Searcharxiv⌕ Search

arXiv subjects

Xian Xiao

Publications and source records attributed to Xian Xiao.

10 recordsLinked to original sources
arXiv2026

IMPA-Net: Meteorology-Aware Multi-Scale Attention and Dynamic Loss for Extreme Convective Radar Nowcasting

Short-range prediction of convective precipitation from weather radar observations is essential for severe weather warnings. However, deep learning models trained with pixel-wise error metrics tend to produce overly smooth forecasts that suppress intense echoes critical for hazard detection. This issue is exacerbated by insufficient multi-scale feature interaction and suboptimal fusion of heterogeneous geophysical inputs. We propose IMPA-Net (Integrated Multi-scale Predictive Attention Network), a deterministic 0-2 hour nowcasting framework that addresses these limitations through meteorologically-informed designs at the input, architecture, and loss function levels. A parameter-free Spatial Mixer reorganizes heterogeneous input channels at the mesoscale-$γ$ neighborhood (~2 km) via deterministic channel permutation, providing a structured cross-field prior. An integrated multi-scale predictive attention module serves as the spatiotemporal translator, capturing dynamics from mesoscale-$β$ to mesoscale-$γ$ scales. A Meteorologically-Aware Dynamic Loss employs three-level asymmetric weighting -- adapting across training epochs, storm intensity, and forecast lead time -- to counteract regression-to-the-mean. Evaluated against seven baselines on a multi-source radar dataset over eastern China, IMPA-Net raises the Heidke Skill Score at $\geq$45 dBZ from 0.049 (SimVP baseline) to 0.143 under matched settings. Relative to pySTEPS, it provides a better trade-off between severe-event detection and false-alarm control. Spectral analysis confirms preserved energy across mesoscale bands where competing methods show progressive smoothing. These improvements are shown within a single domain and convective regime; generalizability to other orographic and climatic regions remains to be tested.

cs.LG↗
arXiv2026

Scalable Back-Propagation-Free Training of Optical Physics-Informed Neural Networks

Physics intelligence and digital twins often require rapid and repeated performance evaluation of various engineering systems (e.g. robots, autonomous vehicles, semiconductor chips) to enable (almost) real-time actions or decision making. This has motivated the development of accelerated partial differential equation (PDE) solvers, in resource-constrained scenarios if the PDE solvers are to be deployed on the edge. Physics-informed neural networks (PINNs) have shown promise in solving high-dimensional PDEs, but the training time on state-of-the-art digital hardware (e.g., GPUs) is still orders-of-magnitude longer than the latency required for enabling real-time decision making. Photonic computing offers a potential solution to address this huge latency gap because of its ultra-high operation speed. However, the lack of photonic memory and the large device sizes prevent training real-size PINNs on photonic chips. This paper proposes a completely back-propagation-free (BP-free) and highly salable framework for training real-size PINNs on silicon photonic platforms. Our approach involves three key innovations: (1) a sparse-grid Stein derivative estimator to avoid the BP in the loss evaluation of a PINN, (2) a dimension-reduced zeroth-order optimization via tensor-train decomposition to achieve better scalability and convergence in BP-free training, and (3) a scalable on-chip photonic PINN training accelerator design using photonic tensor cores. We validate our numerical methods on both low- and high-dimensional PDE benchmarks. Through pre-silicon simulation based on real device parameters, we further demonstrate the significant performance benefit (e.g., real-time training, huge chip area reduction) of our photonic accelerator.

cs.LG↗
arXiv2024

Hairmony: Fairness-aware hairstyle classification

We present a method for prediction of a person's hairstyle from a single image. Despite growing use cases in user digitization and enrollment for virtual experiences, available methods are limited, particularly in the range of hairstyles they can capture. Human hair is extremely diverse and lacks any universally accepted description or categorization, making this a challenging task. Most current methods rely on parametric models of hair at a strand level. These approaches, while very promising, are not yet able to represent short, frizzy, coily hair and gathered hairstyles. We instead choose a classification approach which can represent the diversity of hairstyles required for a truly robust and inclusive system. Previous classification approaches have been restricted by poorly labeled data that lacks diversity, imposing constraints on the usefulness of any resulting enrollment system. We use only synthetic data to train our models. This allows for explicit control of diversity of hairstyle attributes, hair colors, facial appearance, poses, environments and other parameters. It also produces noise-free ground-truth labels. We introduce a novel hairstyle taxonomy developed in collaboration with a diverse group of domain experts which we use to balance our training data, supervise our model, and directly measure fairness. We annotate our synthetic training data and a real evaluation dataset using this taxonomy and release both to enable comparison of future hairstyle prediction approaches. We employ an architecture based on a pre-trained feature extraction network in order to improve generalization of our method to real data and predict taxonomy attributes as an auxiliary task to improve accuracy. Results show our method to be significantly more robust for challenging hairstyles than recent parametric approaches.

cs.CV↗
arXiv2024

Photonic KAN: a Kolmogorov-Arnold network inspired efficient photonic neuromorphic architecture

Kolmogorov-Arnold Networks (KAN) models were recently proposed and claimed to provide improved parameter scaling and interpretability compared to conventional multilayer perceptron (MLP) models. Inspired by the KAN architecture, we propose the Photonic KAN -- an integrated all-optical neuromorphic platform leveraging highly parametric optical nonlinear transfer functions along KAN edges. In this work, we implement such nonlinearities in the form of cascaded ring-assisted Mach-Zehnder Interferometer (MZI) devices. This innovative design has the potential to address key limitations of current photonic neural networks. In our test cases, the Photonic KAN showcases enhanced parameter scaling and interpretability compared to existing photonic neural networks. The photonic KAN achieves approximately 65$\times$ reduction in energy consumption and area, alongside a 50$\times$ reduction in latency compared to previous MZI-based photonic accelerators with similar performance for function fitting task. This breakthrough presents a promising new avenue for expanding the scalability and efficiency of neuromorphic hardware platforms.

physics.optics↗
arXiv2024

Real-Time FJ/MAC PDE Solvers via Tensorized, Back-Propagation-Free Optical PINN Training

Solving partial differential equations (PDEs) numerically often requires huge computing time, energy cost, and hardware resources in practical applications. This has limited their applications in many scenarios (e.g., autonomous systems, supersonic flows) that have a limited energy budget and require near real-time response. Leveraging optical computing, this paper develops an on-chip training framework for physics-informed neural networks (PINNs), aiming to solve high-dimensional PDEs with fJ/MAC photonic power consumption and ultra-low latency. Despite the ultra-high speed of optical neural networks, training a PINN on an optical chip is hard due to (1) the large size of photonic devices, and (2) the lack of scalable optical memory devices to store the intermediate results of back-propagation (BP). To enable realistic optical PINN training, this paper presents a scalable method to avoid the BP process. We also employ a tensor-compressed approach to improve the convergence and scalability of our optical PINN training. This training framework is designed with tensorized optical neural networks (TONN) for scalable inference acceleration and MZI phase-domain tuning for \textit{in-situ} optimization. Our simulation results of a 20-dim HJB PDE show that our photonic accelerator can reduce the number of MZIs by a factor of $1.17\times 10^3$, with only $1.36$ J and $1.15$ s to solve this equation. This is the first real-size optical PINN training framework that can be applied to solve high-dimensional PDEs.

cs.LG↗
arXiv2023

QoS-aware User Association and Transmission Scheduling for Millimeter-Wave Train-ground Communications

With the development of wireless communication, people have put forward higher requirements for train-ground communications in the high-speed railway (HSR) scenarios. With the help of mobile relays (MRs) installed on the roof of the train, the application of Millimeter-Wave (mm-wave) communication which has rich spectrum resources to the train-ground communication system can realize high data rate, so as to meet users' increasing demand for broad-band multimedia access. Also, full-duplex (FD) technology can theoretically double the spectral efficiency. In this paper, we formulate the user association and transmission scheduling problem in the mm-wave train-ground communication system with MR operating in the FD mode as a nonlinear programming problem. In order to maximize the system throughput and the number of users meeting quality of service (QoS) requirements, we propose an algorithm based on coalition game to solve the challenging NP-hard problem, and also prove the convergence and Nash-stable structure of the proposed algorithm. Extensive simulation results demonstrate that the proposed coalition game based algorithm can effectively improve the system throughput and meet the QoS requirements of as many users as possible, so that the communication system has a certain QoS awareness.

cs.IT↗
arXiv2023

Tensorized Optical Multimodal Fusion Network

We propose the first tensorized optical multimodal fusion network architecture with a self-attention mechanism and low-rank tensor fusion. Simulation results show $51.3 \times$ less hardware requirement and $3.7\times 10^{13}$ MAC/J energy efficiency.

eess.SP↗
arXiv2021

Review Of Integrated Photonic Elastic WDM Switches For Data Centers

In this review paper, we present an elaborate discussion on wavelength selective switches and their demonstrations. We also review packaging and electronic photonic integration of switches; a topic neglected in other review papers. We also cover wavelength locking which is paramount in switching networks with many tunable filters.

cs.ET↗
arXiv2021

Izhikevich-Inspired Optoelectronic Neurons with Excitatory and Inhibitory Inputs for Energy-Efficient Photonic Spiking Neural Networks

We designed, prototyped, and experimentally demonstrated, for the first time to our knowledge, an optoelectronic spiking neuron inspired by the Izhikevich model incorporating both excitatory and inhibitory optical spiking inputs and producing optical spiking outputs accordingly. The optoelectronic neurons consist of three transistors acting as electrical spiking circuits, a vertical-cavity surface-emitting laser (VCSEL) for optical spiking outputs, and two photodetectors for excitatory and inhibitory optical spiking inputs. Additional inclusion of capacitors and resistors complete the Izhikevich-inspired optoelectronic neurons, which receive excitatory and inhibitory optical spikes as inputs from other optoelectronic neurons. We developed a detailed optoelectronic neuron model in Verilog-A and simulated the circuit-level operation of various cases with excitatory input and inhibitory input signals. The experimental results closely resemble the simulated results and demonstrate how the excitatory inputs trigger the optical spiking outputs while the inhibitory inputs suppress the outputs. Utilizing the simulated neuron model, we conducted simulations using fully connected (FC) and convolutional neural networks (CNN). The simulation results using MNIST handwritten digits recognition show 90% accuracy on unsupervised learning and 97% accuracy on a supervised modified FC neural network. We further designed a nanoscale optoelectronic neuron utilizing quantum impedance conversion where a 200 aJ/spike input can trigger the output from on-chip nanolasers with 10 fJ/spike. The nanoscale neuron can support a fanout of ~80 or overcome 19 dB excess optical loss while running at 10 GSpikes/second in the neural network, which corresponds to 100x throughput and 1000x energy-efficiency improvement compared to state-of-art electrical neuromorphic hardware such as Loihi and NeuroGrid.

cs.NE↗