This course gives an introduction to ML algorithms. It is great to get an idea about algorithms, but not to learn in detail. One of the main drawbacks for me is that this course has no lecture notes.
Looking to break into machine learning? Need to review the ins and outs of each algorithm? Preparing for an interview? Curious to see how these algorithms are applied to business? As ML practitioners, the true value of ML is not in memorizing complicated formulas. It’s not using the latest, greatest deep learning architecture. It’s knowing when to use an algorithm and how to maximize the impact of that model. It’s knowing how to use them to solve REAL business problems. The true value of ML is not ML. It’s solving important business problems. Whether you need to build a forecasting model that forms the backbone of the ads business, builld a recommender system powering millions of purchases or build a fraud detection system that catches bad apples, this course will arm you with both the ML knowledge AND the know-how on how to apply it to your business problem. We don’t want you to leave this course just knowing ML. We want you to leave this course to leave as ML practitioners.
You will learn the ins and outs of each algorithm and we’ll walk you through examples of the world’s biggest tech companies using these algorithms to apply to their problems. This course touches on several key aspects a practitioner needs in order to be able to aply ML to business problems:
In this introductory section, you’ll learn which of the course is best for you, the types of ML issues businesses need to solve, and additional resources from the course authors.
Linear regression is the most dynamic model out of all we review. It’s an exceptional framework for making predictions and extracting insight into relationships between variables.
One way to limit some of the negatives of linear regression is via regularization. We can regularize a linear regression model using ridge, lasso, and elastic net algorithms.
Like linear regression, logistic regression is one of the most powerful and straightforward models.
Gradient descent is an optimization algorithm that powers many of our ML algorithms. Think of it as a helper algorithm, enabling us to find the best formulation of our ML model.
Decision trees work for classification and regression problems. While individual decision trees don’t typically produce the best predictive outcomes in real life, they are highly interpretable.
Random forest is one of the most popular models for classification and regression. It works exceedingly well with datasets with many categorical values or a mix of categorical and continuous variables.
Gradient-boosted trees are a way for our models to learn from their mistakes. Unlike random forest—where we grow our trees in parallel, then aggregate the results—with gradient-boosted trees, we grow trees sequentially.
XGBoost is a famous open-source gradient-boosting algorithm for supervised learning tasks, including regression and classification problems. Reducing errors during each iteration makes better predictions by combining several decision trees.
K-nearest neighbors is a popular machine learning algorithm for classifying and predicting data. It works by finding the k closest data points to a new, unknown data point and categorizing it based on the most common class among its neighbors.
K-means clustering is the first unsupervised learning method we will introduce. As a reminder, unsupervised learning means that our data doesn’t have specific target labels we try to classify. Instead, we consider data as a whole and see algorithmically what groups can be made from it.
Hierarchical clustering is similar to how you organize files into folders on your computer. Whenever we organize our files into their folders, we perform hierarchical clustering.
Support vector machines (SVMs) work by identifying a hyperplane that best separates the data into different classes or predicts the target variable for regression. It also aims to find the optimal hyperplane that maximizes the margin between the classes while minimizing the misclassification error.
Artificial neural nets (ANNs) are machine learning algorithms based on the structure and function of biological neurons. Layers of interconnected nodes process and transform input data to produce predictions. The neural network learns by adjusting the weights and biases of the connections between the nodes based on the error of the predictions.
Collaborative filtering uses ratings from a group of (collaborative) users to infer the preference of another (filtering) user.
Discover the intricacies of the Naïve Bayes algorithm, exploring Bayes Theorem, its intuitive definition, why it's called "Naïve", the ML process, the pros and cons of this algorithm, a real-life business example, and effective use cases.
“Most courses only explain the concept itself, and they’re done. Not only do we teach you the concept, but we also lay out its pros and cons and when you should use the concept to solidify your intuition. ”
Works at Large music streaming platform
The Machine Learning Algorithms A-Z
with Jeff Li and Ken Jee