Understanding the World Through Data (edX)

Whether you're a high school student or someone switching careers, all you need to get started in this course is a curiosity about the topic of machine learning and a willingness to tinker around with your computer.

The course is taught by modules. Within each module, you'll have access to videos, short exercises, and a final capstone project. In Module 1, you'll begin by looking at different kinds of data. To help you explore the data, you'll dive right into some programming with the Python programming language. You don't need to have any programming background, we will guide you on how to leverage Python to explore and visualize any data.

One kind of data you'll work with is data that relates one variable to another. Coming up with a relationship between two variables—one depending on the other—is at the center of Module 2. In that module, you'll build up some core concepts before seeing your first machine learning algorithm. The goal is to use programming to create models that describe mathematical relationships between data. You'll be able to see how good the model is and use it to make predictions about new data.

In Module 3, you'll see a discussion about where imperfections in collected data might come from. You rarely have perfectly “clean” data sets, so it's important to understand how imperfections impact the model that an algorithm might come up with. To this end, we will introduce the notion of data distributions and build up to the concepts of biased and unbiased noise.

Another kind of data you'll work with is data that belongs in different groups (or classes). Creating a model that predicts what group data belongs in is at the center of Module 4. You'll work through different ways of thinking about this problem and see three different ways of approaching making such groupings (classification).

What you'll learn

- Python programming and the Colab notebook programming environment

- Dependent and independent variables

- Coming up with relationships between data using linear and polynomial regression models

- Recognizing how data is distributed

- How to observe noise in distributions and when to ignore it

- Categorize data into groups with classification models

- And more!

Syllabus

Module 1: How to represent and manipulate data

Examples of numerical data

The Python programming language and the Colab notebook programming environment

Loading datafiles in Colab as dataframes and performing simple operations (selecting rows or columns, filtering data by specific conditions, grouping data, applying functions on the resulting groups)

Finding the correlation between columns of the dataframe

Visualizing the data using line plots, scatter plots, histograms, correlation matrix

Module 2: Reverse engineering nature

Dependent and independent variables and how they correspond to real life scenarios

Intuition for what a linear model is

Intuition for what a polynomial model is

Python libraries that can perform the linear regression on data

Compare the quality of different models (mean-squared-error and R^2 values)

Fitting higher order polynomials

Overfitting

Module 3: Distributions and Latent Variables

Uniform distributions

Gaussian distributions

Distribution mean and standard deviation

Noise in distributions (biased and unbiased noise)

Module 4: How machines think

Categorizing data based on particular conditions being met

Using linear regression to classify a new datapoint as above or below the best fit line

Using a support vector classifier to separate two groups of data and classifying a new datapoint into a group

Using logistic regression to classify data into two groups and finding the probabilities of a new datapoint falling into each group

Understanding how to divide data into training and test sets