Online Learning in Digital Hardware, Deep Convolutional STDP

STDP is a bio-inspired learning rule that can be implemented locally in neurons (without a global supervisor). This will allow the rule to be implemented in massively parallel platforms without extra overhead. In this project, we start from understanding the Convolutional STDP learning rule and Hierarchical multi-layer version of the rule. Also we will try to understand the digital hardware implementation of simple network with a real-time demo. Then we work together to design a multi-layer version of the Convolutional STDP in hardware and explore the extension of layers.

Login to become a member send

Timetable

No timetable published yet.

These are the refrences:
1-Unsupervised Learning of Digit Recognition Using Spike-Timing-Dependent Plasticity, Peter U. Diehl, Matthew Cook
2-Unsupervised learning of visual features through spike timing dependent plasticity, Timothée Masquelier, Simon J Thorpe
3-STDP-based spiking deep neural networks for object recognition,Saeed Reza Kheradpisheh, Mohammad Ganjtabesh, Simon J Thorpe, Timothée Masquelier
4-Hardware Implementation of Convolutional STDP for On-line Visual Feature Learning, A. Yousefzadeh, T. Masquelier, T. Serrano-Gotarredona, and B. Linares-Barranco (this paper has not published yet and it is attached)
5-Video of Real-time demo in FPGA for unsupervised learning: https://youtu.be/VRz1uLXa4KA

ISCAS_ConvSTDP.pdf (257.0 KB) ISCAS_demo_ConvSTDP.pdf (125.8 KB)

Leader

Members