Related Experiment Videos
Spike timing dependent plasticity (STDP) can ameliorate process variations in neuromorphic VLSI
Katherine Cameron1, Vasin Boonsobhak, Alan Murray
1School of Engineering and Electronics, The University of Edinburgh, UK. K.L.Cameron@sms.ed.ac.uk
IEEE Transactions on Neural Networks
|December 14, 2005
Summary
This study introduces a novel transient-detecting very large scale integration (VLSI) pixel designed for visual processing. The adaptive system effectively minimizes variations in VLSI processes for spike-timing-based algorithms.
Area of Science:
- Computer Engineering
- Neuromorphic Engineering
- Artificial Intelligence
Background:
- Visual processing algorithms often rely on spike timing.
- Variations in Very Large Scale Integration (VLSI) manufacturing processes can impact algorithm performance.
- Spike-timing dependent plasticity (STDP) is a biologically inspired learning rule used in adaptive systems.
Purpose of the Study:
- To describe a transient-detecting VLSI pixel suitable for depth-recovery algorithms.
- To develop an adaptive system using STDP to mitigate VLSI process variations.
- To demonstrate the effectiveness of the adaptive system in maintaining algorithm performance.
Main Methods:
- Designed and implemented a transient-detecting VLSI pixel array.
- Coupled the pixel array to an adaptive system based on STDP.
- Utilized 0.35 microm CMOS technology for temporal differentiating pixels and STDP circuits.
- Evaluated the system's ability to adapt to process variations.
Main Results:
- The developed VLSI pixel and STDP system successfully adapted to process variations.
- The adaptation significantly reduced the impact of manufacturing inconsistencies on algorithm performance.
- The system demonstrated robustness without interrupting the core processing functions.
Conclusions:
- The transient-detecting VLSI pixel and adaptive STDP system effectively compensates for process variations.
- This approach is applicable to a wide range of spike-timing driven processing algorithms in VLSI.
- The findings pave the way for more reliable neuromorphic computing hardware.