Search for: All records

With the ever-growing popularity of Graph Neural Networks (GNNs), efficient GNN inference is gaining tremendous attention. Field-Programmable Gate Arrays (FPGAs) are a promising execution platform due to their fine-grained parallelism, low power consumption, reconfigurability, and concurrent execution. Even better, High-Level Synthesis (HLS) tools help bridge the gap between the non-trivial FPGA development efforts and rapid emergence of new GNN models. To enable investigation into how effectively modern HLS tools can accelerate GNN inference, we present GNNHLS, a benchmark suite containing a software stack for data generation and baseline deployment and FPGA implementations of 6 well-tuned GNN HLS kernels. 

A double-edged sword in two-dimensional material science and technology is optically forbidden dark exciton. On the one hand, it is fascinating for condensed matter physics, quantum information processing, and optoelectronics due to its long lifetime. On the other hand, it is notorious for being optically inaccessible from both excitation and detection standpoints. Here, we provide an efficient and low-loss solution to the dilemma by reintroducing photonics bound states in the continuum (BICs) to manipulate dark excitons in the momentum space. In a monolayer tungsten diselenide under normal incidence, we demonstrated a giant enhancement (~1400) for dark excitons enabled by transverse magnetic BICs with intrinsic out-of-plane electric fields. By further employing widely tunable Friedrich-Wintgen BICs, we demonstrated highly directional emission from the dark excitons with a divergence angle of merely 7°. We found that the directional emission is coherent at room temperature, unambiguously shown in polarization analyses and interference measurements. Therefore, the BICs reintroduced as a momentum-space photonic environment could be an intriguing platform to reshape and redefine light-matter interactions in nearby quantum materials, such as low-dimensional materials, otherwise challenging or even impossible to achieve.

 

This paper presents a fabric defect detection network (FabricNet) for automatic fabric defect detection. Our proposed FabricNet incorporates several effective techniques, such as Feature Pyramid Network (FPN), Deformable Convolution (DC) network, and Distance IoU Loss function, into vanilla Faster RCNN to improve the accuracy and speed of fabric defect detection. Our experiment shows that, when optimizations are combined, the FabricNet achieves 62.07% mAP and 97.37% AP50 on DAGM 2007 dataset, and an average prediction speed of 17 frames per second. 

Graph structured data are abundant in the real world. Among different graph types, directed acyclic graphs (DAGs) are of particular interest to machine learning researchers, as many machine learning models are realized as computations on DAGs, including neural networks and Bayesian networks. In this paper, we study deep generative models for DAGs, and propose a novel DAG variational autoencoder (D-VAE). To encode DAGs into the latent space, we leverage graph neural networks. We propose an asynchronous message passing scheme that allows encoding the computations on DAGs, rather than using existing simultaneous message passing schemes to encode local graph structures. We demonstrate the effectiveness of our proposed DVAE through two tasks: neural architecture search and Bayesian network structure learning. Experiments show that our model not only generates novel and valid DAGs, but also produces a smooth latent space that facilitates searching for DAGs with better performance through Bayesian optimization.