Fujitsu Corporation

Liuan Wang

Jun Sun

Wei Fan

Rong Zhou

A method and apparatus of accelerating deep learning, and a deep neural network are provided. The method comprises: randomly initializing weights and biases of a deep neural network as n-bit fixed-point numbers; reducing data in a plurality of layers in the deep neural network that have calculation amounts are greater than a first predetermined threshold as m-bit fixed-point numbers, where m and n are integers and m&#x3c;n, and maintaining data in remaining layers among the plurality of layers as n-bit fixed-point numbers; and training the deep neural network after the reducing, until convergence.

Method and apparatus for accelerating deep learning and deep neural network

A method and a device for simulating atomic dynamics includes setting initial positions for multiple specific atoms in a specific scene; calculating, based on the initial positions, positions of the multiple specific atoms at each time in a first time series by utilizing a Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS) configured with respect to the specific scene, as real positions; calculating, based on the initial positions, positions of the multiple specific atoms at the same time in the first time series by utilizing a generative adversarial network (GAN), as predicted positions; improving a configuration of the GAN based on the real positions and the predicted positions at a same time. Initial positions are settable for multiple atoms to be simulated in a scene; positions of the multiple atoms to be simulated are calculated at each time in a second time series in the scene by utilizing the improved GAN.

Method and device for simulating atomic dynamics

A device and a method for improving a processing speed of a neural network and applications thereof in the neural network where the device includes a processor configured to perform: determining, according to a predetermined processing speed improvement target, a dimension reduction amount of each of one or more parameter matrixes in the neural network obtained through training; preprocessing each parameter matrix based on the dimension reduction amount of the parameter matrix; and retraining the neural network based on a result of the preprocessing to obtain one or more dimension reduced parameter matrixes so as to ensure performance of the neural network meets a predetermined requirement. According to the embodiments of the present disclosure, it is possible to significantly improve the processing speed of the neural network while ensuring the performance of the neural network meets the predetermined requirement.

Device and method for improving processing speed of neural network

A device and a method for electromagnetic field simulation are provided. The image processing device is to obtain, with a first resolution, a first electromagnetic field simulation result of a simulation region, and to derive a second electromagnetic field simulation result with a second resolution, according to the first electromagnetic field simulation result, using a model trained based on a deep learning method. The second resolution is higher than the first resolution.

Growing community of inventors

Beijing, China

Liuan Wang