Existing learning-based visual retouching primarily focuses on improving image quality through end-to-end objective mapping between input and retouched images. However, these approaches often overlook two critical aspects: the progressive nature of image retouching and the subjective aesthetic preferences, resulting in suboptimal visual outcomes. To address this, we introduce Automatic Aesthetic Image Retouching via active reinforcement learning (A$^{3}$3RL) to enhance the visualization experience in two sub-tasks: color enhancement and composition optimization, which are formulated as a unified Markov Decision Process in the proposed A$^{3}$3RL framework. In our approach, each pixel functions as an autonomous agent that determines optimal actions based on aesthetic guidance, engaging in online exploration through immediate pixel-wise and channel-wise feedback from the aesthetic environment. By leveraging a pretrained image aesthetic model, our method ensures that the A$^{3}$3RL process aligns with human aesthetic preferences and adheres to subjective aesthetic principles. The framework integrates pixel-level retouching actions with image-level operations to achieve optimal image sequences through progressive iterations. Extensive experiments demonstrate that our method effectively recalibrates image aesthetics across multiple dimensions: low-level quality metrics (PSNR, SSIM), visual perception (LPIPS), and subjective visual experience (human survey). The results demonstrate high consistency with expert-retouched ground-truth images.

Evaluating the performance of low-light image enhancement (LLE) is highly subjective, thus making integrating human preferences into LLE a necessity. Existing methods fail to consider this and present a series of potentially valid heuristic criteria for training LLE models. In this paper, we propose a new paradigm, i.e., aesthetics-guided low-light image enhancement (ALL-E), which introduces aesthetic preferences to LLE and motivates training in a reinforcement learning framework with an aesthetic reward. Each pixel, functioning as an agent, refines itself by recursive actions. We further present ALL-E+, an extended version of ALL-E, which casts a two-stage aesthetics-guided enhancement and denoising. ALL-E+ achieves low-light enhancement and denoising compensation sequentially in a unified framework, resulting in significant improvements in both subjective visual experience and objective evaluation. Extensive experiments show that integrating aesthetic preferences can further improve the visual experience of enhanced images. Our results on various benchmarks also demonstrate the superiority of our method over state-of-the-art methods.

High-performance key-value (KV) storage is critical for the cloud, providing KV services to various cloud applications. A key challenge for KV services is that workloads of cloud applications are often write-intensive and exhibit highly-skewed characteristics, which result in load imbalance among storage servers thus lowering system performance. To address this problem, prior works set up an in-switch write-back caching mechanism, which adopts a centralized controller to balance writes. However, the limited bandwidth between the switches and the controller is a severe bottleneck for achieving high performance. In this paper, we present Magneto, a novel key-value service architecture for the cloud. At the core of Magneto is a delayed-write mechanism in the switch data plane to absorb frequent write queries for hot items. It effectively balances the load under write-intensive workloads without involving the controller. Magneto also designs a reliability mechanism to ensure system reliability during switch state transitions and failures. We implement a prototype using an FPGA-integrated switch, which has high packet-processing performance and contains enough memory to provide both the in-switch cache and the write buffer. Extensive evaluation shows that Magneto can achieve 8.4x system throughput gains compared to baseline systems when handling a skewed workload consisting of 70% reads and 30% writes. Moreover, it can reduce the load on back-end servers up to 56% in total.

In the tide of explosive development in artificial intelligence (AI), the design of AI System-on-Chips (SoCs) is an urgently pressing issue that needs to be addressed. The application of Design Space Exploration (DSE) methods is paramount in pursuing a sound microarchitecture design and improving the quality of results. However, the high-dimensional design parameters and huge design space, which normally occur in the complicated SoCs for Large Language Model (LLM) tasks, pose a great challenge to existing techniques. In this paper, a novel and explainable Bayesian optimization-based framework MCT-Explorer is proposed. A Monte Carlo Tree Search (MCTS)-based method is utilized to analyze the importance of design parameters, guide the sampling directions, mitigate low-quality performance modeling issues, and further improve optimization efficiency. Besides, an information-guided multi-objective optimization function is adopted to balance the multiple metrics (e.g., Cycle. Area, and Power) for SoC design. Our approach can provide guiding opinions and deeper insights for parameter optimization, thus transcending previous arts and achieving an explainable model. Experiment results demonstrate the extraordinary performance of our framework in various high-dimensional (up to hundreds of parameters) and complicated LLM SoC designs.

Key-value stores are critical for modern data centers, yet they struggle with performance degradation under write-intensive workloads due to frequent cache invalidations. Traditional read-optimized caching mechanisms fail to maintain efficiency as frequent updates lead to obsolete cached data. As a result, queries sent by clients will suffer from long queuing delays or even be dropped and overall system performance will be degraded severely. This paper presents a novel key-value stores architecture, named Freq-absorb, which addresses this issue through a delayed-write mechanism delaying writes commitment to backend servers. By buffering write queries within the switch for a specific period, Freq-absorb reduces the impact of cache invalidations, improves switch hit rates, and enhances overall system performance. We implement a prototype using an FPGA-based switch that acts as the ToR (Top of Rack) switch to achieve cache and the appended write buffer. Our experimental results show that compared to the read cache mechanism, our approach can reduce switch miss from 99% to 48.4% when handling write-intensive workloads, significantly enhancing performance.

Evaluating the performance of low-light image enhancement (LLE) is highly subjective, thus making integrating human preferences into image enhancement a necessity. Existing methods fail to consider this and present a series of potentially valid heuristic criteria for training enhancement models. In this paper, we propose a new paradigm, i.e., aesthetics-guided low-light image enhancement (ALL-E), which introduces aesthetic preferences to LLE and motivates training in a reinforcement learning framework with an aesthetic reward. Each pixel, functioning as an agent, refines itself by recursive actions, i.e., its corresponding adjustment curve is estimated sequentially. Extensive experiments show that integrating aesthetic assessment improves both subjective experience and objective evaluation. Our results on various benchmarks demonstrate the superiority of ALL-E over state-of-the-art methods. Source code: https://dongl-group.github.io/project pages/ALLE.html