Intro to Machine Learning Bootcamp

Course Description

Before the course

• Installation - You will be given detailed instructions on how to install Python onto your machine and set up an environment to run all the code during class.
• Jupyter Notebooks - We will be using the excellent Jupyter Notebook to run most of the code during class. It provides an interactive coding environment to quickly execute code, get feedback, and make notes.
• Assignment - You will be given an assignment on how to use Jupyter Notebooks and write Markdown that is expected to be completed before the start of the course.
• Private Slack Channel - Upon registration, you be given access to Ted's private Slack channel where you can directly communicate with him and ask questions about the course.
• All Access Pass ($399 value) - You will be granted the All Access Pass which provides you lifetime access to all current and future books and non-live courses. Currently, this includes thousands of pages of material, hundreds of exercises, and dozens of hours of video.

Day 1

• The Machine Learning Model - Models are anything that help us represent the real world. You will learn how machine learning models are a subset of mathematical models that learn from data.
• Estimators - We use the excellent scikit-learn library to handle all of our machine learning. scikit-learn uses a generic class of objects called 'Estimators' that do all of the machine learning as well as other tasks that learn from data. You will learn the three step process common to all estimators - import, instantiate, fit.
• Linear regression - You will learn how to model data with linear regression, a simple and practical model that is a useful starting point before diving into more advanced models. You will learn how the parameters of the model are fit and understand what it means to minimize squared error.
• K-Nearest Neighbors - You will learn the simple learning approach taken by K-Nearest Neighbors (KNN) and implement the KNN algorithm by-hand using Python.

Day 2

• Decision Trees - Most people are familiar with decision trees outside of the world of machine leaning. You will learn the exact steps that are taken to build a decision tree and implement one yourself by-hand using Python.
• Random Forests - Random forests are a collection of decision trees and one of the most successful classes of machine learning models. You'll learn the sources of randomness that are injected into the decision tree model to create a collection of distinct trees.

• Model Evaluation - The goal of machine learning is to build a model that provides us intelligence on data that it has yet to encounter. You will learn how to determine how well a model is likely to perform on unseen data by properly evaluating it with techniques such as cross validation.
• Model Selection - You will learn how to choose 'better' models by selecting combinations of hyperparameter values that yield good cross-validated results.

Day 3

• Transformers - We take a step back to learn how to transform our data before machine learning so that we can build a better model. You will learn about different transformations such as filling in missing values, scaling features, and encoding categorical data. 
• Machine Learning Pipelines - You will learn how to build machine learning pipelines that can both transform the data and then do machine learning. This step of this pipeline can then be searched to find the best combination of hyperparameters.
• The ColumnTransformer - The ColumnTransformer provides a direct way to apply different transformations to different features within the data. You will learn how to use this single object to contain all the transformation and machine learning steps.
• Feature engineering - Creating new features is one of the best ways to build a better model. You will learn strategies to engineer new features from existing ones to improve results.

Day 4

• Residual Analysis - Analyzing the residuals (errors) from our models can provide great insight into what types of transformations might improve performance. You'll learn how to diagnose different issues with the residuals and what actions to take to handle them.
• Transforming Target Variable - In order to alleviate some of the issues during residual analysis, it may become necessary to transform the target variable as well as the features. You will learn different strategies on how to transform the target variable to get better model performance.
• Penalized Regression - You will learn to reduce over-fitting during training by applying a penalty term that is proportional to the size of linear regression coefficients.
• Bias-Variance Tradeoff - Penalizing models can reduce their predictive power, but it can also produce more stable models. You will learn about this balance between predictability and model stability which is referred to as the bias-variance tradeoff.

Day 5

• Custom Error Metrics - You will learn how to create custom scoring metrics that are not available in scikit-learn by default. You will use these metrics during model evaluation and selection.
• Complete Machine Learning Pipeline - We will spend time putting together all of the components of the course up to this point to build a single complete machine learning pipeline.
• Model persistence - Building a model is great fun, but we need a way to access it in the future. You will learn how to save a model as a file and retrieve it any time you need to make a new prediction.
• Kaggle Competition - To further help you practice all of the material covered in the course, you will enter a machine learning competition offered on Kaggle. You will learn best practices on how to manage your models and results.

After the course

• Certificate of completion - Upon conclusion of the course, you will be given a direct path towards mastering the fundamentals of machine learning using Python. If you go on to complete all the tasks on this path, you will receive a certificate of completion.
• Lifetime access to material - You will always have access to the material after the course has completed. Ted upgrades and adds to his material on a regular basis. You will always have immediate access to the latest updates.
• Lifetime Slack Access - You will always have access to Slack allowing you to interact with Ted and all previous students.