23th September 2025 (13 Topics)
1
Sep
2
Sep
3
Sep
4
Sep
5
Sep
6
Sep
8
Sep
9
Sep
10
Sep
11
Sep
12
Sep
13
Sep
15
Sep
16
Sep
17
Sep
18
Sep
19
Sep
20
Sep
22
Sep
23
Sep
24
Sep
Optical Computing for AI
Context:
Researchers have demonstrated the use of optical fibres to perform AI tasks faster and more efficiently than conventional electronic computers.
Optical Computing:
- Uses photons instead of electrons to process data, enabling higher speed, bandwidth, and energy efficiency.
Nonlinear Regime of Light:
- High-intensity light pulses in optical fibres create interactions that can perform complex transformations useful for AI computations.
Extreme Learning Machine (ELM):
- A simple neural network with a single hidden layer; output weights are trained in one step. Optical fibres serve as the hidden layer by transforming input data (images) into light pulse fingerprints.
Process Overview:
- Input image downsized and encoded onto light pulses (phase/amplitude at different frequencies).
- Light pulses travel through optical fibres, exploiting nonlinear effects and dispersion.
- Resulting light spectrum (“fingerprint”) acts as the hidden layer input for ELM classification.
Performance:
- Achieved >91% accuracy in the anomalous dispersion regime and >93% in the normal dispersion regime, comparable to conventional ELMs.
Significance:
- Demonstrates potential for faster, energy-efficient AI hardware, enabling optical neural networks and photonic integrated circuits in the future.
Limitations:
- Real-world effects such as light polarization not fully utilized; future research could enhance performance.
More Articles
Courses
Study
Notes
Home
Resources
More