Good news!
"... To tackle the growing energy demands of AI, researchers ... have built something dazzling. Their new chip, called a photonic joint transform correlator (pJTC), swaps electricity for light to handle one of AI's most power-hungry jobs. ..."
"... The prototype chip uses two sets of miniature Fresnel lenses using standard manufacturing processes. These two-dimensional versions of the same lenses found in lighthouses are just a fraction of the width of a human hair. Machine learning data, such as from an image or other pattern-recognition tasks, are converted into laser light on-chip and passed through the lenses. The results are then converted back into a digital signal to complete the AI task.
This lens-based convolution system is not only more computationally efficient, but it also reduces the computing time. Using light instead of electricity has other benefits, too. Sorger’s group designed a chip that could use different colored lasers to process multiple data streams in parallel. ..."
From the abstract:
"Convolutional operations are computationally intensive in artificial intelligence services, and their overhead in electronic hardware limits machine learning scaling. Here, we introduce a photonic joint transform correlator (pJTC) using a near-energy-free on-chip Fourier transformation to accelerate convolution operations. The pJTC reduces computational complexity for both convolution and cross-correlation from O(N4) to O(N2), where N2 is the input data size.
Demonstrating functional Fourier transforms and convolution, this pJTC achieves 98.0% accuracy on an exemplary MNIST inference task.
Furthermore, a wavelength-multiplexed pJTC architecture shows potential for high throughput and energy efficiency, reaching 305 TOPS/W and 40.2 TOPS/mm2, based on currently available foundry processes.
An efficient, compact, and low-latency convolution accelerator promises to advance next-generation AI capabilities across edge demands, high-performance computing, and cloud services."
New light-based chip boosts power efficiency of AI tasks 100 fold (original news release)
New light-based chip boosts power efficiency of AI tasks 100 fold (original news release)
Near-energy-free Photonic Fourier Transformation for Convolution Operation Acceler (preprint, open access)
pJTC Convolutional techniques.
(a) Comparison of spatial convolution, Fourier electrical convolution, and Fourier optical convolution in terms of computational complexity.
(b) Schematic of a JTC, demonstrating how it performs Fourier optical convolution by optically generating the Fourier transform of the combined input Signal and Kernel, detecting the intensity pattern, and producing the auto- and cross-correlation between Signal and Kernel.
(c) Optical microscope image of the fabricated SiPh chiplet from AIM Photonics. (d) Comparison of the initial MNIST image (green line), the output after an ideal Fourier transform (blue line), the output after the actual on-chip lens Fourier transform (yellow line), and the calibrated output obtained from the actual on-chip lens after applying phase correction (pink line). (e) Confusion matrix shows the classification accuracy for 10,000 test MNIST images with 10 percent random temporal delay introduced in the input electrical signal, achieving total accuracy of 95.3 percent.
No comments:
Post a Comment