SAiDL Summer of Code

This repo contains the 2 models for the SSoC project. The first one is a Leaky Intgerate and Fire (LIF) model, which is modeled using numpy and matplotlib in Python3. The second one is a simple 2-layer shallow neural net, it is made to model the given input table for XOR/XNOR output of 2 binary bits using a third binary input bit.

Leaky Integrate and Fire Model

It is a basic Spiking Neuron model based on the following equations :

$\tau\*\frac{du}{dt} = - u + IR$

$u(t^{(f)})= \nu$

$\lim_{t \rightarrow t^{(f)},t > t^{(f)}} = u_{rest}$

$u = u_{rest}\ if\ t^{(f)}<t<t^{(f)}+t_{refrac}$ </t<t^{(f)}+t_{refrac}”>

Results

The observed results are plotted and saved as images in the Leaky Integrate and Fire Model folder.
The current used for each testcase is mentioned below.

Input Values :
I.   I = 2uA for 10ms <= t <= 75ms; I = 0Amp otherwise
II.  I = 1.5uA for 0ms <= t <= 100ms
III. I = 5sin(ωt)uA where ω = 45deg/msec
IV.  I = -2sin(ω1t) + 3sin(ω2t) + cos(ω3t)uA where ω1 = 45deg/msec, ω2 = 60 deg/msec, ω3 = 30 deg/msec

Artificial Neural Network for XNOR/XOR

This is a basic model to predict the XOR/XNOR values of 2 binary single bit inputs based on a third bit (switch). It is a shallow network as the data is ordered and easy to classify.

Input Values

Bit 1	Bit 2	XNOR(0)/XOR(1)	Output
1	1	0	1
1	0	0	0
0	1	0	0
0	0	0	1
1	1	1	0
1	0	1	1
0	1	1	1
0	0	1	0

Author

Dev Churiwala