The researchers at the Technical University of Denmark and the University of Rennes have made advances in advanced computing, specifically focusing on in-memory computing and photonics. The researchers have demonstrated the use of time and wavelength division multiplexing (WDM) to solve four independent tasks at the same time in a single photonic chip. This approach can have a paradigm shift in multi-tasking complex assignments via parallel computing.
Advances in core areas like artificial intelligence, complex climate change or drug discovery simulations, and data-intensive tasks like scientific research have urged the demand for advanced computing. For quite a time now, scientists have been focusing on using light (photons) for data transmission, which makes processing 1000x faster.
While Von Neumann architecture or the stored-program concept, is the foundational design of modern computers, it has limitations. Due to the Von Neumann bottleneck limiting the data transfer speed between the CPU and memory, the study has highlighted the need for alternative architectures.
The Wavelength Division Multiplexing system is based on Time-Delay Reservoir Computing (TDRC) that uses a Silicon Microring Resonator (MRR) to solve multiple tasks simultaneously on a single photonic chip. The tasks include Time-series prediction, waveform signal classification, wireless channel equalization, and radar signal prediction. Testing showed that the system could efficiently handle up to 10 simultaneous instances of the same task, with excellent performance.
The authors stress that the system size and its complexity could be reduced significantly by using time and wavelength division multiplexing.
Related:
- Scaled-up LLMs are more prone to sensible yet wrong answers
- This 5D Memory Crystal can store 360 TB of data for billions of years
“By adjusting the input power and frequency of each optical channel, we can achieve levels of performance for each of the tasks that are comparable to those quoted in state-of-the-art reports focusing on single-task operation,” says the study.
“We also quantify the memory capacity and nonlinearity of each parallelized RC and relate these properties to the performance of each task. Finally, we provide insight into the impact of the feedback mechanism on the performance of the system.”
The final version of the study is yet to be published.
Journal Reference:
Giron Castro, B. J., Peucheret, C., Zibar, D., & Da Ros, F. Multi-task Photonic Reservoir Computing: Wavelength Division Multiplexing for Parallel Computing with a Silicon Microring Resonator. Advanced Optical Technologies, 13, 1471239. DOI: https://doi.org/10.3389/aot.2024.1471239