Hardware image signal processing (ISP) transforms RAW inputs into high-quality RGB images through a series of processing modules, each with numerous tunable parameters. Traditionally, these parameters are manually tuned by imaging experts, a time-consuming and subjective process. Recent deep learning approaches predict ISP parameters, but often treat the process as a black box and overlook the intrinsic relationships among ISP modules. To address these fundamental issues, we introduce a novel ISP parameter optimization model based on single-agent reinforcement learning (RL) (i.e., SARL-ISP), formulating the hardware ISP parameter tuning as a sequential optimization problem. During the optimization process, the agent updates ISP parameter tuning strategies for different tasks through interaction with the environment. In order to explore the influence of the sequential structure of hardware ISP modules and the coupling relationships among ISP parameters on the tuning process, we further propose a sequential ISP framework based on collaborative multi-agent RL (i.e., MARL-ISP). Specifically, the serialized parameter tuning module (SPTM) realistically simulates the process of manual prediction and module pipeline. Additionally, the feature selection module (FSM) facilitates the transmission and fusion of agent features, thereby selecting more appropriate feature inputs for downstream tasks. Extensive experiments across various tasks (e.g., object detection, instance segmentation) validate the effectiveness and efficiency of our models. Even with minimal training data, our models also outperform current state-of-the-art methods in both quantitative metrics and qualitative evaluations.

Hardware image signal processing (ISP), aiming at converting RAW inputs to RGB images, consists of a series of processing blocks, each with multiple parameters. Traditionally, ISP parameters are manually tuned in isolation by imaging experts according to application-specific quality and performance metrics, which is time-consuming and biased towards human perception due to complex interaction with the output image. Since the relationship between any single parameter’s variation and the output performance metric is a complex, non-linear function, optimizing such a large number of ISP parameters is challenging. To address this challenge, we propose a novel Sequential ISP parameter optimization model, called the RL-SeqISP model, which utilizes deep reinforcement learning to jointly optimize all ISP parameters for a variety of imaging applications. Concretely, inspired by the sequential tuning process of human experts, the proposed model can progressively enhance image quality by seamlessly integrating information from both the image feature space and the parameter space. Furthermore, a dynamic parameter optimization module is introduced to avoid ISP parameters getting stuck into local optima, which is able to more effectively guarantee the optimal parameters resulting from the sequential learning strategy. These merits of the RL-SeqISP model as well as its high efficiency are substantiated by comprehensive experiments on a wide range of downstream tasks, including two visual analysis tasks (instance segmentation and object detection), and image quality assessment (IQA), as compared with representative methods both quantitatively and qualitatively. In particular, even using only 10% of the training data, our model outperforms other SOTA methods by an average of 7% mAP on two visual analysis tasks.

Human leukocyte antigen (HLA) can encode the human major histocompatibility complex (MHC) proteins and play a key role in adaptive and innate immunity. Emerging clinical evidences suggest that the presentation of tumor neoantigens and neoantigen-specific T cell response associated with MHC class I molecules are of key importance to activate the adaptive immune systemin cancer immunotherapy. Therefore, accurate HLA typing is very essential for the clinical application of immunotherapy. In this study, we conducted performance evaluations of 4 widely used HLA typing tools (OptiType, Phlat, Polysolver and seq2hla) for predicting HLA class Ia genes from WES and RNA-seq data of 28 cancer patients. HLA genotyping data using PCR-SBT method was firstly obtained as the golden standard and was subsequently compared with HLA typing data by using NGS techniques. For both WES data and RNA-seq data, OptiType showed the highest accuracy for HLA-Ia typing than the other 3 programs at 2-digit and 4-digit resolution. Additionally, HLA typing accuracy from WES data was higher than from RNA-seq data (99.11% for WES data versus 96.42% for RNA-seq data). The accuracy of HLA-Ia typing by OptiType can reach 100% with the average depth of HLA gene regions >20x. Besides, the accuracy of 2-digit and 4-digit HLA-Ia typing based on control samples was higher than tumor tissues. In conclusion, OptiType by using WES data from control samples with the high average depth (>20x) of HLA gene regions can present a probably superior performance for HLA-Ia typing, enabling its application in cancer immunotherapy.

In this paper, we firstly review the concerned issues on dynamic reconfiguration of non-stop dataflow systems with QoS requirement and what the related work have been done on these issues. Then we present our research by focusing on reconfiguration automation – automated reconfiguration planning and automated execution scheduling with taking the concerned issues on board. Significantly the automated planning removes the reliance of reconfiguration correctness on change administrators. Structures, such as Version-Visibility, Restrict-Segment and Route-Map, along with planning and scheduling algorithms, are proposed in this paper as the fundamentals to automate reconfiguration for maintaining the system’s QoS and preserving the application integrity. All the proposed structures and algorithms have been fully implemented as an architectural meta-model called the RDF (Reconfigurable Dataflow) model. In this paper a dataflow system: the DEDSS (Data Encryption & Digital Signature System) is studied as a typical case on the RDF to demonstrate the effectiveness of our approaches to automating seamless and QoS-ensured reconfiguration of non-stop dataflow systems.

Although many approaches to dynamic reconfiguration have been proposed, how the impact of reconfiguration on system QoS can be controlled has not been addressed adequately so far. In this paper, we propose an approach to dynamic reconfiguration of distributed data flow systems. It is an improvement of our previous work for dynamic reconfiguration with QoS management into distributed environment. Our approach has three features. First, it uses version control, flow trace, and reconfiguration scheduling to avoid logical and physical impact on system QoS. Second, it plans a reconfiguration in a totally automatic way. Third, it executes a reconfiguration by a decentralized protocol to reduce reconfiguration time in distributed environment. We adopt our approach to the reconfiguration of a real world application, the data encryption and digital signature system. Experimental results show that our approach has significant advantages in impact control comparing with other existing approaches.