Markov chain Monte Carlo (MCMC) algorithms are widely used in Bayesian inference to compute the posterior distribution of complex models, facilitating sampling from probability distributions. However, the computational burden of evaluating the likelihood function in MCMC poses significant challenges in big data applications. To address this, sub-sampling methods have been introduced to approximate the target distribution by using subsets of the data rather than the entire dataset. Unfortunately, these methods often lead to biased samples, making them impractical for real-world applications. This paper proposes a novel scaling MCMC method that achieves exact sampling by utilizing a subset (mini-batch) of the data with locally bounded approximations of the target distribution. Our method adaptively adjusts the mini-batch size by automatically tuning a hyperparameter based on the sample acceptance ratio, ensuring optimal balance between sample efficiency and computational cost. Moreover, we introduce a highly optimized hardware architecture to efficiently implement the proposed MCMC method onto FPGA. Our accelerator is evaluated on an AMD Zynq UltraScale+ FPGA device using a Bayesian logistic regression model on the MNIST dataset. The results demonstrate that our design achieves unbiased sampling with a 47.6× speedup over the standard MCMC design, while also significantly reducing estimation errors compared to state-of-the-art MCMC methods.

Convolutional neural networks (CNNs) have demonstrated outstanding accuracy among a range of machine learning tasks. However, the huge computational overhead limits their deployability in real-time applications. For this reason, parallel computing has been extensively employed to accelerate CNNs in parallel computing devices, such as GPUs and field programmable gate arrays (FPGAs), by unrolling multiple loop operations of convolutional layers. Nevertheless, existing CNN accelerators can hardly exploit different parallelisms offered by the CNN algorithms efficiently, since their degrees of parallelism are fixed at different dimensions and layers. In this article, we propose the DCP-CNN, an FPGA-based CNN accelerator which implements the CNN with Dynamic Computing Parallelism degrees. DCP-CNN employs a parallel computing architecture which dynamically allocates the computing resources between different data dimensions of each layer based on layer size, to ensure that all computing units are working to full capacity and thus achieve optimal compute efficiency (CE). Furthermore, in order to boost the performance of throughput, we propose a design space exploration (DSE) framework based on the simulated annealing method, which automatically generates the parallelism degrees between different dimensions of the network layers, according to the resource constraints and CNN structure. On Intel Stratix 10 GX650 FPGA, the proposed DCP-CNN achieves the throughput of more than 800 Gop/s and the CE of 72%–98%, which outperforms the existing state-of-the-art FPGA-based CNN accelerators.

Computing convolutional layers in the frequency domain using fast Fourier transformation (FFT) has been demonstrated to be effective in reducing the computational complexity of convolutional neural networks (CNNs). Nevertheless, the main challenge of this approach lies in the frequent and repeated transformations between the spatial and frequency domains due to the absence of nonlinear functions in the spectral domain, as such it makes the benefit less attractive for low-latency inference, especially on embedded platforms. To overcome the drawbacks in the existing FFT-based convolution, we propose a fully spectral CNN using a novel spectral-domain adaptive rectified linear unit (ReLU) layer, which completely removes the compute-intensive transformations between the spatial and frequency domains within the network. The proposed fully spectral CNNs maintain the nonlinearity of the spatial CNNs while taking into account the hardware efficiency. We then propose a deeply customized and compute-efficient hardware architecture to accelerate the fully spectral CNN inference on field programmable gate array (FPGA). Different hardware optimizations, such as spectral-domain intralayer and interlayer pipeline techniques, are introduced to further improve the performance of throughput. To achieve a load-balanced pipeline, a design space exploration (DSE) framework is proposed to optimize the resource allocation between hardware modules according to the resource constraints. On an Intel's Arria 10 SX160 FPGA, our optimized accelerator achieves a throughput of 204 Gop/s with 80% of compute efficiency. Compared with the state-of-the-art spatial and FFT-based implementations on the same device, our accelerator is 4× ∼ 6.6× and 3.0× ∼ 4.4× faster while maintaining a similar level of accuracy across different benchmark datasets.

Over the past few years, 2-D convolutional neural networks (CNNs) have demonstrated their great success in a wide range of 2-D computer vision applications, such as image classification and object detection. At the same time, 3-D CNNs, as a variant of 2-D CNNs, have shown their excellent ability to analyze 3-D data, such as video and geometric data. However, the heavy algorithmic complexity of 2-D and 3-D CNNs imposes a substantial overhead over the speed of these networks, which limits their deployment in real-life applications. Although various domain-specific accelerators have been proposed to address this challenge, most of them only focus on accelerating 2-D CNNs, without considering their computational efficiency on 3-D CNNs. In this article, we propose a unified hardware architecture to accelerate both 2-D and 3-D CNNs with high hardware efficiency. Our experiments demonstrate that the proposed accelerator can achieve up to 92.4% and 85.2% multiply-accumulate efficiency on 2-D and 3-D CNNs, respectively. To improve the hardware performance, we propose a hardware-friendly quantization approach called static block floating point (BFP), which eliminates the frequent representation conversions required in traditional dynamic BFP arithmetic. Comparing with the integer linear quantization using zero-point, the static BFP quantization can decrease the logic resource consumption of the convolutional kernel design by nearly 50% on a field-programmable gate array (FPGA). Without time-consuming retraining, the proposed static BFP quantization is able to quantize the precision to 8-bit mantissa with negligible accuracy loss. As different CNNs on our reconfigurable system require different hardware and software parameters to achieve optimal hardware performance and accuracy, we also propose an automatic tool for parameter optimization. Based on our hardware design and optimization, we demonstrate that the proposed accelerator can achieve 3.8-5.6 times higher energy efficiency than graphics processing unit (GPU) implementation. Comparing with the state-of-the-art FPGA-based accelerators, our design achieves higher generality and up to 1.4-2.2 times higher resource efficiency on both 2-D and 3-D CNNs.

Recent advances in algorithm-hardware co-design for deep neural networks (DNNs) have demonstrated their potential in automatically designing neural architectures and hardware designs. Nevertheless, it is still a challenging optimization problem due to the expensive training cost and the time-consuming hardware implementation, which makes the exploration on the vast design space of neural architecture and hardware design intractable. In this paper, we demonstrate that our proposed approach is capable of locating designs on the Pareto frontier. This capability is enabled by a novel three-phase co-design framework, with the following new features: (a) decoupling DNN training from the design space exploration of hardware architecture and neural architecture, (b) providing a hardware-friendly neural architecture space by considering hardware characteristics in constructing the search cells, (c) adopting Gaussian process to predict accuracy, latency and power consumption to avoid time-consuming synthesis and place-and-route processes. In comparison with the manually-designed ResNet101, InceptionV2 and MobileNetV2, we can achieve up to 5% higher accuracy with up to$3\times$speed up on the ImageNet dataset. Compared with other state-of-the-art co-design frameworks, our found network and hardware configuration can achieve 2% (~ 6% higher accuracy,$2\times\sim 26\times$smaller latency and$8.5\times$higher energy efficiency.

Neural networks (NNs) have demonstrated their potential in a variety of domains ranging from computer vision (CV) to natural language processing. Among various NNs, two-dimensional (2-D) and three-dimensional (3-D) convolutional NNs (CNNs) have been widely adopted for a broad spectrum of applications, such as image classification and video recognition, due to their excellent capabilities in extracting 2-D and 3-D features. However, standard 2-D and 3-D CNNs are not able to capture their model uncertainty which is crucial for many safety-critical applications, including healthcare and autonomous driving. In contrast, Bayesian CNNs (BayesCNNs), as a variant of CNNs, have demonstrated their ability to express uncertainty in their prediction via a mathematical grounding. Nevertheless, BayesCNNs have not been widely used in industrial practice due to their compute requirements stemming from sampling and subsequent forward passes through the whole network multiple times. As a result, these requirements significantly increase the amount of computation and memory consumption in comparison to standard CNNs. This article proposes a novel field-programmable gate array (FPGA)-based hardware architecture to accelerate both 2-D and 3-D BayesCNNs based on Monte Carlo dropout (MCD). Compared with other state-of-the-art accelerators for BayesCNNs, the proposed design can achieve up to four times higher energy efficiency and nine times better compute efficiency. An automatic framework capable of supporting partial Bayesian inference is proposed to explore the tradeoff between algorithm and hardware performance. Extensive experiments are conducted to demonstrate that our framework can effectively find the optimal implementations in the design space.

Due to the huge success and rapid development of convolutional neural networks (CNNs), there is a growing demand for hardware accelerators that accommodate a variety of CNNs to improve their inference latency and energy efficiency, in order to enable their deployment in real-time applications. Among popular platforms, field-programmable gate arrays (FPGAs) have been widely adopted for CNN acceleration because of their capability to provide superior energy efficiency and low-latency processing, while supporting high reconfigurability, making them favorable for accelerating rapidly evolving CNN algorithms. This article introduces a highly customized streaming hardware architecture that focuses on improving the compute efficiency for streaming applications by providing full-stack acceleration of CNNs on FPGAs. The proposed accelerator maps most computational functions, that is, convolutional and deconvolutional layers into a singular unified module, and implements the residual and concatenative connections between the functions with high efficiency, to support the inference of mainstream CNNs with different topologies. This architecture is further optimized through exploiting different levels of parallelism, layer fusion, and fully leveraging digital signal processing blocks (DSPs). The proposed accelerator has been implemented on Intel's Arria 10 GX1150 hardware and evaluated with a wide range of benchmark models. The results demonstrate a high performance of over 1.3 TOP/s of throughput, up to 97% of compute [multiply-accumulate (MAC)] efficiency, which outperforms the state-of-the-art FPGA accelerators.

Computing convolutional layers in frequency domain can largely reduce the computation overhead for training and inference of convolutional neural networks (CNNs). However, existing designs with such an idea require repeated spatial- and frequency-domain transforms due to the absence of nonlinear functions in the frequency domain, as such it makes the benefit less attractive for low-latency inference. This paper presents a fully spectral CNN approach by proposing a novel adaptive Rectified Linear Unit (ReLU) activation in spectral domain. The proposed design maintains the non-linearity in the network while taking into account the hardware efficiency in algorithm level. The spectral model size is further optimized by merging and fusing layers. Then, a customized hardware architecture is proposed to implement the designed spectral network on FPGA device with DSP optimizations for 8-bit fixed point multipliers. Our hardware accelerator is implemented on Intel's Arria 10 device and applied to the MNIST, SVHN, AT&T and CIFAR-10 datasets. Experimental results show a speed improvement of 6 × ~ 10 × and 4 × ~ 5.7 × compared to state-of-the-art spatial or FFT-based designs respectively, while achieving similar accuracy across the benchmark datasets.

Existing FPGA accelerators for short read mapping often fail to utilize the complete biological information in sequencing data for simple hardware design, leading to missed or incorrect alignment. Furthermore, their performance may not be optimized across hardware platforms. This paper proposes a novel alignment pipeline that considers all information in sequencing data for biologically accurate acceleration of short read mapping. To ensure the performance of the proposed design optimized across different platforms, we accelerate the memory-bound operations which have been a bottleneck in short read mapping. Specifically, we partition the FM-index into buckets. The length of each bucket is equal to an optimal multiple of the memory burst size and is determined through data-driven exploration. A tool has been developed to obtain the optimal parameters of the design for different hardware platforms to enhance performance optimization. Experimental results indicate that our design maximizes alignment accuracy compared to the state-of-the-art software Bowtie, mapping reads 4.48x as fast. Compared to the previous hardware aligner, our achieved accuracy is 97.7% which reports 4.48 M more valid alignments with a similar speed.

Neural architecture search (NAS) aims to find the optimal neural network automatically for different scenarios. Among various NAS methods, the differentiable NAS (DNAS) approach has demonstrated its effectiveness in terms of searching cost and final accuracy. However, most of previous efforts focus on applying DNAS to GPU or CPU platforms, and its potential is less exploited on the FPGA. In this paper, we first propose a novel FPGA-based CNN accelerator. An accurate performance model of the proposed hardware design is also introduced. To improve accuracy as well as hardware performance, we then apply DNAS and encapsulate the proposed performance model into the objective function. Based on our FPGA design and NAS method, the experiments demonstrate that the network generated by NAS achieves nearly 95% accuracy on CIFAR-10, while decreasing latency by nearly 12 times compared with existing work.

Three-dimensional convolutional neural networks (3D CNNs) have demonstrated their outstanding classification accuracy for human action recognition (HAR). However, the large number of computations and parameters in 3D CNNs limits their deployability in real-life applications. To address this challenge, this paper adopts an algorithm-hardware co-design method by proposing an efficient 3D CNN building unit called 3D-1 bottleneck residual block (3D-1 BRB) at the algorithm level, and a corresponding FPGA-based hardware architecture called F-E3D at the hardware level. Based on 3D-1 BRB, a novel 3D CNN model called E3DNet is developed, which achieves nearly 37 times reduction in model size and 5% improvement in accuracy compared to standard 3D CNNs on the UCF101 dataset. Together with several hardware optimizations, including 3D fused BRB, online blocking and kernel reuse, the proposed F-E3D is nearly 13 times faster than a previous FPGA design for 3D CNNs, with performance and accuracy comparable to other state-of-the-art 3D CNN models on GPU platforms while requiring only 7% of their energy consumption.

Long Short-Term Memory (LSTM) networks have been deployed in speech recognition, natural language processing and financial calculations in recent years, and are beginning to be used in systems where low latency and low power are required. To meet such requirements, we propose a stall-free hardware architecture by reorganising the order of operations in an LSTM system and develop a unique blocking-batching strategy to reuse the LSTM weights fetched from external memory to optimise the benefits of on-chip memory with a limited size for a large machine learning model. Evaluation results show that our architecture can achieve up to 20.8 GOPS/W, which would be among the highest for FPGA designs targeting LSTM systems with weights stored in off-chip memory. Comparing to the state-of-the-art design using off-chip memory to store the weights, we achieve 1.65 times higher performance-per-watt efficiency and 1.60 times higher performance-per-DSP efficiency. When compared with CPU and GPU implementation, our novel hardware architecture is 23.7 and 1.3 times faster while consuming 208 and 19.2 times lower energy respectively, which shows that our approach contributes to high performance and low power FPGA-based LSTM systems.

Training Deep Neural Networks (DNNs) requires a significant amount of time and resources to obtain acceptable results, which severely limits its deployment in resource-limited platforms. This paper proposes DarkFPGA, a novel customizable framework to efficiently accelerate the entire DNN training on a single FPGA platform. First, we explore batch-level parallelism to enable efficient training on FPGAs. Second, we devise a novel hardware architecture optimised by a batch-oriented data pattern and tiling techniques to effectively exploit parallelism. Moreover, an analytical model is developed to determine the optimal design parameters for the DarkFPGA accelerator with respect to a specific network specification and FPGA resource constraints. Our results show that the accelerator is able to perform about 11 times faster than CPU training and about a third of the energy consumption than GPU training using 8-bit integers for training VGG-like networks on the CIFAR dataset for the Maxeler MAX5 platform.

Among popular techniques in remote sensing image (RSI) segmentation, Deep Neural Networks (DNNs) have gained increasing interest but often require high computation complexity, which largely limits their applicability in on-board space platforms. Therefore, various dedicated hardware designs on FPGAs have been developed to accelerate DNNs. However, it imposes difficulty on the design of efficient accelerators for DNN-based segmentation algorithms, since they need to perform both convolution and deconvolution which are two fundamentally different types of operations. This paper proposes a uniform architecture to efficiently implement both convolution and deconvolution in one vector multiplication module. This architecture is further optimized through exploiting different levels of parallelism and layer fusion to achieve low latency for RSI segmentation tasks. Moreover, an optimized DNN model is developed for real-time RSI segmentation, which shows preferable accuracy compared to other methods. The proposed hardware accelerator efficiently implements the DNN model on Intel's Arria 10 device, demonstrating 1578 GOPS of throughput and 17.4 ms of latency, i.e., 57 images per second.

Deploying a deep neural network model on a reconfigurable platform, such as an FPGA, is challenging due to the enormous design spaces of both network models and hardware design. A neural network model has various layer types, connection patterns and data representations, and the corresponding implementation can be customised with different architectural and modular parameters. Rather than manually exploring this design space, it is more effective to automate optimisation throughout an end-to-end compilation process. This paper provides an overview of recent literature proposing novel approaches to achieve this aim. We organise materials to mirror a typical compilation flow: front end, platform-independent optimisation and back end. Design templates for neural network accelerators are studied with a specific focus on their derivation methodologies. We also review previous work on network compilation and optimisation for other hardware platforms to gain inspiration regarding FPGA implementation. Finally, we propose some future directions for related research.

Convolutional Neural Networks (CNNs) based algorithms have been successful in solving image recognition problems, showing very large accuracy improvement. In recent years, deconvolution layers are widely used as key components in the state-of-the-art CNNs for end-to-end training and models to support tasks such as image segmentation. However, the deconvolution algorithms are computationally intensive which limits their applicability to real time applications. Particularly, there has been little research on the efficient implementations of deconvolution algorithms on FPGA platforms. In this work, we propose and develop fully customized deconvolution architecture for CNN-based segmentation algorithms. Besides, memory sharing between the computation modules is proposed for the FPGA-based CNN accelerator as well as for other optimization techniques. Furthermore, a hardware mapping framework is developed to automatically generate the high-throughput hardware design for any given CNN model on the target device. Finally, we implement our designs on Xilinx Zynq-7030 and the deconvolution accelerator achieves a performance of 25.6 GOPS under 200MHz working frequency and a performance density of 0.064 GOPS/DSP using 32-bit quantization, which significantly outperforms previous designs on FPGAs. A real-time application of scene segmentation on Cityscapes Dataset is used to evaluate our CNN accelerator on Zynq-7030 board, and the system achieves a performance of 57.2 GOPS and 0.143 GOPS/DSP using 16-bit quantization, and supports up to 2 frames per second for 512x512 image inputs with a power consumption of only 3.2W.

This paper introduces ADAM, an approach for merging multiple FPGA designs into a single hardware design, so that multiple place-and-route tasks can be replaced by a single task to speed up functional evaluation of designs, especially during the development process. ADAM has three key elements. First, a novel approximate maximum common subgraph detection algorithm with linear time complexity to maximize sharing of resources in the merged design. Second, a prototype tool implementing this common subgraph detection algorithm for dataflow graphs derived from Verilog designs; this tool would also generate the appropriate control circuits to enable selection of the original designs at runtime. Third, a comprehensive analysis of compilation time versus degree of similarity to identify the optimized user parameters for the proposed approach. Experimental results show that ADAM can reduce compilation time by around 5 times when each design is 95% similar to the others, and the compilation time is reduced from 1 hour to 10 minutes in the case of binomial filters.

Human action recognition (HAR) has been widely employed in various applications such as autonomous cars and intelligent video surveillance. Among the algorithms proposed for HAR, the 3D-CNNs algorithm can achieve the best accuracy. However, its algorithmic complexity imposes a substantial overhead over the speed of these networks, which limits their deployment in real-life applications. This paper proposes a novel customizable architecture for 3D-CNNs based on block floating-point (BFP) arithmetic, where the utilization of BFP significantly reduces the bitwidth while eliminating the need to retrain the network. Optimizations such as locality exploration and block alignment with 3D blocking are performed to improve performance and accuracy. An analytical model and tool are developed to predict the optimized parameters for hardware customization based on user constraints such as FPGA resources or accuracy requirement. Experiments show that without retraining, a 15-bit mantissa design using single-precision accumulation on a Xilinx ZC706 device can be 8.2 times faster than an Intel i7-950 processor at 3.07 GHz with only 0.4% accuracy loss.

Convolutional neural network (CNN)-based object detection has been widely employed in various applications such as autonomous driving and intelligent video surveillance. However, the computational complexity of conventional convolution hinders its application in embedded systems. Recently, a mobile-friendly CNN model SSDLite-MobileNetV2 (SSDLiteM2) has been proposed for object detection. This model consists of a novel layer called bottleneck residual block (BRB). Although SSDLiteM2 contains far fewer parameters and computations than conventional CNN models, its performance on embedded devices still cannot meet the requirements of real-time processing. This paper proposes a novel FPGA-based architecture for SSDLiteM2 in combination with hardware optimizations including fused BRB, processing element (PE) sharing and load-balanced channel pruning. Moreover, a novel quantization scheme called partial quantization has been developed, which partially quantizes SSDLiteM2 to 8 bits with only 1.8% accuracy loss. Experiments show that the proposed design on a Xilinx ZC706 device can achieve up to 65 frames per second with 20.3 mean average precision on the COCO dataset.

Generative adversarial networks (GANs) are a class of artificial intelligence algorithms used in unsupervised machine learning, implemented by a system of two neural networks: a generative network (generator) and a discriminative network (discriminator). These two networks compete with each other to perform better at their respective tasks. The generator is typically a deconvolutional neural network and the discriminator is a convolutional neural network (CNN). Deconvolution performs a fundamentally new type of mathematical operation which differs from convolution. While the FPGA-based CNN accelerators have been widely studied in prior work, the acceleration of deconvolutional networks on FPGA is rarely explored. This paper proposes a novel parametrized deconvolutional architecture based on an FPGA-friendly method, in contrast to the transposed convolution implementation in CPUs and GPUs. Hardware design templates which map this architecture to FPGAs are provided with configurable deconvolutional layer parameters. Furthermore, a memory-efficient architecture with a new tiling method is proposed to accelerate the generator of GANs, by storing all intermediate data in on-chip memories and significantly reducing off-chip data transfers. The performance of the proposed accelerator is evaluated using a variety of GANs on a Xilinx Zynq 706 board, which shows 2.3x higher speed and 8.2x off-chip memory access reduction than an optimized Vanilla FPGA design. Compared to the respective implementations on CPUs and GPUs, the achieved improvements are in the range of 30x-92x in speed over an Intel 8-core i7-950 CPU, and 8x-108x in terms of Performance-per-Watt over an NVIDIA Titan X GPU.

Convolutional Neural Networks-- (CNNs) based algorithms have been successful in solving image recognition problems, showing very large accuracy improvement. In recent years, deconvolution layers are widely used as key components in the state-of-the-art CNNs for end-to-end training and models to support tasks such as image segmentation and super resolution. However, the deconvolution algorithms are computationally intensive, which limits their applicability to real-time applications. Particularly, there has been little research on the efficient implementations of deconvolution algorithms on FPGA platforms that have been widely used to accelerate CNN algorithms by practitioners and researchers due to their high performance and power efficiency. In this work, we propose and develop deconvolution architecture for efficient FPGA implementation. FPGA-based accelerators are proposed for both deconvolution and CNN algorithms. Besides, memory sharing between the computation modules is proposed for the FPGA-based CNN accelerator as well as for other optimization techniques. A non-linear optimization model based on the performance model is introduced to efficiently explore the design space to achieve optimal processing speed of the system and improve power efficiency. Furthermore, a hardware mapping framework is developed to automatically generate the low-latency hardware design for any given CNN model on the target device. Finally, we implement our designs on Xilinx Zynq ZC706 board and the deconvolution accelerator achieves a performance of 90.1 giga operations per second (GOPS) under 200MHz working frequency and a performance density of 0.10 GOPS/DSP using 32-bit quantization, which significantly outperforms previous designs on FPGAs. A real-time application of scene segmentation on Cityscapes Dataset is used to evaluate our CNN accelerator on Zynq ZC706 board, and the system achieves a performance of 107 GOPS and 0.12 GOPS/DSP using 16-bit quantization and supports up to 17 frames per second for 512 × 512 image inputs with a power consumption of only 9.6W.

Markov Chain Monte Carlo (MCMC) based methods have been the main tool for Bayesian Inference for some years now, and recently they find increasing applications in modern statistics and machine learning. Nevertheless, with the availability of large datasets and increasing complexity of Bayesian models, MCMC methods are becoming prohibitively expensive for real-world problems. At the heart of these methods, lies the computation of likelihood functions that requires access to all input data points in each iteration of the method. Current approaches, based on data subsampling, aim to accelerate these algorithms by reducing the number of the data points for likelihood evaluations at each MCMC iteration. However the existing work doesn't consider the properties of modern memory hierarchies, but treats the memory as one monolithic storage space. This paper proposes a communication-aware MCMC framework that takes into account the underlying performance of the memory subsystem. The framework is based on a novel subsampling algorithm that utilises an unbiased likelihood estimator based on Probability Proportional-to-Size (PPS) sampling, allowing information on the performance of the memory system to be taken into account during the sampling stage. The proposed MCMC sampler is mapped to an FPGA device and its performance is evaluated using the Bayesian logistic regression model on MNIST dataset. The proposed system achieves a 3.37× speed up over a highly optimised traditional FPGA design, therefore the risk in the estimates based on the generated samples is largely decreased.

Genetic sequence alignment has always been a computational challenge in bioinformatics. Depending on the problem size, software-based aligners can take multiple CPU-days to process the sequence data, creating a bottleneck point in bioinformatic analysis flow. Reconfigurable accelerator can achieve high performance for such computation by providing massive parallelism, but at the expense of programming flexibility and thus has not been commensurately used by practitioners. Therefore, this paper aims to provide a thorough survey of the proposed accelerators by giving a qualitative categorization based on their algorithms and speedup. A comprehensive comparison between work is also presented so as to guide selection for biologist, and to provide insight on future research direction for FPGA scientists.

Markov Chain Monte Carlo (MCMC) based methods have been the main tool used for Bayesian Inference by practitioners and researchers due to their flexibility and theoretical properties that guarantee unbiased sampling-based estimates. Nevertheless, with the availability of large data sets and the constant need to develop more complex models that better capture the targeted problem, significant computational challenges have been presented. Current approaches, based on multi-core CPUs, GPUs, and FPGAs, aim to accelerate the execution time of the MCMC methods using subsampling techniques or custom precision arithmetic, resulting to biased estimates. In this work, a novel FPGA-based construction is proposed that utilises the custom precision support of FPGA devices in order to accelerate the computations, guaranteeing at the same time asymptotically unbiased estimates. Key to this approach is the extension of the parameter space by an extra parameter that indicates the required precision in the computation of the likelihood of a data point. The work proposes an FPGA architecture for the above algorithm, as well as discuss its tuning for maximising the performance of the system. The performance of the FPGA-mapped sampler is evaluated using two Bayesian logistic regression case studies of varying complexity, which show significant speedups compared to existing FPGAand CPU-based works that utilise double floating point arithmetic, without any bias on the sampling-based estimates.

Markov chain Monte Carlo (MCMC) is one of the most popular and important tools to generate random samples from probability distributions over many variables which occur frequently in Bayesian inference. However, MCMC cannot be practically applied to models with large data sets because of the prohibitive costly likelihood evaluations for each data point. Previous solutions propose to compute the likelihood approximately, e.g. by sub-sampling data or by using custom precision implementations. These methods involve a trade-off between bias in the output and sampling speed; therefore they cannot guarantee unbiased sampling, which is critical in many applications. This paper introduces a novel mixed precision MCMC accelerator for FPGAs, which simulates from the exact probability distribution in contrast to existing approximate MCMC samplers. An auxiliary binary variable is appended to each data point to indicate the corresponding likelihood term evaluation in full or reduced precision. The proposed method guarantees unbiased samples, while the large majority of likelihood computations are performed in reduced precision. Moreover, a tailored FPGA architecture for the algorithm is introduced, and its performance is evaluated using two Bayesian logistic regression case studies of varying complexity: a 2-dimension synthetic problem and MNIST classification with 12-dimension parameters. The achieved speedups over double-precision FPGA designs are 4.21× and 4.76× respectively.

Particle filters (PFs) are a set of algorithms that implement recursive Bayesian filtering, which represent the posterior distribution by a set of weighted samples. Resampling is a fundamental operation in PF algorithms. It consists of taking a population of samples and reconstructing it based on the weights attached to each sample, favouring the samples with large weights. However, resampling is computationally intensive when the number of samples is large and, most importantly, it is not inherently parallelizable like the other steps of the particle filter. Parallel computing devices such as Graphics Processing Units (GPUs) and Field Programmable Gate Arrays (FPGAs) have been proposed to accelerate resampling. In this paper, we propose novel parallel architectures that map four state-of-the-art resampling algorithms (systematic, residual systematic, Metropolis and Rejection resampling) to a FPGA. FPGA-specific optimisations are introduced to further optimize the performance of the above systems. The proposed architectures are implemented in a Virtex-6 LX240T FPGA device with half-utilization of logic resources. Compared to the respective state-of-the-art implementations on an NVIDIA K20 GPU, the achieved speedups are in the range of 1.7x-49x.