Superconducting qubits are among the most promising candidates for building quantum information processors. Yet, they are often limited by slow and error-prone qubit readout-a critical factor in achieving high-fidelity operations. While current methods, including deep neural networks, enhance readout accuracy, they typically lack support for mid-circuit measurements essential for quantum error correction, and they usually rely on large, resource-intensive network models. This paper presents KLiNQ, a novel qubit readout architecture leveraging lightweight neural networks optimized via knowledge distillation. Our approach achieves around a $99 \%$ reduction in model size compared to the baseline while maintaining a qubitstate discrimination accuracy of $91 \%$. KLiNQ facilitates rapid, independent qubit-state readouts that enable mid-circuit measurements by assigning a dedicated, compact neural network for each qubit. Implemented on the Xilinx UltraScale+ FPGA, our design can perform the discrimination within 32 ns. The results demonstrate that compressed neural networks can maintain highfidelity independent readout while enabling efficient hardware implementation, advancing practical quantum computing.

As modern HPC systems are typically composed of fat and rich compute nodes, it is usually difficult to fully utilize all node resources with a single application. Co-scheduling, i.e., co-executing multiple complementary applications (or jobs) on the same node in a space sharing manner, is a promising solution and thus has been widely studied in the past decade. As one major drawback of co-scheduling is that it induces the interference effects among co-located applications due to contention among shared resources, the industry has started to support several resource/traffic partitioning features, e.g., in shared caches or memory controllers, on modern commercial processors. Recent studies proposed effective approaches to make use of these advanced features, however, the interactions between these features and (1) job scheduling decisions as well as (2) NUMA (Non-Uniform Memory Access) effects were generally overlooked. This paper explicitly targets these two missing pieces and comprehensively harmonizes the following decisions using reinforcement learning: (a) job selections for co-execution from a given job queue; and (b) diverse resource assignments to co-executed jobs, leveraging emerging hardware partitioning features, while taking NUMA-awareness into account. Our evaluation result demonstrates that our approach can improve the total system throughput by up to 78.1% over time sharing-based naive scheduling.

GPU-based heterogeneous architectures are now commonly used in HPC clusters. Due to their architectural simplicity specialized for data-level parallelism, GPUs can offer much higher computational throughput and memory bandwidth than CPUs in the same generation do. However, as the available resources in GPUs have increased exponentially over the past decades, it has become increasingly difficult for a single program to fully utilize them. As a consequence, the industry has started supporting several resource partitioning features in order to improve the resource utilization by co-scheduling multiple programs on the same GPU die at the same time.Driven by the technological trend, this paper focuses on hierarchical resource partitioning on modern GPUs, and as an example, we utilize a combination of two different features available on recent NVIDIA GPUs in a hierarchical manner: MPS (Multi-Process Service), a finer-grained logical partitioning; and MIG (Multi-Instance GPU), a coarse-grained physical partitioning. We propose a method for comprehensively co-optimizing the setup of hierarchical partitioning and the selection of co-scheduling groups from a given set of jobs, based on reinforcement learning using their profiles. Our thorough experimental results demonstrate that our approach can successfully set up job concurrency, partitioning, and co-scheduling group selections simultaneously. This results in a maximum throughput improvement by a factor of 1.87 compared to the time-sharing scheduling.

In federated learning, all networked clients contribute to the model training cooperatively. However, with model sizes increasing, even sharing the trained partial models often leads to severe communication bottlenecks in underlying networks, especially when communicated iteratively. In this paper, we introduce a federated learning framework FedD3 requiring only one-shot communication by integrating dataset distillation instances. Instead of sharing model updates in other federated learning approaches, FedD3 allows the connected clients to distill the local datasets independently, and then aggregates those decentralized distilled datasets (e.g. a few unrecognizable images) from networks for model training. Our experimental results show that FedD3 significantly outperforms other federated learning frameworks in terms of needed communication volumes, while it provides the additional benefit to be able to balance the trade-off between accuracy and communication cost, depending on usage scenario or target dataset. For instance, for training an AlexNet model on CIFAR-10 with 10 clients under non-independent and identically distributed (Non-IID) setting, FedD3 can either increase the accuracy by over 71% with a similar communication volume, or save 98% of communication volume, while reaching the same accuracy, compared to other one-shot federated learning approaches.

In traditional target tracking methods, the angle error and range error are often measured by the empirical value, while observation noise is a constant. In this paper, the angle error and range error are analyzed. They are influenced by the signal-to-noise ratio (SNR). Therefore, a model related to SNR has been established, in which the SNR information is applied for target tracking. Combined with an advanced nonlinear filter method, the extended Kalman filter method based on the SNR model (SNR-EKF) and the unscented Kalman filter method based on the SNR model (SNR-UKF) are proposed. There is little difference between the SNR-EKF and SNR-UKF methods in position precision, but the SNR-EKF method has advantages in computation time and the SNR-UKF method has advantages in velocity precision. Simulation results show that target tracking methods based on the SNR model can greatly improve the tracking performance compared with traditional tracking methods. The target tracking accuracy and convergence speed of the proposed methods have significant improvements.

We present a practical framework for reproducing omnidirectional incident illumination conditions with complex spectra using a light stage with multispectral LED lights. For lighting acquisition, we augment standard RGB panoramic photography with one or more observations of a color chart with numerous reflectance spectra. We then solve for how to drive the multispectral light sources so that they best reproduce the appearance of the color charts in the original lighting. Even when solving for non-negative intensities, we show that accurate lighting reproduction is achievable using just four or six distinct LED spectra for a wide range of incident illumination spectra. A significant benefit of our approach is that it does not require the use of specialized equipment (other than the light stage) such as monochromators, spectroradiometers, or explicit knowledge of the LED power spectra, camera spectral response functions, or color chart reflectance spectra. We describe two simple devices for multispectral lighting capture, one for slow measurements of detailed angular spectral detail, and one for fast measurements with coarse angular detail. We validate the approach by realistically compositing real subjects into acquired lighting environments, showing accurate matches to how the subject would actually look within the environments, even for those including complex multispectral illumination. We also demonstrate dynamic lighting capture and playback using the technique.