Applied Machine Learning: Foundations

Generate impactful insights with the power of machine learning. Get the foundational skills needed to efficiently solve nearly any kind of machine learning problem.

Anyone who can write basic Python is capable of fitting a simple machine learning model on a clean dataset. The competitive edge comes in the ability to customize and optimize those models for specific problems. The workflow used to build effective machine learning models and the methods used to optimize those models are typically not algorithm or problem specific. In this course, the first installment in the two-part Applied Machine Learning series, instructor Derek Jedamski digs into the foundations of machine learning, from exploratory data analysis to evaluating a model to ensure it generalizes to unseen examples. Instead of zeroing in on any specific machine learning algorithm, Derek focuses on giving you the tools to efficiently solve nearly any kind of machine learning problem.

Introduction

• Leveraging machine learning
• What you should know
• What tools you need
• Using the exercise files

1. Machine Learning Basics

• What is machine learning?
• What kind of problems can this help you solve?
• Why Python?
• Machine learning vs. Deep learning vs. Artificial intelligence
• Demos of machine learning in real life
• Common challenges

2. Exploratory Data Analysis and Data Cleaning

• Why do we need to explore and clean our data?
• Exploring continuous features
• Plotting continuous features
• Continuous data cleaning
• Exploring categorical features
• Plotting categorical features
• Categorical data cleaning

3. Measuring Success

• Why do we split up our data?
• Split data for train/validation/test set
• What is cross-validation?
• Establish an evaluation framework

4. Optimizing a Model

• Bias/Variance tradeoff
• What is underfitting?
• What is overfitting?
• Finding the optimal tradeoff
• Hyperparameter tuning
• Regularization

5. End-to-End Pipeline

• Overview of the process
• Clean continuous features
• Clean categorical features
• Split data into train/validation/test set
• Fit a basic model using cross-validation
• Tune hyperparameters
• Evaluate results on validation set
• Final model selection and evaluation on test set

Conclusion

• Next steps