Adaptive-Computation-and-Machine-Learning

This repository will consist of all the programming assessments i had for a Wits University honours course(Adaptive Computation and Machine Learning)

In Assignment 1 we were required to:

Implement linear regression and see it work on a simple dataset of our choice. Furthermore what was required is:

1. Basic implementation of linear regression
2. Exploring the effect of different learning rates on convergence
3. Implementation of feature scaling, feature standardization and regularization for improved learning
4. Data visualization to understand the working of algorithm and other steps (applicable to all the above stages)

In Assignment 2 we were required to:

Implementing a neural network from first principles and exploring use of different activation functions and different network architectures. Furthermore what was required is:

1. Basic implementation of neural network trained using back-propagation

• Provide clear comments for each sections/phases of your code.

2. Exploring the effect of different activation functions

• You are required to implement at least two activation functions and provide visualizations to help understand their impact/use.

3. Exploring the effect of network size on generalizability (eg. number of hidden layers, number of hidden neurons)

• You are required to explore at least two options (eg. two hidden layers vs one hidden layer, x neurons vs y neurons in the hidden layer/layers)

4. Effective use of data visualization and analysis of the results to understand the working of algorithm and other steps

• This is applicable to all the above stages. eg. learning curves, no. of hidden layers/neurons vs convergence, activation functions etc.

In the Project we were required to:

Research and implement at least two machine learning algorithms. Furthermore what was required is:

1. Explore the different machine learning paradigms, tools, algorithms that can be used to solve the recognised project idea/question
2. Finalise on a ML technique/algorithm.Note: You are required to only explore one model/algorithm. However, if you do want to perform a comparative analysis between couple of algorithms you are free to do so.
3. Implement the baseline algorithms, run the experiments and understand the first phase of outcomes.
4. Employ optimization techniques(for both model parameters, algorithm parameters).
5. Analyse the final results.