Dataset Information

Cross-layer transmission realized by light-emitting memristor for constructing ultra-deep neural network with transfer learning ability.

ABSTRACT: Deep neural networks have revolutionized several domains, including autonomous driving, cancer detection, and drug design, and are the foundation for massive artificial intelligence models. However, hardware neural network reports still mainly focus on shallow networks (2 to 5 layers). Implementing deep neural networks in hardware is challenging due to the layer-by-layer structure, resulting in long training times, signal interference, and low accuracy due to gradient explosion/vanishing. Here, we utilize negative ultraviolet photoconductive light-emitting memristors with intrinsic parallelism and hardware-software co-design to achieve electrical information's optical cross-layer transmission. We propose a hybrid ultra-deep photoelectric neural network and an ultra-deep super-resolution reconstruction neural network using light-emitting memristors and cross-layer block, expanding the networks to 54 and 135 layers, respectively. Further, two networks enable transfer learning, approaching or surpassing software-designed networks in multi-dataset recognition and high-resolution restoration tasks. These proposed strategies show great potential for high-precision multifunctional hardware neural networks and edge artificial intelligence.

SUBMITTER: Chen Z

PROVIDER: S-EPMC10908859 | biostudies-literature | 2024 Mar

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Cross-layer transmission realized by light-emitting memristor for constructing ultra-deep neural network with transfer learning ability.

Chen Zhenjia Z Lin Zhenyuan Z Yang Ji J Chen Cong C Liu Di D Shan Liuting L Hu Yuanyuan Y Guo Tailiang T Chen Huipeng H

Nature communications 20240302 1

Deep neural networks have revolutionized several domains, including autonomous driving, cancer detection, and drug design, and are the foundation for massive artificial intelligence models. However, hardware neural network reports still mainly focus on shallow networks (2 to 5 layers). Implementing deep neural networks in hardware is challenging due to the layer-by-layer structure, resulting in long training times, signal interference, and low accuracy due to gradient explosion/vanishing. Here, ...[more]

PMID: 38431669

Similar Datasets

Project description:Light colors may affect poultry behaviors, well-being and performance. However, preferences of layer pullets for light colors are not fully understood. This study was conducted to investigate the pullet preferences for four light-emitting diode colors, including white, red, green and blue, in a lighting preference test system. The system contained four identical compartments each provided with a respective light color. The pullets were able to move freely between the adjacent compartments. A total of three groups of 20 Chinese domestic Jingfen layer pullets (54 to 82 days of age) were used for the test. Pullet behaviors were continuously recorded and summarized for each light color/compartment into daily time spent (DTS), daily percentage of time spent (DPTS), daily times of visit (DTV), duration per visit, daily feed intake (DFI), daily feeding time (DFT), feeding rate (FR), distribution of pullet occupancy and hourly time spent. The results showed that the DTS (h/pullet·per day) were 3.9±0.4 under white, 1.4±0.3 under red, 2.2±0.3 under green and 4.5±0.4 under blue light, respectively. The DTS corresponded to 11.7% to 37.6% DPTS in 12-h lighting periods. The DTV (times/pullet·per day) were 84±5 under white, 48±10 under red, 88±10 under green and 94±8 under blue light. Each visit lasted 1.5 to 3.2 min. The DFI (g/pullet·per day) were 27.6±1.7 under white, 7.1±1.6 under red, 15.1±1.1 under green and 23.1±2.0 under blue light. The DFT was 0.18 to 0.65 h/pullet·per day and the FR was 0.57 to 0.75 g/min. For most of the time during the lighting periods, six to 10 birds stayed under white, and one to five birds stayed under red, green and blue light. Pullets preferred to stay under blue light when the light was on and under white light 4 h before the light off. Overall, pullets preferred blue light the most and red light the least. These findings substantiate the preferences of layer pullets for light colors, providing insights for use in the management of light-emitting diode colors to meet pullet needs.

Dataset Information

Cross-layer transmission realized by light-emitting memristor for constructing ultra-deep neural network with transfer learning ability.

Publications

Cross-layer transmission realized by light-emitting memristor for constructing ultra-deep neural network with transfer learning ability.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets