Skip to content

Latest commit

 

History

History
7 lines (5 loc) · 784 Bytes

README.md

File metadata and controls

7 lines (5 loc) · 784 Bytes

Cluster of FPAAs to recognize images using neural networks

This software manages a cluster of 40 Anadigm FPAAs placed on 10 Quad Apex V2.0 development boards, allowing the prototyping of analog applications using FPAAs currently commercially available FPAAs prior to being implemented through more advanced FPAA devices or custom IC design, using the Dynamic Reconfiguration capability of the FPAAs.

Currently. this cluster classifies 28×28 MNIST images through a 19-8-6-4 Analog NN with 19-input metrics based on DCT compression.

Acknowledgements

This work has been supported by MICINN under grant PID2021-123041OB-I00, the CM under grant S2018/TCS-4423, the EU (FEDER) and the Spanish MINECO under grant RTI2018-093684-B-I00 and by Banco Santander under grant PR26/16-20B-1.