Utilizing optical neural network to establish high-performance OR and XOR logic gates. (March 2023)
- Record Type:
- Journal Article
- Title:
- Utilizing optical neural network to establish high-performance OR and XOR logic gates. (March 2023)
- Main Title:
- Utilizing optical neural network to establish high-performance OR and XOR logic gates
- Authors:
- Lin, Chu-En
Sun, Ching-Pao
Chen, Chii-Chang - Abstract:
- Abstract: The optical–neural-network logic gates using unsupervised learning method and supervised learning method are investigated. The structures of the optical neurons using self-connection configuration and interconnection configuration are proposed. The performance of the AND, OR, NAND, NOR and XOR logic gates are analyzed. According to our simulation results, the bit error ratio (BER) of the optical neurons using the interconnection configuration is lower than that using self-connection configuration. For OR logic gate, the best performance is BER = 6.54%. For XOR logic gate, the best performance is BER < 4.89 × 10 −5 . The results show that the proposed optical structure can work for different logic gates by tuning the parameters of the couplers and the phase shifters. Highlights: The optical structures of our optical neural networks with interconnection and self-connection configurations are proposed. The synergy of two neurons which process the input data together is better than the neurons working independently. The OR and XOR logic gates, which are trained by unsupervised learning method are designed using the same optical structure with different coupling ratios of couples and different phase delays of phase modulators. These two logic gates can be fabricated with the same manufacturing process. Since the optical setup can achieve OR and XOR logic devices, the manufacturing process and the cost of this optical integrated circuit could be simpler and cheaperAbstract: The optical–neural-network logic gates using unsupervised learning method and supervised learning method are investigated. The structures of the optical neurons using self-connection configuration and interconnection configuration are proposed. The performance of the AND, OR, NAND, NOR and XOR logic gates are analyzed. According to our simulation results, the bit error ratio (BER) of the optical neurons using the interconnection configuration is lower than that using self-connection configuration. For OR logic gate, the best performance is BER = 6.54%. For XOR logic gate, the best performance is BER < 4.89 × 10 −5 . The results show that the proposed optical structure can work for different logic gates by tuning the parameters of the couplers and the phase shifters. Highlights: The optical structures of our optical neural networks with interconnection and self-connection configurations are proposed. The synergy of two neurons which process the input data together is better than the neurons working independently. The OR and XOR logic gates, which are trained by unsupervised learning method are designed using the same optical structure with different coupling ratios of couples and different phase delays of phase modulators. These two logic gates can be fabricated with the same manufacturing process. Since the optical setup can achieve OR and XOR logic devices, the manufacturing process and the cost of this optical integrated circuit could be simpler and cheaper than electric circuits. Moreover, the computing speed of our optical logic gates could be higher than that of electronic circuits. … (more)
- Is Part Of:
- Engineering applications of artificial intelligence. Volume 119(2023)
- Journal:
- Engineering applications of artificial intelligence
- Issue:
- Volume 119(2023)
- Issue Display:
- Volume 119, Issue 2023 (2023)
- Year:
- 2023
- Volume:
- 119
- Issue:
- 2023
- Issue Sort Value:
- 2023-0119-2023-0000
- Page Start:
- Page End:
- Publication Date:
- 2023-03
- Subjects:
- Unsupervised learning method -- Supervised learning method -- Optical neural networks -- Reservoir computing -- Integrated optical device
Engineering -- Data processing -- Periodicals
Artificial intelligence -- Periodicals
Expert systems (Computer science) -- Periodicals
Ingénierie -- Informatique -- Périodiques
Intelligence artificielle -- Périodiques
Systèmes experts (Informatique) -- Périodiques
Artificial intelligence
Engineering -- Data processing
Expert systems (Computer science)
Periodicals
620.00285 - Journal URLs:
- http://www.sciencedirect.com/science/journal/09521976 ↗
http://www.elsevier.com/journals ↗ - DOI:
- 10.1016/j.engappai.2022.105788 ↗
- Languages:
- English
- ISSNs:
- 0952-1976
- Deposit Type:
- Legaldeposit
- View Content:
- Available online (eLD content is only available in our Reading Rooms) ↗
- Physical Locations:
- British Library DSC - 3755.704500
British Library DSC - BLDSS-3PM
British Library HMNTS - ELD Digital store - Ingest File:
- 25681.xml