impact on task accuracy.
As the machine learning and systems communities strive to achieve higher energy-efficiency through custom deep neural network (DNN) accelerators, varied precision or quantization levels, and model compression techniques, there is a need for design space exploration frameworks that incorporate quantization-aware processing elements into the accelerator design space while having accurate and fast power, performance, and area models. In this work, we present QUIDAM, a highly parameterized quantization-aware DNN accelerator and model co-exploration framework. Our framework can facilitate future research on design space exploration of DNN accelerators for various design choices such as bit precision, processing element type, scratchpad sizes of processing elements, global buffer size, number of total processing elements, and DNN configurations. Our results show that different bit precisions and processing element types lead to significant differences in terms of performance per area and energy. Specifically, our framework identifies a wide range of design points where performance per area and energy varies more than 5x and 35x, respectively. With the proposed framework, we show that lightweight processing elements achieve on par accuracy results and up to 5.7x more performance per area and energy improvement when compared to the best INT16 based implementation. Finally, due to the efficiency of the pre-characterized power, performance, and area models, QUIDAM can speed up the design exploration process by 3-4 orders of magnitude as it removes the need for expensive synthesis and characterization of each design.
One of the main focuses in distributed learning is communication efficiency, since model aggregation at each round of training can consist of millions to billions of parameters. Several model compression methods, such as gradient quantization and sparsification, have been proposed to improve the communication efficiency of model aggregation. However, the information-theoretic minimum communication cost for a given distortion of gradient estimators is still unknown. In this paper, we study the fundamental limit of communication cost of model aggregation in distributed learning from a rate-distortion perspective. By formulating the model aggregation as a vector Gaussian CEO problem, we derive the rate region bound and sum-rate-distortion function for the model aggregation problem, which reveals the minimum communication rate at a particular gradient distortion upper bound. We also analyze the communication cost at each iteration and total communication cost based on the sum-rate-distortion function with the gradient statistics of real-world datasets. It is found that the communication gain by exploiting the correlation between worker nodes is significant for SignSGD, and a high distortion of gradient estimator can achieve low total communication cost in gradient compression.
This technical report describes the details of our TASK1A submission of the DCASE2021 challenge. The goal of the task is to design an audio scene classification system for device-imbalanced datasets under the constraints of model complexity. This report introduces four methods to achieve the goal. First, we propose Residual Normalization, a novel feature normalization method that uses instance normalization with a shortcut path to discard unnecessary device-specific information without losing useful information for classification. Second, we design an efficient architecture, BC-ResNet-Mod, a modified version of the baseline architecture with a limited receptive field. Third, we exploit spectrogram-to-spectrogram translation from one to multiple devices to augment training data. Finally, we utilize three model compression schemes: pruning, quantization, and knowledge distillation to reduce model complexity. The proposed system achieves an average test accuracy of 76.3% in TAU Urban Acoustic Scenes 2020 Mobile, development dataset with 315k parameters, and average test accuracy of 75.3% after compression to 61.0KB of non-zero parameters. We extend this work to [1].
With the improvement of AI chips (e.g., GPU, TPU, and NPU) and the fast development of the Internet of Things (IoT), some robust deep neural networks (DNNs) are usually composed of millions or even hundreds of millions of parameters. Such a large model may not be suitable for directly deploying on low computation and low capacity units (e.g., edge devices). Knowledge distillation (KD) has recently been recognized as a powerful model compression method to decrease the model parameters effectively. The central concept of KD is to extract useful information from the feature maps of a large model (i.e., teacher model) as a reference to successfully train a small model (i.e., student model) in which the model size is much smaller than the teacher one. Although many KD methods have been proposed to utilize the information from the feature maps of intermediate layers in the teacher model, most did not consider the similarity of feature maps between the teacher model and the student model. As a result, it may make the student model learn useless information. Inspired by the attention mechanism, we propose a novel KD method called representative teacher key (RTK) that not only considers the similarity of feature maps but also filters out the useless information to improve the performance of the target student model. In the experiments, we validate our proposed method with several backbone networks (e.g., ResNet and WideResNet) and datasets (e.g., CIFAR10, CIFAR100, SVHN, and CINIC10). The results show that our proposed RTK can effectively improve the classification accuracy of the state-of-the-art attention-based KD method.
Deep convolutional neural network with increased number of parameters has achieved improved precision in task of object detection on natural images, where objects of interests are annotated with horizontal boundary boxes. On aerial images captured from the bird-view perspective, these improvements on model architecture and deeper convolutional layers can also boost the performance on oriented object detection task. However, it is hard to directly apply those state-of-the-art object detectors on the devices with limited computation resources, which necessitates lightweight models through model compression. In order to address this issue, we present a model compression method for rotated object detection on aerial images by knowledge distillation, namely KD-RNet. With a well-trained teacher oriented object detector with a large number of parameters, the obtained object category and location information are both transferred to a compact student network in KD-RNet by collaborative training strategy. Transferring the category information is achieved by knowledge distillation on predicted probability distribution, and a soft regression loss is adopted for handling displacement in location information transfer. The experimental result on a large-scale aerial object detection dataset (DOTA) demonstrates that the proposed KD-RNet model can achieve improved mean-average precision (mAP) with reduced number of parameters, at the same time, KD-RNet boost the performance on providing high quality detections with higher overlap with groundtruth annotations.