Tensor processing supports most artificial intelligence systems, requiring arithmetic beyond regular math. Unlike electronic processors that handle such operations sequentially, the new optical method encodes digital data in the amplitude and phase of light waves.
During the process, light fields interact to carry out core operations of deep learning, such as matrix and tensor multiplications. Multiple wavelengths can be used to extend computing to higher order tensors.
"Imagine you are a customs officer who must inspect every parcel through multiple machines with different functions and then sort them into the right bins," said Dr. Yufeng Zhang from Aalto University's Photonics Group. "Normally, you would process each parcel one by one. Our optical computing method merges all parcels and all machines together - we create multiple optical hooks that connect each input to its correct output. With just one operation, one pass of light, all inspections and sorting happen instantly and in parallel."
The system does not require active electronic switching since the computations occur naturally as light propagates. Professor Zhipei Sun, head of the Photonics Group, stated, "This approach can be implemented on almost any optical platform."
The researchers plan to integrate this method directly into photonic chips to allow low-power AI processing. Dr. Zhang estimates industry-scale adoption for this technique within three to five years.
Research Report:Direct tensor processing with coherent light
Related Links
Aalto University
Computer Chip Architecture, Technology and Manufacture
Nano Technology News From SpaceMart.com
| Subscribe Free To Our Daily Newsletters |
| Subscribe Free To Our Daily Newsletters |
