A naukri.com initiative
Medium
7d
33
An artificial memory with 11 quantum states based on spin that mimics a human synapse is created
- A new artificial memory based on spintronics, with 11 stable states, has been created by researchers at the National Taiwan University and published in Advanced Science.
- This memory mimics a human synapse and is controlled solely by electric current, making it more energy-efficient and conducive to brain-like computing.
- The memory is designed to bridge the gap between processing and memory, mirroring the simultaneous information processing and storage in the human brain.
- The device operates using tilted magnetic anisotropy, allowing for precise control of electron spin and simulating synaptic enhancement and depression.
- The memory states are read using the anomalous Hall effect, with a minimal inter-cycle variation of only 2% for reliable distinction between the 11 states.
- The device's performance was tested in a convolutional neural network for image classification tasks, achieving an accuracy of 81.51% post-training quantization.
- This close performance alignment with digital models showcases the potential of physical devices in implementing artificial intelligence in neuromorphic hardware.
- By utilizing spin as the basis of memory operation, the device offers precise control, energy efficiency, high information density, and good compatibility with semiconductor processes.
- Challenges remain in integrating the memory into functional systems and improving the readout methods for better performance in large-scale applications.
- Overall, this research demonstrates the transformative potential of quantum physics and spintronics in advancing artificial intelligence and brain-inspired computing.
Read Full Article
2 Likes
For uninterrupted reading, download the app