Linear regression is one of the simplest statistical analysis methods used in Machine Learning. In this problem set, we use linear regression to identify a function that would fit the trend of the data. This function would then serve as a predictor with a margin of error in predicting future specific values based on the identified regression function.

Link to Code

Logistic regression utilizes a special exponential function called the sigmoid function. The sigmoid function is reliant on logarithms to create a graph that would approach a minimum of 0 and a maximum of 1. Using this, we can compare probabilities to this sigmoid function to create binary decision making, allowing our program to decide between yes or no, true or false, and 0 or 1.

Link to Code (Version 1)

Link to Code (Version 2)

Multi-class Classification is a classification system built on logistic regression. For each classification, a logistic regression is used to identify the likelihood of it being any of the categories and using the highest probability category as the output.

Link to Code

Neural networks is a method in artificial intelligence wherein, in contrast to machine learning, features, parameters, and indicators are unknown. Neural networks require annotated and labeled data using which the neural network will identify significant indicators by itself. Neural networks are patterned to human intelligence wherein we have the ability to differentiate and identify without having rigid indicators. We can differentiate dogs and cats based on the face, nose, tails, and ears by intuition but neural networks cannot. Neural networks will identify these differentiable features, the face, nose, tails, and ears, then would base all future decision on these features.

Link to Code