11 May 2022 in Future Generation Computer Systems
18 April 2022 in IEEE Transactions on Instrumentation and Measurement
In order to further improve the output voltage amplitude and repetitive operation reliability of nanosecond pulse generator (PG) based on avalanche transistor (AT), the method of power synthesis of modularized Marx circuits with transmission line transformer (TLT) is employed in this article. The modified positive $4\times11$ -stage Marx circuit introducing base-triggering method is designed and implemented to improve the operation reliability. The triggering synchronization and the consistence of output characteristics of multiple Marx modules with an identical trigger pulse are validated. The feasibility of combining pulses generated from four Marx modules with an optimized four-stage TLT is demonstrated. The influences of the type and quantity of magnetic core and the superimposed stage number of TLT on the output performance of the generator are investigated. The operation characteristics of the generator including output parameters, trigger time-delay, and loop efficiency in the whole working range are evaluated. The experimental results show that the working range of the overall device has been widened significantly with the injection of additional initial carrier. Accordingly, the generator prototype is capable of achieving the output performance with an adjustable voltage amplitude in the range of 7.8–26.7 kV on a matched 300- $\Omega $ resistive load and 12.2–38.9 kV on the high-resistance load, respectively, a 10%–90% rise time of 3.6 ns, and a full-width at half-maximum (FWHM) of 12.2 ns. A long-term test with the maximum repetition rate of 3 kHz is conducted to realize high-voltage output and high-frequency operation simultaneously.
20 October 2021 in CCF Transactions on High Performance Computing
6 January 2025 in IEEE Transactions on Fuzzy Systems
Graph neural networks (GNNs) have been widely used in many fields. Inductive learning has replaced transductive learning as the current mainstream paradigm for GNN training due to its higher memory efficiency, computing speed, and stronger generalization. Neighbor node sampling as a key step in GNN inductive learning is crucial to the model performance. However, existing samplers only focus on how to sample nodes from the adjacency matrix, ignoring the fact that different neighbors have different impacts on the target node at different moments. They usually aggregate the neighbor information in a simple way such as averaging or summing, which limits the information representation, robustness, and generalization. In order to address the shortcomings of existing graph inductive learning samplers, this article proposes a dynamic fuzzy sampler (DFS) based on a Gaussian fuzzy system. The DFS fully takes into account the diversity of nodes in the graph-structured data, and efficiently models and handles the uncertainties and fuzziness of the mutual information of various nodes at different moments. Specifically, DFS first innovatively constructs a learnable Gaussian fuzzy set system for determining the membership degree of different neighbors to the target node at different moments. Subsequently, DFS aggregates the target node embeddings and membership-weighted neighbor embeddings to update the target node's features, which makes the target node utilize the sampling information more effectively. The aggregated target node effectively captures the graph structure information and neighbor node information, which can facilitate the subsequent GNN-based graph representation model with stronger representation and generalization capabilities. Our supervised and self-supervised experimental results on graph datasets of different sizes show that DFS has consistently excellent performance, significantly outperforming other state-of-the-art sampling schemes. DFS achieves up to 1.90% and 9.52% F1-score improvement compared to the state-of-the-art schemes on small- and large-scale graphs, respectively.
12 October 2021 in Scientific Reports
Due to the increase in computing power, it is possible to improve the feature extraction and data fitting capabilities of DNN networks by increasing their depth and model complexity. However, the big data and complex models greatly increase the training overhead of DNN, so accelerating their training process becomes a key task. The Tianhe-3 peak speed is designed to target E-class, and the huge computing power provides a potential opportunity for DNN training. We implement and extend LeNet, AlexNet, VGG, and ResNet model training for a single MT-2000+ and FT-2000+ compute nodes, as well as extended multi-node clusters, and propose an improved gradient synchronization process for Dynamic Allreduce communication optimization strategy for the gradient synchronization process base on the ARM architecture features of the Tianhe-3 prototype, providing experimental data and theoretical basis for further enhancing and improving the performance of the Tianhe-3 prototype in large-scale distributed training of neural networks.
11 February 2022 in Future Generation Computer Systems