Training and inference of large-scale machine learning models place diverse demands on compute systems, with increasingly higher requirements for processing power, memory capacity, and memory bandwidth. Current compute systems mainly comprise CUs tightly coupled to a fixed amount of high-bandwidth memory (HBM), which needs to be placed in close proximity to the compute die due to constraints related to signal integrity, energy consumption and limited die periphery. This tight coupling limits the system's ability to flexibly adapt to varying compute and memory demands. In this work, we start by exploring the design space of co-packaged photonic I/Os and identifying compute die's shoreline width as a critical architectural constraint. We then present SiPAM, a silicon photonic accelerated memory pooling architecture that integrates dense wavelength division multiplexed (DWDM) photonic I/Os directly along the compute die's shoreline. By replacing local HBMs and conventional electrical networking interfaces with photonic I/Os, we build an unified high-bandwidth photonic communication domain that enables reconfigurable access to both memory and network resources. This architectural design decouples memory and network scaling from compute packaging constraints and allows the system to flexibly adapt to workload demands. We then develop a quantitative optimization methodology to efficiently allocate compute and memory resources based on workloads and parallelism strategies. Our evaluations show that SiPAM scales effectively with model size and I/O bandwidth, improves compute utilization across GPU and memory generations, and achieves up to 3.5×speedup in end-to-end iteration time and up to 5.46× improvement in system efficiency compared to state-of-the-art Nvidia B100 GPU systems.

While traditional datacenters rely on static, electrically switched fabrics, Optical Circuit Switch (OCS)-enabled reconfigurable networks offer dynamic bandwidth allocation and lower power consumption. This work introduces a quantitative framework for evaluating reconfigurable networks in large-scale AI systems, guiding the adoption of various OCS and link technologies by analyzing trade-offs in reconfiguration latency, link bandwidth provisioning, and OCS placement. Using this framework, we develop two in-workload reconfiguration strategies and propose an OCS-enabled, multi-dimensional all-to-all topology that supports hybrid parallelism with improved energy efficiency. Our evaluation demonstrates that with state-of-the-art per-GPU bandwidth, the optimal in-workload strategy achieves up to 2.3 × improvement over the commonly used one-shot approach when reconfiguration latency is low (<100 μ s). However, with sufficiently high bandwidth, one-shot reconfiguration can achieve comparable performance without requiring in-workload reconfiguration. Additionally, our proposed topology improves performance–power efficiency, achieving up to 1.75 × better trade-offs than Fat-Tree and 3D-Torus–based OCS network architectures.

Partial-label learning is a kind of weakly supervised learning with inexact labels, where for each training example, we are given a set of candidate labels instead of only one true label. Recently, various approaches on partial-label learning have been proposed under different generation models of candidate label sets. However, these methods require relatively strong distributional assumptions on the generation models. When the assumptions do not hold, the performance of the methods is not guaranteed theoretically. In this letter, we propose the notion of properness on partial labels. We show that this proper partial-label learning framework requires a weaker distributional assumption and includes many previous partial-label learning settings as special cases. We then derive a unified unbiased estimator of the classification risk. We prove that our estimator is risk consistent, and we also establish an estimation error bound. Finally, we validate the effectiveness of our algorithm through experiments.

Beam hopping technology can flexibly allocate system resources to achieve on-demand coverage in satellite communication system. However, most of the current research mainly focused on the improvement of system capacity while ignoring the optimization of user delay and quality of service (QoS). In this paper, we propose a maximal user service weight gain (maxUSWG) resource allocation algorithm in beam hopping satellite communication system. Specifically, the user service weight gain is determined by the current traffic demand and delay sensitivity of the cell. We propose to use the combination weights method for multiple attribute decision making based on maximizing deviations to calculate the weight of traffic demand and delay sensitivity. After the weighted decision attributes are obtained, the user service weight gain can be combined and evaluated. In order to maximize the user service weight gain, we allocate resources to the cell with larger weight gain until the power is exhausted or the number of working beams reaches a certain threshold. The simulation results show that the algorithm maxUSWG can effectively improve the QoS, increase the system throughput and improve the resource utilization.

This paper provides a mathematical model of data center performance based on the recently introduced Quantitative Theory of Bottleneck Structures (QTBS). Using the model, we prove that if the traffic pattern is \textit{interference-free}, there exists a unique optimal design that both minimizes maximum flow completion time and yields maximal system-wide throughput. We show that interference-free patterns correspond to the important set of patterns that display data locality properties and use these theoretical insights to study three widely used interconnects---fat-trees, folded-Clos and dragonfly topologies. We derive equations that describe the optimal design for each interconnect as a function of the traffic pattern. Our model predicts, for example, that a 3-level folded-Clos interconnect with radix 24 that routes 10\% of the traffic through the spine links can reduce the number of switches and cabling at the core layer by 25\% without any performance penalty. We present experiments using production TCP/IP code to empirically validate the results and provide tables for network designers to identify optimal designs as a function of the size of the interconnect and traffic pattern.

We study the problem of learning from aggregate observations where supervision signals are given to sets of instances instead of individual instances, while the goal is still to predict labels of unseen individuals. A well-known example is multiple instance learning (MIL). In this paper, we extend MIL beyond binary classification to other problems such as multiclass classification and regression. We present a general probabilistic framework that accommodates a variety of aggregate observations, e.g., pairwise similarity/triplet comparison for classification and mean/difference/rank observation for regression. Simple maximum likelihood solutions can be applied to various differentiable models such as deep neural networks and gradient boosting machines. Moreover, we develop the concept of consistency up to an equivalence relation to characterize our estimator and show that it has nice convergence properties under mild assumptions. Experiments on three problem settings --- classification via triplet comparison and regression via mean/rank observation indicate the effectiveness of the proposed method.