jasha64/snn-fpga

Spiking Neural Network RTL Implementation

My "Programming CV"

In this repo is my historical code, starting from 2012 (when I first learned programming) and organized chronologically.

For easy maintenance, this repo only contains small programs; my large projects are split out into separate git repos such as jasha64/ARM9-compatible-soft-CPU-core, and are hyperlinked below. Should you be interested, do check them out.

Contents:

2012-2013 Basic C++ programming

2014-2017 NOI

• "NOI" stands for "National Olympiad in Informatics", a series of algorithm competitions for middle school students in China organized by the China Computer Federation (CCF) [Contest website]

• It's my senior high school years. I participated in NOI, won second prize in NOI 2016 Winter Camp, and won a bronze medal in APIO 2016 (Asia-Pacific Informatics Olympiad)

2017-2021 Bachelor in Computer Science

• Course projects
  • Spring 2019, Intro to Computer Systems: Completed an x86 emulator "NEMU" in C, which illustrates x86 CPU behavior and runs x86 programs like GDB does.
  • Spring 2019, Databases: Developed an Online Judge with a MySQL backend, a PHP frontend, and an independent judger program.
  • Fall 2019, Hardware Experiments: Did experiments with battery stack monitor boards, somatosensory controllers, and motor control boards.
  • Fall 2019, Intro to Artificial Intelligence: An expert system, which asks users' requirements and recommend railway transfer schemes for them
  • Computer Architecture, Computer Networks, Operating Systems, Digital Logic, Computer Graphics, etc. (Those that are not hyperlinked can be found inside this very repo jasha64/jasha64)
• Summer vacation 2019, ARM9 soft CPU core
• Summer vacation 2020, "Vanilla" Pipelined MIPS32 CPU
  • A high-performance pipelined MIPS32 CPU core with caches and TLB, capable of running PMON BIOS
  • "Vanilla" CPU is our work in the NSCSCC Contest [Contest Website], and won 2nd prize
• Spring 2021, Spiking Neural Network RTL Implementation

snn-fpga's Issues

Input_mem BRAM contents

Hi Jasha64, I ran the testbench that you had provided for the spiking neural network but the only digits the code is able to detect are 6 and 9. You are giving the 28x28 image data by doing a rate encoding of the pixel intensities in the image. Can you please clarify what digits from the MNIST dataset are encoded in the input_mem Block RAM? I see that the BRAM comprises the spiking pattern for the 784 pixel for 200 timesteps but I am unable to understand that at a particular timestep which digit in the MNIST dataset is being encoded. Kindly help.