Model predictive Path-Following Control (MPFC) is a viable option for motion systems in many application domains. However, despite considerable progress on tailored numerical methods for predictive control, the real-time implementation of predictive control and MPFC on small-scale autonomous platforms with low-cost embedded hardware remains challenging. While usual stabilizing MPC formulations lead to static feedback laws, the MPFC feedback turns out to be dynamic as the path parameter acts as an internal controller variable. In this paper, we leverage deep learning to implement predictive path-following control on microcontrollers. We show that deep neural networks can approximate the dynamic MPFC feedback law accurately. Moreover, we illustrate and tackle the challenges that arise if the target platform employs limited precision arithmetic. Specifically, we draw upon a post-stabilization with an additional feedback law to attenuate undesired quantization effects. Simulation examples underpin the efficacy of the proposed approach.
Quantization plays a critical role in digital signal processing systems, allowing the representation of continuous amplitude signals with a finite number of bits. However, accurately representing signals requires a large number of quantization bits, which causes severe cost, power consumption, and memory burden. A promising way to address this issue is task-based quantization. By exploiting the task information for the overall system design, task-based quantization can achieve satisfying performance with low quantization costs. In this work, we apply task-based quantization to multi-user signal recovery and present a hardware prototype implementation. The prototype consists of a tailored configurable combining board, and a software-based processing and demonstration system. Through experiments, we verify that with proper design, the task-based quantization achieves a reduction of 25 fold in memory by reducing from 16 receivers with 16 bits each to 2 receivers with 5 bits each, without compromising signal recovery performance.
Bayesian Optimization Mixed-Precision Neural Architecture Search (BOMP-NAS) is an approach to quantization-aware neural architecture search (QA-NAS) that leverages both Bayesian optimization (BO) and mixed-precision quantization (MP) to efficiently search for compact, high performance deep neural networks. The results show that integrating quantization-aware fine-tuning (QAFT) into the NAS loop is a necessary step to find networks that perform well under low-precision quantization: integrating it allows a model size reduction of nearly 50\% on the CIFAR-10 dataset. BOMP-NAS is able to find neural networks that achieve state of the art performance at much lower design costs. This study shows that BOMP-NAS can find these neural networks at a 6x shorter search time compared to the closest related work.
Deep neural networks (DNNs) are nowadays ubiquitous in many domains such as computer vision. However, due to their high latency, the deployment of DNNs hinges on the development of compression techniques such as quantization which consists in lowering the number of bits used to encode the weights and activations. Growing concerns for privacy and security have motivated the development of data-free techniques, at the expanse of accuracy. In this paper, we identity the uniformity of the quantization operator as a limitation of existing approaches, and propose a data-free non-uniform method. More specifically, we argue that to be readily usable without dedicated hardware and implementation, non-uniform quantization shall not change the nature of the mathematical operations performed by the DNN. This leads to search among the continuous automorphisms of $(\mathbb{R}_+^*,\times)$, which boils down to the power functions defined by their exponent. To find this parameter, we propose to optimize the reconstruction error of each layer: in particular, we show that this procedure is locally convex and admits a unique solution. At inference time, we show that our approach, dubbed PowerQuant, only require simple modifications in the quantized DNN activation functions. As such, with only negligible overhead, it significantly outperforms existing methods in a variety of configurations.
Nanopore sequencers generate electrical raw signals in real-time while sequencing long genomic strands. These raw signals can be analyzed as they are generated, providing an opportunity for real-time genome analysis. An important feature of nanopore sequencing, Read Until, can eject strands from sequencers without fully sequencing them, which provides opportunities to computationally reduce the sequencing time and cost. However, existing works utilizing Read Until either 1) require powerful computational resources that may not be available for portable sequencers or 2) lack scalability for large genomes, rendering them inaccurate or ineffective. We propose RawHash, the first mechanism that can accurately and efficiently perform real-time analysis of nanopore raw signals for large genomes using a hash-based similarity search. To enable this, RawHash ensures the signals corresponding to the same DNA content lead to the same hash value, regardless of the slight variations in these signals. RawHash achieves an accurate hash-based similarity search via an effective quantization of the raw signals such that signals corresponding to the same DNA content have the same quantized value and, subsequently, the same hash value. We evaluate RawHash on three applications: 1) read mapping, 2) relative abundance estimation, and 3) contamination analysis. Our evaluations show that RawHash is the only tool that can provide high accuracy and high throughput for analyzing large genomes in real-time. When compared to the state-of-the-art techniques, UNCALLED and Sigmap, RawHash provides 1) 25.8x and 3.4x better average throughput and 2) significantly better accuracy for large genomes, respectively. Source code is available at https://github.com/CMU-SAFARI/RawHash.
Neural-based image and video codecs are significantly more power-efficient when weights and activations are quantized to low-precision integers. While there are general-purpose techniques for reducing quantization effects, large losses can occur when specific entropy coding properties are not considered. This work analyzes how entropy coding is affected by parameter quantizations, and provides a method to minimize losses. It is shown that, by using a certain type of coding parameters to be learned, uniform quantization becomes practically optimal, also simplifying the minimization of code memory requirements. The mathematical properties of the new representation are presented, and its effectiveness is demonstrated by coding experiments, showing that good results can be obtained with precision as low as 4~bits per network output, and practically no loss with 8~bits.
This paper focuses on designing robust symbol-level precoding (SLP) in an overlay cognitive radio (CR) network, where the primary and secondary networks transmit signals concurrently. When the primary base station (PBS) shares data and perfect channel state information (CSI) with the cognitive base station (CBS), we derive an SLP approach that minimizes the CR transmission power and satisfies symbol-wise Safety Margin (SM) constraints of both primary users (PUs) and cognitive users (CUs). The resulting optimization has a quadratic objective and linear inequality (LI) constraints, which can be solved by standard convex methods. For the case of imperfect CSI from the PBS, we propose robust SLP schemes. First, with a norm-bounded CSI error model to approximate the uncertain channels, we adopt a max-min philosophy to conservatively achieve robust SLP constraints. Second, we use the additive quantization noise model (AQNM) to describe the quantized PBS CSI and employ a stochastic constraint to formulate the problem. Both robust approaches also result in a quadratic objective with LI constraints. Simulation results show that, rather than simply trying to eliminate the network's cross-interference, the proposed robust SLP schemes enable the primary and secondary networks to aid each other in meeting their quality of service constraints.
Today, visual data is often analyzed by a neural network without any human being involved, which demands for specialized codecs. For standard-compliant codec adaptations towards certain information sinks, HEVC or VVC provide the possibility of frequency-specific quantization with scaling lists. This is a well-known method for the human visual system, where scaling lists are derived from psycho-visual models. In this work, we employ scaling lists when performing VVC intra coding for neural networks as information sink. To this end, we propose a novel data-driven method to obtain optimal scaling lists for arbitrary neural networks. Experiments with Mask R-CNN as information sink reveal that coding the Cityscapes dataset with the proposed scaling lists result in peak bitrate savings of 8.9 % over VVC with constant quantization. By that, our approach also outperforms scaling lists optimized for the human visual system. The generated scaling lists can be found under https://github.com/FAU-LMS/VCM_scaling_lists.
Ensemble learning has gain attention in resent deep learning research as a way to further boost the accuracy and generalizability of deep neural network (DNN) models. Recent ensemble training method explores different training algorithms or settings on multiple sub-models with the same model architecture, which lead to significant burden on memory and computation cost of the ensemble model. Meanwhile, the heurtsically induced diversity may not lead to significant performance gain. We propose a new prespective on exploring the intrinsic diversity within a model architecture to build efficient DNN ensemble. We make an intriguing observation that pruning and quantization, while both leading to efficient model architecture at the cost of small accuracy drop, leads to distinct behavior in the decision boundary. To this end, we propose Heterogeneously Compressed Ensemble (HCE), where we build an efficient ensemble with the pruned and quantized variants from a pretrained DNN model. An diversity-aware training objective is proposed to further boost the performance of the HCE ensemble. Experiemnt result shows that HCE achieves significant improvement in the efficiency-accuracy tradeoff comparing to both traditional DNN ensemble training methods and previous model compression methods.
In this research work, we have proposed a thermal tiny-YOLO multi-class object detection (TTYMOD) system as a smart forward sensing system that should remain effective in all weather and harsh environmental conditions using an end-to-end YOLO deep learning framework. It provides enhanced safety and improved awareness features for driver assistance. The system is trained on large-scale thermal public datasets as well as newly gathered novel open-sourced dataset comprising of more than 35,000 distinct thermal frames. For optimal training and convergence of YOLO-v5 tiny network variant on thermal data, we have employed different optimizers which include stochastic decent gradient (SGD), Adam, and its variant AdamW which has an improved implementation of weight decay. The performance of thermally tuned tiny architecture is further evaluated on the public as well as locally gathered test data in diversified and challenging weather and environmental conditions. The efficacy of a thermally tuned nano network is quantified using various qualitative metrics which include mean average precision, frames per second rate, and average inference time. Experimental outcomes show that the network achieved the best mAP of 56.4% with an average inference time/ frame of 4 milliseconds. The study further incorporates optimization of tiny network variant using the TensorFlow Lite quantization tool this is beneficial for the deployment of deep learning architectures on the edge and mobile devices. For this study, we have used a raspberry pi 4 computing board for evaluating the real-time feasibility performance of an optimized version of the thermal object detection network for the automotive sensor suite. The source code, trained and optimized models and complete validation/ testing results are publicly available at https://github.com/MAli-Farooq/Thermal-YOLO-And-Model-Optimization-Using-TensorFlowLite.
Data-free quantization aims to achieve model quantization without accessing any authentic sample. It is significant in an application-oriented context involving data privacy. Converting noise vectors into synthetic samples through a generator is a popular data-free quantization method, which is called generative data-free quantization. However, there is a difference in attention between synthetic samples and authentic samples. This is always ignored and restricts the quantization performance. First, since synthetic samples of the same class are prone to have homogenous attention, the quantized network can only learn limited modes of attention. Second, synthetic samples in eval mode and training mode exhibit different attention. Hence, the batch-normalization statistics matching tends to be inaccurate. ACQ is proposed in this paper to fix the attention of synthetic samples. An attention center position-condition generator is established regarding the homogenization of intra-class attention. Restricted by the attention center matching loss, the attention center position is treated as the generator's condition input to guide synthetic samples in obtaining diverse attention. Moreover, we design adversarial loss of paired synthetic samples under the same condition to prevent the generator from paying overmuch attention to the condition, which may result in mode collapse. To improve the attention similarity of synthetic samples in different network modes, we introduce a consistency penalty to guarantee accurate BN statistics matching. The experimental results demonstrate that ACQ effectively improves the attention problems of synthetic samples. Under various training settings, ACQ achieves the best quantization performance. For the 4-bit quantization of Resnet18 and Resnet50, ACQ reaches 67.55% and 72.23% accuracy, respectively.
Generative models with discrete latent representations have recently demonstrated an impressive ability to learn complex high-dimensional data distributions. However, their performance relies on a long sequence of tokens per instance and a large number of codebook entries, resulting in long sampling times and considerable computation to fit the categorical posterior. To address these issues, we propose the Masked Vector Quantization (MVQ) framework which increases the representational capacity of each code vector by learning mask configurations via a stochastic winner-takes-all training regime called Multiple Hypothese Dropout (MH-Dropout). On ImageNet 64$\times$64, MVQ reduces FID in existing vector quantization architectures by up to $68\%$ at 2 tokens per instance and $57\%$ at 5 tokens. These improvements widen as codebook entries is reduced and allows for $7\textit{--}45\times$ speed-up in token sampling during inference. As an additional benefit, we find that smaller latent spaces lead to MVQ identifying transferable visual representations where multiple can be smoothly combined.
In recent years, Convolutional Neural Networks (CNNs) have become the standard class of deep neural network for image processing, classification and segmentation tasks. However, the large strides in accuracy obtained by CNNs have been derived from increasing the complexity of network topologies, which incurs sizeable performance and energy penalties in the training and inference of CNNs. Many recent works have validated the effectiveness of parameter quantization, which consists in reducing the bit width of the network's parameters, to enable the attainment of considerable performance and energy efficiency gains without significantly compromising accuracy. However, it is difficult to compare the relative effectiveness of different quantization methods. To address this problem, we introduce RedBit, an open-source framework that provides a transparent, extensible and easy-to-use interface to evaluate the effectiveness of different algorithms and parameter configurations on network accuracy. We use RedBit to perform a comprehensive survey of five state-of-the-art quantization methods applied to the MNIST, CIFAR-10 and ImageNet datasets. We evaluate a total of 2300 individual bit width combinations, independently tuning the width of the network's weight and input activation parameters, from 32 bits down to 1 bit (e.g., 8/8, 2/2, 1/32, 1/1, for weights/activations). Upwards of 20000 hours of computing time in a pool of state-of-the-art GPUs were used to generate all the results in this paper. For 1-bit quantization, the accuracy losses for the MNIST, CIFAR-10 and ImageNet datasets range between [0.26%, 0.79%], [9.74%, 32.96%] and [10.86%, 47.36%] top-1, respectively. We actively encourage the reader to download the source code and experiment with RedBit, and to submit their own observed results to our public repository, available at https://github.com/IT-Coimbra/RedBit.
Quantization of a probability measure means representing it with a finite set of Dirac masses that approximates the input distribution well enough (in some metric space of probability measures). Various methods exists to do so, but the situation of quantizing a conditional law has been less explored. We propose a method, called DCMQ, involving a Huber-energy kernel-based approach coupled with a deep neural network architecture. The method is tested on several examples and obtains promising results.
The coordination of robotic swarms and the remote wireless control of industrial systems are among the major use cases for 5G and beyond systems: in these cases, the massive amounts of sensory information that needs to be shared over the wireless medium can overload even high-capacity connections. Consequently, solving the effective communication problem by optimizing the transmission strategy to discard irrelevant information can provide a significant advantage, but is often a very complex task. In this work, we consider a prototypal system in which an observer must communicate its sensory data to an actor controlling a task (e.g., a mobile robot in a factory). We then model it as a remote Partially Observable Markov Decision Process (POMDP), considering the effect of adopting semantic and effective communication-oriented solutions on the overall system performance. We split the communication problem by considering an ensemble Vector Quantized Variational Autoencoder (VQ-VAE) encoding, and train a Deep Reinforcement Learning (DRL) agent to dynamically adapt the quantization level, considering both the current state of the environment and the memory of past messages. We tested the proposed approach on the well-known CartPole reference control problem, obtaining a significant performance increase over traditional approaches
Automatic gym activity recognition on energy- and resource-constrained wearable devices removes the human-interaction requirement during intense gym sessions - like soft-touch tapping and swiping. This work presents a tiny and highly accurate residual convolutional neural network that runs in milliwatt microcontrollers for automatic workouts classification. We evaluated the inference performance of the deep model with quantization on three resource-constrained devices: two microcontrollers with ARM-Cortex M4 and M7 core from ST Microelectronics, and a GAP8 system on chip, which is an open-sourced, multi-core RISC-V computing platform from GreenWaves Technologies. Experimental results show an accuracy of up to 90.4% for eleven workouts recognition with full precision inference. The paper also presents the trade-off performance of the resource-constrained system. While keeping the recognition accuracy (88.1%) with minimal loss, each inference takes only 3.2 ms on GAP8, benefiting from the 8 RISC-V cluster cores. We measured that it features an execution time that is 18.9x and 6.5x faster than the Cortex-M4 and Cortex-M7 cores, showing the feasibility of real-time on-board workouts recognition based on the described data set with 20 Hz sampling rate. The energy consumed for each inference on GAP8 is 0.41 mJ compared to 5.17 mJ on Cortex-M4 and 8.07 mJ on Cortex-M7 with the maximum clock. It can lead to longer battery life when the system is battery-operated. We also introduced an open data set composed of fifty sessions of eleven gym workouts collected from ten subjects that is publicly available.
Deep Neural Networks (DNNs) and Spiking Neural Networks (SNNs) are both known for their susceptibility to adversarial attacks. Therefore, researchers in the recent past have extensively studied the robustness and defense of DNNs and SNNs under adversarial attacks. Compared to accurate SNNs (AccSNN), approximate SNNs (AxSNNs) are known to be up to 4X more energy-efficient for ultra-low power applications. Unfortunately, the robustness of AxSNNs under adversarial attacks is yet unexplored. In this paper, we first extensively analyze the robustness of AxSNNs with different structural parameters and approximation levels under two gradient-based and two neuromorphic attacks. Then, we propose two novel defense methods, i.e., precision scaling and approximate quantization-aware filtering (AQF), for securing AxSNNs. We evaluated the effectiveness of these two defense methods using both static and neuromorphic datasets. Our results demonstrate that AxSNNs are more prone to adversarial attacks than AccSNNs, but precision scaling and AQF significantly improve the robustness of AxSNNs. For instance, a PGD attack on AxSNN results in a 72\% accuracy loss compared to AccSNN without any attack, whereas the same attack on the precision-scaled AxSNN leads to only a 17\% accuracy loss in the static MNIST dataset (4X robustness improvement). Similarly, a Sparse Attack on AxSNN leads to a 77\% accuracy loss when compared to AccSNN without any attack, whereas the same attack on an AxSNN with AQF leads to only a 2\% accuracy loss in the neuromorphic DVS128 Gesture dataset (38X robustness improvement).
Text-to-speech (TTS) and voice conversion (VC) are two different tasks both aiming at generating high quality speaking voice according to different input modality. Due to their similarity, this paper proposes UnifySpeech, which brings TTS and VC into a unified framework for the first time. The model is based on the assumption that speech can be decoupled into three independent components: content information, speaker information, prosody information. Both TTS and VC can be regarded as mining these three parts of information from the input and completing the reconstruction of speech. For TTS, the speech content information is derived from the text, while in VC it's derived from the source speech, so all the remaining units are shared except for the speech content extraction module in the two tasks. We applied vector quantization and domain constrain to bridge the gap between the content domains of TTS and VC. Objective and subjective evaluation shows that by combining the two task, TTS obtains better speaker modeling ability while VC gets hold of impressive speech content decoupling capability.
Semantic communications learned on background knowledge bases (KBs) have been identified as a promising technology for communications between intelligent agents. Existing works assume that transceivers of semantic communications share the same KB. However, intelligent transceivers may suffer from the communication burden or worry about privacy leakage to exchange data in KBs. Besides, the transceivers may independently learn from the environment and dynamically update their KBs, leading to timely sharing of the KBs infeasible. All these cause the mismatch between the KBs, which may result in a semantic-level misunderstanding on the receiver side. To address this issue, we propose a transceiver cooperative learning-assisted semantic communication (TCL-SC) scheme against mismatched KBs. In TCL-SC, the transceivers cooperatively train semantic encoder and decoder neuron networks (NNs) of the same structure based on their own KBs. They periodically share the parameters of NNs. To reduce the communication overhead of parameter sharing, parameter quantization is adopted. Moreover, we discuss the impacts of the number of communication rounds on the performance of semantic communication systems. Experiments on real-world data demonstrate that our proposed TCL-SC can reduce the semantic-level misunderstanding on the receiver side caused by the mismatch between the KBs, especially at the low signal-to-noise (SNR) ratio regime.
Large-scale Transformer models are known for their exceptional performance in a range of tasks, but training them can be difficult due to the requirement for communication-intensive model parallelism. One way to improve training speed is to compress the message size in communication. Previous approaches have primarily focused on compressing gradients in a data parallelism setting, but compression in a model-parallel setting is an understudied area. We have discovered that model parallelism has fundamentally different characteristics than data parallelism. In this work, we present the first empirical study on the effectiveness of compression methods for model parallelism. We implement and evaluate three common classes of compression algorithms - pruning-based, learning-based, and quantization-based - using a popular Transformer training framework. We evaluate these methods across more than 160 settings and 8 popular datasets, taking into account different hyperparameters, hardware, and both fine-tuning and pre-training stages. We also provide analysis when the model is scaled up. Finally, we provide insights for future development of model parallelism compression algorithms.