README for a Cerebellum-Inspired Deep Learning Algorithm
Cerebellum-inspired spiking neural network(DLCISNN), a hybrid of Spiking Neural Network and Deep learning networks with the architecture and firing behaviour inspired from cerebellar microcircuitry. This README includes the software requirements and instructions to test the algorithm.
Developed by Asha Vijayan and Shyam Diwakar
Computational Neuroscience lab, Amrita School of Biotechnology, Amrita Vishwa Vidyapeetham, Kollam, Kerala, India.
www.amrita.edu/compneuro
Last updated on 27-February 2018
Required software and Instructions
- MATLAB
Install MATLAB 2014 or higher
Download all the .m files and the decoder folder into a single folder
Add 'decoder' folder to path in MATLAB.
Right clicking the 'decoder' folder;
Click on 'Add to path';
Click 'Selected folders and subfolder'.
- Dataset
Provide a dataset of format .xlsx with feature label and class label as first row.
Followed by the datapoints have to be numerical(Categorical data has to be represented as numbers). Zero cannot be used as a number/ label
This folder has 4 datasets.
1. ASD_adolescent.xlsx
2. ASD_adult.xlsx
3. ASD_child.xlsx
4. Iris.xlsx
How to compile the program
- Open the main program (DLCISNN.m) and submit the filename in line 11
- If the classification involves more than 2 class label like in Iris dataset, Open the folder 'decoder'; Open the decoding program, comment the conditional statements under the heading 'For 2 class labels' and uncomment the statements under 'For 3 class labels'
