Deep convolutional neural networks (CNNs) are computationally and memory intensive. In CNNs, intensive multiplication can have resource implications that may challenge the ability for effective deployment of inference on resource-constrained edge devices. This paper proposes GhostShiftAddNet, where the motivation is to implement a hardware-efficient deep network: a multiplication-free CNN with fewer redundant features. We introduce a new bottleneck block, GhostSA, that converts all multiplications in the block to cheap operations. The bottleneck uses an appropriate number of bit-shift filters to process intrinsic feature maps, then applies a series of transformations that consist of bit-wise shifts with addition operations to generate more feature maps that fully learn to capture information underlying intrinsic features. We schedule the number of bit-shift and addition operations for different hardware platforms. We conduct extensive experiments and ablation studies with desktop and embedded (Jetson Nano) devices for implementation and measurements. We demonstrate the proposed GhostSA block can replace bottleneck blocks in the backbone of state-of-the-art networks architectures and gives improved performance on image classification benchmarks. Further, our GhostShiftAddNet can achieve higher classification accuracy with fewer FLOPs and parameters (reduced by up to 3x) than GhostNet. When compared to GhostNet, inference latency on the Jetson Nano is improved by 1.3x and 2x on the GPU and CPU respectively. Code is available open-source on \url{https://github.com/JIABI/GhostShiftAddNet}.
This paper presents a convex-analytic framework to learn sparse graphs from data. While our problem formulation is inspired by an extension of the graphical lasso using the so-called combinatorial graph Laplacian framework, a key difference is the use of a nonconvex alternative to the $\ell_1$ norm to attain graphs with better interpretability. Specifically, we use the weakly-convex minimax concave penalty (the difference between the $\ell_1$ norm and the Huber function) which is known to yield sparse solutions with lower estimation bias than $\ell_1$ for regression problems. In our framework, the graph Laplacian is replaced in the optimization by a linear transform of the vector corresponding to its upper triangular part. Via a reformulation relying on Moreau's decomposition, we show that overall convexity is guaranteed by introducing a quadratic function to our cost function. The problem can be solved efficiently by the primal-dual splitting method, of which the admissible conditions for provable convergence are presented. Numerical examples show that the proposed method significantly outperforms the existing graph learning methods with reasonable CPU time.
We present distributed algorithms for training dynamic Graph Neural Networks (GNN) on large scale graphs spanning multi-node, multi-GPU systems. To the best of our knowledge, this is the first scaling study on dynamic GNN. We devise mechanisms for reducing the GPU memory usage and identify two execution time bottlenecks: CPU-GPU data transfer; and communication volume. Exploiting properties of dynamic graphs, we design a graph difference-based strategy to significantly reduce the transfer time. We develop a simple, but effective data distribution technique under which the communication volume remains fixed and linear in the input size, for any number of GPUs. Our experiments using billion-size graphs on a system of 128 GPUs shows that: (i) the distribution scheme achieves up to 30x speedup on 128 GPUs; (ii) the graph-difference technique reduces the transfer time by a factor of up to 4.1x and the overall execution time by up to 40%
The deployment of artificial neural networks-based optical channel equalizers on edge-computing devices is critically important for the next generation of optical communication systems. However, this is still a highly challenging problem, mainly due to the computational complexity of the artificial neural networks (NNs) required for the efficient equalization of nonlinear optical channels with large dispersion-induced memory. To implement the NN-based optical channel equalizer in hardware, a substantial complexity reduction is needed, while we have to keep an acceptable performance level of the simplified NN model. In this work, we address the complexity reduction problem by applying pruning and quantization techniques to an NN-based optical channel equalizer. We use an exemplary NN architecture, the multi-layer perceptron (MLP), to mitigate the impairments for 30GBd 1000km transmission over a standard single-mode fiber, and demonstrate that it is feasible to reduce the equalizer's memory by up to 87.12%, and its complexity by up to 78.34%, without noticeable performance degradation. In addition to this, we accurately define the computational complexity of a compressed NN-based equalizer in the digital signal processing (DSP) sense. Further, we examine the impact of using different CPU and GPU settings on the power consumption and latency for the compressed equalizer. We also verify the developed technique experimentally, by implementing the reduced NN equalizer on two standard edge-computing hardware units: Raspberry Pi 4 and Nvidia Jetson Nano, which are used to process the data generated via simulating the signal's propagation down the optical-fiber system.
Accurate and complete terrain maps enhance the awareness of autonomous robots and enable safe and optimal path planning. Rocks and topography often create occlusions and lead to missing elevation information in the Digital Elevation Map (DEM). Currently, these occluded areas are either fully avoided during motion planning or the missing values in the elevation map are filled-in using traditional interpolation, diffusion or patch-matching techniques. These methods cannot leverage the high-level terrain characteristics and the geometric constraints of line of sight we humans use intuitively to predict occluded areas. We introduce a self-supervised learning approach capable of training on real-world data without a need for ground-truth information to reconstruct the occluded areas in the DEMs. We accomplish this by adding artificial occlusion to the incomplete elevation maps constructed on a real robot by performing ray casting. We first evaluate a supervised learning approach on synthetic data for which we have the full ground-truth available and subsequently move to several real-world datasets. These real-world datasets were recorded during exploration of both structured and unstructured terrain with a legged robot, and additionally in a planetary scenario on Lunar analogue terrain. We state a significant improvement compared to the baseline methods both on synthetic terrain and for the real-world datasets. Our neural network is able to run in real-time on both CPU and GPU with suitable sampling rates for autonomous ground robots. We motivate the applicability of reconstructing occlusion in elevation maps with preliminary motion planning experiments.
Modern deep learning (DL) applications are built using DL libraries and frameworks such as TensorFlow and PyTorch. These frameworks have complex parameters and tuning them to obtain good training and inference performance is challenging for typical users, such as DL developers and data scientists. Manual tuning requires deep knowledge of the user-controllable parameters of DL frameworks as well as the underlying hardware. It is a slow and tedious process, and it typically delivers sub-optimal solutions. In this paper, we treat the problem of tuning parameters of DL frameworks to improve training and inference performance as a black-box optimization problem. We then investigate applicability and effectiveness of Bayesian optimization (BO), genetic algorithm (GA), and Nelder-Mead simplex (NMS) to tune the parameters of TensorFlow's CPU backend. While prior work has already investigated the use of Nelder-Mead simplex for a similar problem, it does not provide insights into the applicability of other more popular algorithms. Towards that end, we provide a systematic comparative analysis of all three algorithms in tuning TensorFlow's CPU backend on a variety of DL models. Our findings reveal that Bayesian optimization performs the best on the majority of models. There are, however, cases where it does not deliver the best results.
Deep learning is a technique for machine learning using multi-layer neural networks. It has been used for image synthesis and image recognition, but in recent years, it has also been used for various social detection and social labeling. In this analysis, we compared (1) the number of Iterations per minute between the GPU and CPU when using the VGG model and the NIN model, and (2) the number of Iterations per minute by the number of pixels when using the VGG model, using an image with 128 pixels. When the number of pixels was 64 or 128, the processing time was almost the same when using the GPU, but when the number of pixels was changed to 256, the number of iterations per minute decreased and the processing time increased by about three times. In this case study, since the number of pixels becomes core dumping when the number of pixels is 512 or more, we can consider that we should consider improvement in the vector calculation part. If we aim to achieve 8K highly saturated computer graphics using neural networks, we will need to consider an environment that allows computation even when the size of the image becomes even more highly saturated and massive, and parallel computation when performing image recognition and tuning.
Constraint acquisition systems such as QuAcq and MultiAcq can assist non-expert users to model their problems as constraint networks by classifying (partial) examples as positive or negative. For each negative example, the former focuses on one constraint of the target network, while the latter can learn a maximum number of constraints. Two bottlenecks of the acquisition process where both these algorithms encounter problems are the large number of queries required to reach convergence, and the high cpu times needed to generate queries, especially near convergence. In this paper we propose algorithmic and heuristic methods to deal with both these issues. We first describe an algorithm, called MQuAcq, that blends the main idea of MultiAcq into QuAcq resulting in a method that learns as many constraints as MultiAcq does after a negative example, but with a lower complexity. A detailed theoretical analysis of the proposed algorithm is also presented. %We also present a technique that boosts the performance of constraint acquisition by reducing the number of queries significantly. Then we turn our attention to query generation which is a significant but rather overlooked part of the acquisition process. We describe %in detail how query generation in a typical constraint acquisition system operates, and we propose heuristics for improving its efficiency. Experiments from various domains demonstrate that our resulting algorithm that integrates all the new techniques does not only generate considerably fewer queries than QuAcq and MultiAcq, but it is also by far faster than both of them, in average query generation time as well as in total run time, and also largely alleviates the premature convergence problem.
This article describes an efficient end-to-end speech translation (E2E-ST) framework based on non-autoregressive (NAR) models. End-to-end speech translation models have several advantages over traditional cascade systems such as inference latency reduction. However, conventional AR decoding methods are not fast enough because each token is generated incrementally. NAR models, however, can accelerate the decoding speed by generating multiple tokens in parallel on the basis of the token-wise conditional independence assumption. We propose a unified NAR E2E-ST framework called Orthros, which has an NAR decoder and an auxiliary shallow AR decoder on top of the shared encoder. The auxiliary shallow AR decoder selects the best hypothesis by rescoring multiple candidates generated from the NAR decoder in parallel (parallel AR rescoring). We adopt conditional masked language model (CMLM) and a connectionist temporal classification (CTC)-based model as NAR decoders for Orthros, referred to as Orthros-CMLM and Orthros-CTC, respectively. We also propose two training methods to enhance the CMLM decoder. Experimental evaluations on three benchmark datasets with six language directions demonstrated that Orthros achieved large improvements in translation quality with a very small overhead compared with the baseline NAR model. Moreover, the Conformer encoder architecture enabled large quality improvements, especially for CTC-based models. Orthros-CTC with the Conformer encoder increased decoding speed by 3.63x on CPU with translation quality comparable to that of an AR model.
Improving Transformer efficiency has become increasingly attractive recently. A wide range of methods has been proposed, e.g., pruning, quantization, new architectures and etc. But these methods are either sophisticated in implementation or dependent on hardware. In this paper, we show that the efficiency of Transformer can be improved by combining some simple and hardware-agnostic methods, including tuning hyper-parameters, better design choices and training strategies. On the WMT news translation tasks, we improve the inference efficiency of a strong Transformer system by 3.80X on CPU and 2.52X on GPU. The code is publicly available at https://github.com/Lollipop321/mini-decoder-network.
We propose an energy-efficient controller to minimize the energy consumption of a mobile robot by dynamically manipulating the mechanical and computational actuators of the robot. The mobile robot performs real-time vision-based applications based on an event-based camera. The actuators of the controller are CPU voltage/frequency for the computation part and motor voltage for the mechanical part. We show that independently considering speed control of the robot and voltage/frequency control of the CPU does not necessarily result in an energy-efficient solution. In fact, to obtain the highest efficiency, the computation and mechanical parts should be controlled together in synergy. We propose a fast hill-climbing optimization algorithm to allow the controller to find the best CPU/motor configuration at run-time and whenever the mobile robot is facing a new environment during its travel. Experimental results on a robot with Brushless DC Motors, Jetson TX2 board as the computing unit, and a DAVIS-346 event-based camera show that the proposed control algorithm can save battery energy by an average of 50.5%, 41%, and 30%, in low-complexity, medium-complexity, and high-complexity environments, over baselines.
Quantization has been proven to be a vital method for improving the inference efficiency of deep neural networks (DNNs). However, it is still challenging to strike a good balance between accuracy and efficiency while quantizing DNN weights or activation values from high-precision formats to their quantized counterparts. We propose a new method called elastic significant bit quantization (ESB) that controls the number of significant bits of quantized values to obtain better inference accuracy with fewer resources. We design a unified mathematical formula to constrain the quantized values of the ESB with a flexible number of significant bits. We also introduce a distribution difference aligner (DDA) to quantitatively align the distributions between the full-precision weight or activation values and quantized values. Consequently, ESB is suitable for various bell-shaped distributions of weights and activation of DNNs, thus maintaining a high inference accuracy. Benefitting from fewer significant bits of quantized values, ESB can reduce the multiplication complexity. We implement ESB as an accelerator and quantitatively evaluate its efficiency on FPGAs. Extensive experimental results illustrate that ESB quantization consistently outperforms state-of-the-art methods and achieves average accuracy improvements of 4.78%, 1.92%, and 3.56% over AlexNet, ResNet18, and MobileNetV2, respectively. Furthermore, ESB as an accelerator can achieve 10.95 GOPS peak performance of 1k LUTs without DSPs on the Xilinx ZCU102 FPGA platform. Compared with CPU, GPU, and state-of-the-art accelerators on FPGAs, the ESB accelerator can improve the energy efficiency by up to 65x, 11x, and 26x, respectively.
[retracted] We found out that the difference was dependent on the Chainer library, and does not replicate with another library (pytorch) which indicates that the results are probably due to a bug in Chainer, rather than being hardware-dependent. -- old abstract Deep neural networks often present uncertainties such as hardware- and software-derived noise and randomness. We studied the effects of such uncertainty on learning outcomes, with a particular focus on the function of graphics processing units (GPUs), and found that GPU-induced uncertainty increased learning accuracy of a certain deep neural network. When training a predictive deep neural network using only the CPU without the GPU, the learning error is higher than when training the same number of epochs using the GPU, suggesting that the GPU plays a different role in the learning process than just increasing the computational speed. Because this effect cannot be observed in learning by a simple autoencoder, it could be a phenomenon specific to certain types of neural networks. GPU-specific computational processing is more indeterminate than that by CPUs, and hardware-derived uncertainties, which are often considered obstacles that need to be eliminated, might, in some cases, be successfully incorporated into the training of deep neural networks. Moreover, such uncertainties might be interesting phenomena to consider in brain-related computational processing, which comprises a large mass of uncertain signals.
Autonomous marine vehicles play an essential role in many ocean science and engineering applications. Planning time and energy optimal paths for these vehicles to navigate in stochastic dynamic ocean environments is essential to reduce operational costs. In some missions, they must also harvest solar, wind, or wave energy (modeled as a stochastic scalar field) and move in optimal paths that minimize net energy consumption. Markov Decision Processes (MDPs) provide a natural framework for sequential decision-making for robotic agents in such environments. However, building a realistic model and solving the modeled MDP becomes computationally expensive in large-scale real-time applications, warranting the need for parallel algorithms and efficient implementation. In the present work, we introduce an efficient end-to-end GPU-accelerated algorithm that (i) builds the MDP model (computing transition probabilities and expected one-step rewards); and (ii) solves the MDP to compute an optimal policy. We develop methodical and algorithmic solutions to overcome the limited global memory of GPUs by (i) using a dynamic reduced-order representation of the ocean flows, (ii) leveraging the sparse nature of the state transition probability matrix, (iii) introducing a neighbouring sub-grid concept and (iv) proving that it is sufficient to use only the stochastic scalar field's mean to compute the expected one-step rewards for missions involving energy harvesting from the environment; thereby saving memory and reducing the computational effort. We demonstrate the algorithm on a simulated stochastic dynamic environment and highlight that it builds the MDP model and computes the optimal policy 600-1000x faster than conventional CPU implementations, making it suitable for real-time use.
Deep reinforcement learning (RL) is a powerful framework to train decision-making models in complex environments. However, RL can be slow as it requires repeated interaction with a simulation of the environment. In particular, there are key system engineering bottlenecks when using RL in complex environments that feature multiple agents with high-dimensional state, observation, or action spaces. We present WarpDrive, a flexible, lightweight, and easy-to-use open-source RL framework that implements end-to-end deep multi-agent RL on a single GPU (Graphics Processing Unit), built on PyCUDA and PyTorch. Using the extreme parallelization capability of GPUs, WarpDrive enables orders-of-magnitude faster RL compared to common implementations that blend CPU simulations and GPU models. Our design runs simulations and the agents in each simulation in parallel. It eliminates data copying between CPU and GPU. It also uses a single simulation data store on the GPU that is safely updated in-place. WarpDrive provides a lightweight Python interface and flexible environment wrappers that are easy to use and extend. Together, this allows the user to easily run thousands of concurrent multi-agent simulations and train on extremely large batches of experience. Through extensive experiments, we verify that WarpDrive provides high-throughput and scales almost linearly to many agents and parallel environments. For example, WarpDrive yields 2.9 million environment steps/second with 2000 environments and 1000 agents (at least 100x higher throughput compared to a CPU implementation) in a benchmark Tag simulation. As such, WarpDrive is a fast and extensible multi-agent RL platform to significantly accelerate research and development.
Isaac Gym offers a high performance learning platform to train policies for wide variety of robotics tasks directly on GPU. Both physics simulation and the neural network policy training reside on GPU and communicate by directly passing data from physics buffers to PyTorch tensors without ever going through any CPU bottlenecks. This leads to blazing fast training times for complex robotics tasks on a single GPU with 2-3 orders of magnitude improvements compared to conventional RL training that uses a CPU based simulator and GPU for neural networks. We host the results and videos at \url{https://sites.google.com/view/isaacgym-nvidia} and isaac gym can be downloaded at \url{https://developer.nvidia.com/isaac-gym}.
The Bidirectional Encoder Representations from Transformers (BERT) model has been radically improving the performance of many Natural Language Processing (NLP) tasks such as Text Classification and Named Entity Recognition (NER) applications. However, it is challenging to scale BERT for low-latency and high-throughput industrial use cases due to its enormous size. We successfully optimize a Query-Title Relevance (QTR) classifier for deployment via a compact model, which we name BERT Bidirectional Long Short-Term Memory (BertBiLSTM). The model is capable of inferring an input in at most 0.2ms on CPU. BertBiLSTM exceeds the off-the-shelf BERT model's performance in terms of accuracy and efficiency for the aforementioned real-world production task. We achieve this result in two phases. First, we create a pre-trained model, called eBERT, which is the original BERT architecture trained with our unique item title corpus. We then fine-tune eBERT for the QTR task. Second, we train the BertBiLSTM model to mimic the eBERT model's performance through a process called Knowledge Distillation (KD) and show the effect of data augmentation to achieve the resembling goal. Experimental results show that the proposed model outperforms other compact and production-ready models.
BERT-based information retrieval models are expensive, in both time (query latency) and computational resources (energy, hardware cost), making many of these models impractical especially under resource constraints. The reliance on a query encoder that only performs tokenization and on the pre-processing of passage representations at indexing, has allowed the recently proposed TILDE method to overcome the high query latency issue typical of BERT-based models. This however is at the expense of a lower effectiveness compared to other BERT-based re-rankers and dense retrievers. In addition, the original TILDE method is characterised by indexes with a very high memory footprint, as it expands each passage into the size of the BERT vocabulary. In this paper, we propose TILDEv2, a new model that stems from the original TILDE but that addresses its limitations. TILDEv2 relies on contextualized exact term matching with expanded passages. This requires to only store in the index the score of tokens that appear in the expanded passages (rather than all the vocabulary), thus producing indexes that are 99% smaller than those of TILDE. This matching mechanism also improves ranking effectiveness by 24%, without adding to the query latency. This makes TILDEv2 the state-of-the-art passage re-ranking method for CPU-only environments, capable of maintaining query latency below 100ms on commodity hardware.
We consider energy minimization for data-intensive applications run on large number of servers, for given performance guarantees. We consider a system, where each incoming application is sent to a set of servers, and is considered to be completed if a subset of them finish serving it. We consider a simple case when each server core has two speed levels, where the higher speed can be achieved by higher power for each core independently. The core selects one of the two speeds probabilistically for each incoming application request. We model arrival of application requests by a Poisson process, and random service time at the server with independent exponential random variables. Our model and analysis generalizes to today's state-of-the-art in CPU energy management where each core can independently select a speed level from a set of supported speeds and corresponding voltages. The performance metrics under consideration are the mean number of applications in the system and the average energy expenditure. We first provide a tight approximation to study this previously intractable problem and derive closed form approximate expressions for the performance metrics when service times are exponentially distributed. Next, we study the trade-off between the approximate mean number of applications and energy expenditure in terms of the switching probability.
Neural architecture search can discover neural networks with good performance, and One-Shot approaches are prevalent. One-Shot approaches typically require a supernet with weight sharing and predictors that predict the performance of architecture. However, the previous methods take much time to generate performance predictors thus are inefficient. To this end, we propose FOX-NAS that consists of fast and explainable predictors based on simulated annealing and multivariate regression. Our method is quantization-friendly and can be efficiently deployed to the edge. The experiments on different hardware show that FOX-NAS models outperform some other popular neural network architectures. For example, FOX-NAS matches MobileNetV2 and EfficientNet-Lite0 accuracy with 240% and 40% less latency on the edge CPU. FOX-NAS is the 3rd place winner of the 2020 Low-Power Computer Vision Challenge (LPCVC), DSP classification track. See all evaluation results at https://lpcv.ai/competitions/2020. Search code and pre-trained models are released at https://github.com/great8nctu/FOX-NAS.