vineeths96/svm-and-neural-networks

In this repository, we will explore and compare different methods of learning non-linear classifiers such as SVMs and Neural Networks.

Pattern Recognition and Neural Networks

SVMs and Neural Networks
Explore the repository»

View Problem Statement View Report

tags : svm, nonlinear classifier, scikit-learn, tensorflow, MNIST, fmri, alzeimers, epilepsy, EEG

About The Repository

This repository holds the python implementation files for Assignment #3 for E1 213 Pattern Recognition and Neural Networks offered at the Indian Institute of Science (IISc), Bangalore. In this assignment we will explore support vector machines (SVM) and neural networks. The following methods have been implemented across the problems.

Problem 1 deals with learning an SVM and neural network for synthetic two-class data with class conditional densities uniform for label noise level 0%, 20%, and 40%. Problem 2 deals with learning an SVM and neural network for synthetic five-class data for label noise level 0%, 10%, and 25%. Problem 3 deals with learning an SVM and neural network for the popular MNIST dataset. Problem 4 deals with learning an SVM and neural network for classifying Alzeimer’s and Normal subjects using fMRI recordings for two different parcellations of brain. Problem 5 deals with learning an SVM and neural network for a five-class classification using Epileptic seizure EEG data.

Built With

This project was built with

• python v3.7
• The list of libraries used for developing this project is available at requirements.txt.

Getting Started

Clone the repository into a local machine using

git clone https://github.com/vineeths96/SVM-and-Neural-Networks

Prerequisites

Please install required libraries by running the following command (preferably within a virtual environment).

pip install -r requirements.txt

Instructions to run

There are four python files - problem_1.py, problem_2.py, problem_3.py , problem_4.py and problem_5.py - each corresponding to the particular problem in the Problem Statement. Each problem has their corresponding implementation files under a python package with the same name. Each package has python modules and functions to load data, train a model, test it, and write the performance metrics to an output file at ./results with the same file name.

Running the program

python problem_<QUES_NUM>.py

Results

View Report for the results and detailed discussions.

License

Distributed under the MIT License. See LICENSE for more information.

Contact

Vineeth S - [email protected]

Project Link: https://github.com/vineeths96/SVM-and-Neural-Networks