The continued need for improvements in accuracy, throughput, and efficiency of Deep Neural Networks has resulted in a multitude of methods that make the most of custom architectures on FPGAs. These include the creation of hand-crafted networks and the use of quantization and pruning to reduce extraneous network parameters. However, with the potential of static solutions already well exploited, we propose to shift the focus to using the varying difficulty of individual data samples to further improve efficiency and reduce average compute for classification. Input-dependent computation allows for the network to make runtime decisions to finish a task early if the result meets a confidence threshold. Early-Exit network architectures have become an increasingly popular way to implement such behaviour in software. We create: A Toolflow for Hardware Early-Exit Network Automation (ATHEENA), an automated FPGA toolflow that leverages the probability of samples exiting early from such networks to scale the resources allocated to different sections of the network. The toolflow uses the data-flow model of fpgaConvNet, extended to support Early-Exit networks as well as Design Space Exploration to optimize the generated streaming architecture hardware with the goal of increasing throughput/reducing area while maintaining accuracy. Experimental results on three different networks demonstrate a throughput increase of $2.00\times$ to $2.78\times$ compared to an optimized baseline network implementation with no early exits. Additionally, the toolflow can achieve a throughput matching the same baseline with as low as $46\%$ of the resources the baseline requires.
Human-Object Interaction (HOI) detection, which localizes and infers relationships between human and objects, plays an important role in scene understanding. Although two-stage HOI detectors have advantages of high efficiency in training and inference, they suffer from lower performance than one-stage methods due to the old backbone networks and the lack of considerations for the HOI perception process of humans in the interaction classifiers. In this paper, we propose Vision Transformer based Pose-Conditioned Self-Loop Graph (ViPLO) to resolve these problems. First, we propose a novel feature extraction method suitable for the Vision Transformer backbone, called masking with overlapped area (MOA) module. The MOA module utilizes the overlapped area between each patch and the given region in the attention function, which addresses the quantization problem when using the Vision Transformer backbone. In addition, we design a graph with a pose-conditioned self-loop structure, which updates the human node encoding with local features of human joints. This allows the classifier to focus on specific human joints to effectively identify the type of interaction, which is motivated by the human perception process for HOI. As a result, ViPLO achieves the state-of-the-art results on two public benchmarks, especially obtaining a +2.07 mAP performance gain on the HICO-DET dataset. The source codes are available at https://github.com/Jeeseung-Park/ViPLO.
We develop an end-to-end workflow for the training and implementation of co-designed neural networks (NNs) for efficient field-programmable gate array (FPGA) and application-specific integrated circuit (ASIC) hardware. Our approach leverages Hessian-aware quantization (HAWQ) of NNs, the Quantized Open Neural Network Exchange (QONNX) intermediate representation, and the hls4ml tool flow for transpiling NNs into FPGA and ASIC firmware. This makes efficient NN implementations in hardware accessible to nonexperts, in a single open-sourced workflow that can be deployed for real-time machine learning applications in a wide range of scientific and industrial settings. We demonstrate the workflow in a particle physics application involving trigger decisions that must operate at the 40 MHz collision rate of the CERN Large Hadron Collider (LHC). Given the high collision rate, all data processing must be implemented on custom ASIC and FPGA hardware within a strict area and latency. Based on these constraints, we implement an optimized mixed-precision NN classifier for high-momentum particle jets in simulated LHC proton-proton collisions.
In this paper, we study the Tsallis entropic regularized optimal transport in the continuous setting and establish fundamental results such as the $\Gamma$-convergence of the Tsallis regularized optimal transport to the Monge--Kantorovich problem as the regularization parameter tends to zero. In addition, using the quantization and shadow arguments developed by Eckstein--Nutz, we derive the convergence rate of the Tsallis entropic regularization and provide explicit constants. Furthermore, we compare these results with the well-known case of the Kullback--Leibler (KL) divergence regularization and show that the KL regularization achieves the fastest convergence rate in the Tsallis framework.
This paper considers distributed optimization algorithms, with application in binary classification via distributed support-vector-machines (D-SVM) over multi-agent networks subject to some link nonlinearities. The agents solve a consensus-constraint distributed optimization cooperatively via continuous-time dynamics, while the links are subject to strongly sign-preserving odd nonlinear conditions. Logarithmic quantization and clipping (saturation) are two examples of such nonlinearities. In contrast to existing literature that mostly considers ideal links and perfect information exchange over linear channels, we show how general sector-bounded models affect the convergence to the optimizer (i.e., the SVM classifier) over dynamic balanced directed networks. In general, any odd sector-bounded nonlinear mapping can be applied to our dynamics. The main challenge is to show that the proposed system dynamics always have one zero eigenvalue (associated with the consensus) and the other eigenvalues all have negative real parts. This is done by recalling arguments from matrix perturbation theory. Then, the solution is shown to converge to the agreement state under certain conditions. For example, the gradient tracking (GT) step size is tighter than the linear case by factors related to the upper/lower sector bounds. To the best of our knowledge, no existing work in distributed optimization and learning literature considers non-ideal link conditions.
With the development of high-definition display devices, the practical scenario of Super-Resolution (SR) usually needs to super-resolve large input like 2K to higher resolution (4K/8K). To reduce the computational and memory cost, current methods first split the large input into local patches and then merge the SR patches into the output. These methods adaptively allocate a subnet for each patch. Quantization is a very important technique for network acceleration and has been used to design the subnets. Current methods train an MLP bit selector to determine the propoer bit for each layer. However, they uniformly sample subnets for training, making simple subnets overfitted and complicated subnets underfitted. Therefore, the trained bit selector fails to determine the optimal bit. Apart from this, the introduced bit selector brings additional cost to each layer of the SR network. In this paper, we propose a novel method named Content-Aware Bit Mapping (CABM), which can remove the bit selector without any performance loss. CABM also learns a bit selector for each layer during training. After training, we analyze the relation between the edge information of an input patch and the bit of each layer. We observe that the edge information can be an effective metric for the selected bit. Therefore, we design a strategy to build an Edge-to-Bit lookup table that maps the edge score of a patch to the bit of each layer during inference. The bit configuration of SR network can be determined by the lookup tables of all layers. Our strategy can find better bit configuration, resulting in more efficient mixed precision networks. We conduct detailed experiments to demonstrate the generalization ability of our method. The code will be released.
In this paper, we propose an ultrafast automated model compression framework called SeerNet for flexible network deployment. Conventional non-differen-tiable methods discretely search the desirable compression policy based on the accuracy from exhaustively trained lightweight models, and existing differentiable methods optimize an extremely large supernet to obtain the required compressed model for deployment. They both cause heavy computational cost due to the complex compression policy search and evaluation process. On the contrary, we obtain the optimal efficient networks by directly optimizing the compression policy with an accurate performance predictor, where the ultrafast automated model compression for various computational cost constraint is achieved without complex compression policy search and evaluation. Specifically, we first train the performance predictor based on the accuracy from uncertain compression policies actively selected by efficient evolutionary search, so that informative supervision is provided to learn the accurate performance predictor with acceptable cost. Then we leverage the gradient that maximizes the predicted performance under the barrier complexity constraint for ultrafast acquisition of the desirable compression policy, where adaptive update stepsizes with momentum are employed to enhance optimality of the acquired pruning and quantization strategy. Compared with the state-of-the-art automated model compression methods, experimental results on image classification and object detection show that our method achieves competitive accuracy-complexity trade-offs with significant reduction of the search cost.
Nowadays, data is represented by vectors. Retrieving those vectors, among millions and billions, that are similar to a given query is a ubiquitous problem, known as similarity search, of relevance for a wide range of applications. Graph-based indices are currently the best performing techniques for billion-scale similarity search. However, their random-access memory pattern presents challenges to realize their full potential. In this work, we present new techniques and systems for creating faster and smaller graph-based indices. To this end, we introduce a novel vector compression method, Locally-adaptive Vector Quantization (LVQ), that uses per-vector scaling and scalar quantization to improve search performance with fast similarity computations and a reduced effective bandwidth, while decreasing memory footprint and barely impacting accuracy. LVQ, when combined with a new high-performance computing system for graph-based similarity search, establishes the new state of the art in terms of performance and memory footprint. For billions of vectors, LVQ outcompetes the second-best alternatives: (1) in the low-memory regime, by up to 20.7x in throughput with up to a 3x memory footprint reduction, and (2) in the high-throughput regime by 5.8x with 1.4x less memory.
In this paper, we show that a binary latent space can be explored for compact yet expressive image representations. We model the bi-directional mappings between an image and the corresponding latent binary representation by training an auto-encoder with a Bernoulli encoding distribution. On the one hand, the binary latent space provides a compact discrete image representation of which the distribution can be modeled more efficiently than pixels or continuous latent representations. On the other hand, we now represent each image patch as a binary vector instead of an index of a learned cookbook as in discrete image representations with vector quantization. In this way, we obtain binary latent representations that allow for better image quality and high-resolution image representations without any multi-stage hierarchy in the latent space. In this binary latent space, images can now be generated effectively using a binary latent diffusion model tailored specifically for modeling the prior over the binary image representations. We present both conditional and unconditional image generation experiments with multiple datasets, and show that the proposed method performs comparably to state-of-the-art methods while dramatically improving the sampling efficiency to as few as 16 steps without using any test-time acceleration. The proposed framework can also be seamlessly scaled to $1024 \times 1024$ high-resolution image generation without resorting to latent hierarchy or multi-stage refinements.
Transformers' compute-intensive operations pose enormous challenges for their deployment in resource-constrained EdgeAI / tinyML devices. As an established neural network compression technique, quantization reduces the hardware computational and memory resources. In particular, fixed-point quantization is desirable to ease the computations using lightweight blocks, like adders and multipliers, of the underlying hardware. However, deploying fully-quantized Transformers on existing general-purpose hardware, generic AI accelerators, or specialized architectures for Transformers with floating-point units might be infeasible and/or inefficient. Towards this, we propose SwiftTron, an efficient specialized hardware accelerator designed for Quantized Transformers. SwiftTron supports the execution of different types of Transformers' operations (like Attention, Softmax, GELU, and Layer Normalization) and accounts for diverse scaling factors to perform correct computations. We synthesize the complete SwiftTron architecture in a $65$ nm CMOS technology with the ASIC design flow. Our Accelerator executes the RoBERTa-base model in 1.83 ns, while consuming 33.64 mW power, and occupying an area of 273 mm^2. To ease the reproducibility, the RTL of our SwiftTron architecture is released at https://github.com/albertomarchisio/SwiftTron.
The vulnerability in the algorithm supply chain of deep learning has imposed new challenges to image retrieval systems in the downstream. Among a variety of techniques, deep hashing is gaining popularity. As it inherits the algorithmic backend from deep learning, a handful of attacks are recently proposed to disrupt normal image retrieval. Unfortunately, the defense strategies in softmax classification are not readily available to be applied in the image retrieval domain. In this paper, we propose an efficient and unsupervised scheme to identify unique adversarial behaviors in the hamming space. In particular, we design three criteria from the perspectives of hamming distance, quantization loss and denoising to defend against both untargeted and targeted attacks, which collectively limit the adversarial space. The extensive experiments on four datasets demonstrate 2-23% improvements of detection rates with minimum computational overhead for real-time image queries.
Recently, self-supervised learning (SSL) has achieved tremendous success in learning image representation. Despite the empirical success, most self-supervised learning methods are rather "inefficient" learners, typically taking hundreds of training epochs to fully converge. In this work, we show that the key towards efficient self-supervised learning is to increase the number of crops from each image instance. Leveraging one of the state-of-the-art SSL method, we introduce a simplistic form of self-supervised learning method called Extreme-Multi-Patch Self-Supervised-Learning (EMP-SSL) that does not rely on many heuristic techniques for SSL such as weight sharing between the branches, feature-wise normalization, output quantization, and stop gradient, etc, and reduces the training epochs by two orders of magnitude. We show that the proposed method is able to converge to 85.1% on CIFAR-10, 58.5% on CIFAR-100, 38.1% on Tiny ImageNet and 58.5% on ImageNet-100 in just one epoch. Furthermore, the proposed method achieves 91.5% on CIFAR-10, 70.1% on CIFAR-100, 51.5% on Tiny ImageNet and 78.9% on ImageNet-100 with linear probing in less than ten training epochs. In addition, we show that EMP-SSL shows significantly better transferability to out-of-domain datasets compared to baseline SSL methods. We will release the code in https://github.com/tsb0601/EMP-SSL.
Quantization has emerged as an essential technique for deploying deep neural networks (DNNs) on devices with limited resources. However, quantized models exhibit vulnerabilities when exposed to various noises in real-world applications. Despite the importance of evaluating the impact of quantization on robustness, existing research on this topic is limited and often disregards established principles of robustness evaluation, resulting in incomplete and inconclusive findings. To address this gap, we thoroughly evaluated the robustness of quantized models against various noises (adversarial attacks, natural corruptions, and systematic noises) on ImageNet. Extensive experiments demonstrate that lower-bit quantization is more resilient to adversarial attacks but is more susceptible to natural corruptions and systematic noises. Notably, our investigation reveals that impulse noise (in natural corruptions) and the nearest neighbor interpolation (in systematic noises) have the most significant impact on quantized models. Our research contributes to advancing the robust quantization of models and their deployment in real-world scenarios.
With the advent of general-purpose speech representations from large-scale self-supervised models, applying a single model to multiple downstream tasks is becoming a de-facto approach. However, the pooling problem remains; the length of speech representations is inherently variable. The naive average pooling is often used, even though it ignores the characteristics of speech, such as differently lengthed phonemes. Hence, we design a novel pooling method to squash acoustically similar representations via vector quantization, which does not require additional training, unlike attention-based pooling. Further, we evaluate various unsupervised pooling methods on various self-supervised models. We gather diverse methods scattered around speech and text to evaluate on various tasks: keyword spotting, speaker identification, intent classification, and emotion recognition. Finally, we quantitatively and qualitatively analyze our method, comparing it with supervised pooling methods.
Exploring the expected quantizing scheme with suitable mixed-precision policy is the key point to compress deep neural networks (DNNs) in high efficiency and accuracy. This exploration implies heavy workloads for domain experts, and an automatic compression method is needed. However, the huge search space of the automatic method introduces plenty of computing budgets that make the automatic process challenging to be applied in real scenarios. In this paper, we propose an end-to-end framework named AutoQNN, for automatically quantizing different layers utilizing different schemes and bitwidths without any human labor. AutoQNN can seek desirable quantizing schemes and mixed-precision policies for mainstream DNN models efficiently by involving three techniques: quantizing scheme search (QSS), quantizing precision learning (QPL), and quantized architecture generation (QAG). QSS introduces five quantizing schemes and defines three new schemes as a candidate set for scheme search, and then uses the differentiable neural architecture search (DNAS) algorithm to seek the layer- or model-desired scheme from the set. QPL is the first method to learn mixed-precision policies by reparameterizing the bitwidths of quantizing schemes, to the best of our knowledge. QPL optimizes both classification loss and precision loss of DNNs efficiently and obtains the relatively optimal mixed-precision model within limited model size and memory footprint. QAG is designed to convert arbitrary architectures into corresponding quantized ones without manual intervention, to facilitate end-to-end neural network quantization. We have implemented AutoQNN and integrated it into Keras. Extensive experiments demonstrate that AutoQNN can consistently outperform state-of-the-art quantization.
With the growing concern about the security and privacy of smart grid systems, cyberattacks on critical power grid components, such as state estimation, have proven to be one of the top-priority cyber-related issues and have received significant attention in recent years. However, cyberattack detection in smart grids now faces new challenges, including privacy preservation and decentralized power zones with strategic data owners. To address these technical bottlenecks, this paper proposes a novel Federated Learning-based privacy-preserving and communication-efficient attack detection framework, known as FedDiSC, that enables Discrimination between power System disturbances and Cyberattacks. Specifically, we first propose a Federated Learning approach to enable Supervisory Control and Data Acquisition subsystems of decentralized power grid zones to collaboratively train an attack detection model without sharing sensitive power related data. Secondly, we put forward a representation learning-based Deep Auto-Encoder network to accurately detect power system and cybersecurity anomalies. Lastly, to adapt our proposed framework to the timeliness of real-world cyberattack detection in SGs, we leverage the use of a gradient privacy-preserving quantization scheme known as DP-SIGNSGD to improve its communication efficiency. Extensive simulations of the proposed framework on publicly available Industrial Control Systems datasets demonstrate that the proposed framework can achieve superior detection accuracy while preserving the privacy of sensitive power grid related information. Furthermore, we find that the gradient quantization scheme utilized improves communication efficiency by 40% when compared to a traditional federated learning approach without gradient quantization which suggests suitability in a real-world scenario.
Learning-based video compression has been extensively studied over the past years, but it still has limitations in adapting to various motion patterns and entropy models. In this paper, we propose multi-mode video compression (MMVC), a block wise mode ensemble deep video compression framework that selects the optimal mode for feature domain prediction adapting to different motion patterns. Proposed multi-modes include ConvLSTM-based feature domain prediction, optical flow conditioned feature domain prediction, and feature propagation to address a wide range of cases from static scenes without apparent motions to dynamic scenes with a moving camera. We partition the feature space into blocks for temporal prediction in spatial block-based representations. For entropy coding, we consider both dense and sparse post-quantization residual blocks, and apply optional run-length coding to sparse residuals to improve the compression rate. In this sense, our method uses a dual-mode entropy coding scheme guided by a binary density map, which offers significant rate reduction surpassing the extra cost of transmitting the binary selection map. We validate our scheme with some of the most popular benchmarking datasets. Compared with state-of-the-art video compression schemes and standard codecs, our method yields better or competitive results measured with PSNR and MS-SSIM.
The recent advances in machine learning and deep neural networks have made them attractive candidates for wireless communications functions such as channel estimation, decoding, and downlink channel state information (CSI) compression. However, most of these neural networks are large and inefficient making it a barrier for deployment in practical wireless systems that require low-latency and low memory footprints for individual network functions. To mitigate these limitations, we propose accelerated and compressed efficient neural networks for massive MIMO CSI feedback. Specifically, we have thoroughly investigated the adoption of network pruning, post-training dynamic range quantization, and weight clustering to optimize CSI feedback compression for massive MIMO systems. Furthermore, we have deployed the proposed model compression techniques on commodity hardware and demonstrated that in order to achieve inference gains, specialized libraries that accelerate computations for sparse neural networks are required. Our findings indicate that there is remarkable value in applying these model compression techniques and the proposed joint pruning and quantization approach reduced model size by 86.5% and inference time by 76.2% with minimal impact to model accuracy. These compression methods are crucial to pave the way for practical adoption and deployments of deep learning-based techniques in commercial wireless systems.
Transformer-based models, exemplified by GPT-3, ChatGPT, and GPT-4, have recently garnered considerable attention in both academia and industry due to their promising performance in general language tasks. Nevertheless, these models typically involve computationally encoding processes, and in some cases, decoding processes as well, both of which are fundamentally large-scale matrix multiplication. These operations bring the inevitable challenges of massive computation resources and huge memory footprint, usually requiring at least 10^23 FLOPs and hundreds of gigabytes, respectively. A common method to address this issue is to reduce the computational and memory requirements by applying layerwise quantization to the transformer, replacing the usual fp32 data type with a low-bit equivalent. Unfortunately, this method often leads to decreased model accuracy and necessitates time-consuming retraining. Such retraining not only requires fine-tuning skills but also substantial computational resources, posing challenges for users. To specifically tackle these issues, we propose BCT, a framework of blockwise compression for transformers without retraining, aiming to facilitate model deployment. Unlike layerwise compression methods, BCT achieves finer compression of the entire transformer by operating blockwise. This method mitigates data distribution deviation caused by quantization, eliminating the requirement for retraining. BCT effectively compresses all components of the model, including but not limited to the embedding, matrix multiplication, GELU, Softmax, layer normalization, and intermediate results. In a case study, an efficient model is compressed by BCT achieving up to 7.988x compression. Subsequently, we also evaluate it on several General Language Understanding Evaluation (GLUE) datasets.
We propose two distributed iterative algorithms that can be used to solve, in finite time, the distributed optimization problem over quadratic local cost functions in large-scale networks. The first algorithm exhibits synchronous operation whereas the second one exhibits asynchronous operation. Both algorithms share salient features. Specifically, the algorithms operate exclusively with quantized values, which means that the information stored, processed and exchanged between neighboring nodes is subject to deterministic uniform quantization. The algorithms rely on event-driven updates in order to reduce energy consumption, communication bandwidth, network congestion, and/or processor usage. Finally, once the algorithms converge, nodes distributively terminate their operation. We prove that our algorithms converge in a finite number of iterations to the exact optimal solution depending on the quantization level, and we present applications of our algorithms to (i) optimal task scheduling for data centers, and (ii) global model aggregation for distributed federated learning. We provide simulations of these applications to illustrate the operation, performance, and advantages of the proposed algorithms. Additionally, it is shown that our proposed algorithms compare favorably to algorithms in the current literature. Quantized communication and asynchronous updates increase the required time to completion, but finite-time operation is maintained.