The assignment consists of three problems based on basic machine learning and computer vision.

Numerous problems in these areas are well studied in statistics and applied mathematics.

Solutions are to be implemented in python by filling out required functions in each python file.

A basic framework for data i/o and evaluation is already provided (see header comments in each python file).

Please note that all required libraries are already imported so please DO NOT import anything new.

The four problems are briefly discussed below.

1. Gradient Descent: This is a popular optimization problem to find solutions to differentiable equations.

Typically, learning problems involve minimizing a cost function by appropriately setting model parameters.

In this task, we are given a set of (noisy) points on a line and we wish to retrieve model parameters (intercept and slope) through gradient descent.

Please refer to 'gradient_descent.py' and inline comments for further details.

2. Eigenfaces: This is a popular application of learning a basis representation of input data.

The application of this technique is the basis for simple recognition/compression algorithms.

In this task, we want to learn orthonormal basis using PCA of images that correspond to faces.

Please refer to 'eigenfaces.py' and inline comments for further details.

3. Classification: This is among the basic tasks of machine learning problems.

Here, we will learn a classifier to using groundtruth labels on the training data to be able to distinguish between two object classes.

You will use the scikit library to learn two classifiers (svm and random forest).

Feel free to explore the parameters of both models to maximize classifier performance.

Please refer to 'classification.py' and inline comments for further details.

4. Disparity map: This is among the basic tasks of 3D computer vision

Here, given two differnce perspectives of the same scene, we will reconstruct an approximate of the depth map.

This is called the disparity map (higher disparity is similar to lower depth).

You will use the scikit library to implement the module. Feel free to explore the parameters 'downsample' and 'patchsize'

Please refer to disparity.py and inline comments for further details.