stefanogenettiunitn / machinelearning_passerini

The purpose of this repository is to write a handbook about the topics covered by professor Andrea Passerini in his course: Machine Learning.

Machine Learning Notes 🤖📚

Welcome to the Machine Learning Notes repository! Here you'll find comprehensive notes on various topics covered in the Machine Learning course of the Master Degree program at the University of Trento. Whether you're a student studying for exams or someone eager to dive into the world of machine learning, you're in the right place! 🎓

In this updated version, I've added additional content, including a new chapter, and fixed all typographical errors to provide you with the best learning experience possible. 🚀

Contents 📋

1. Introduction to Machine Learning 🤖
2. Decision Trees 🌳
3. K-nearest Neighbors 🏘️
4. Linear Algebra ➗
5. Probability Theory 🎲
6. Evaluation ✔️
7. Parameter Estimation 🔍
8. Bayesian Networks 🔄
9. Inference in BN 🧮
10. Learning BN 📖
11. Naive Bayes 🤞
12. Linear Discriminant Functions ➡️
13. Support Vector Machines 🛠️
14. Non-linear SVMs 🔄
15. Kernel Machines ⚙️
16. Deep Learning 🧠
17. Ensemble Methods 🎭
18. Unsupervised Learning 🧩
19. Reinforcement Learning 🎮

Download 📥

To download a PDF version of these notes, click on the image below:

Machine Learning Notes

Feel free to explore, learn, and contribute to this repository! Let's dive into the fascinating world of machine learning together! 🤗💻

machinelearning_passerini's Issues

A couple of mistakes in chapter 12.7 Generative linear classifiers

Hello, I found a couple of potential mistakes in the chapter 12.7 about Generative linear classifiers.

• Page 151, at the end:

  $\displaystyle \prod^{K}_{k=1}$ : $x_j$ is a discrete variable with $k$ possible values

  There's a little typo, it should be " $K$ possible values".

• Page 152, at the beginning:

  $θ_{ky_i}$ parameter of the kth value for $y_i$ class. It is the probability that feature $k$ of $x_j$ is true given $y_i$. In essence $θ_{ky_i}$ is raised to the power of $1$ when $x_j$ has the kth feature, otherwise $θ_{ky_i}$ is raised to the power of $0$. (Figure 12.12)

  From my understanding, $k$ is not a feature. Instead, it represents the index of a value within a set of $K$ possible states that the actual feature $x_j$ can take. Therefore, the statement should be revised as:

  $θ_{ky_i}$ parameter of the $k$-th value for $y_i$ class. It is the probability that the feature $x_j$ takes the $k$-th value given $y_i$. In essence $θ_{ky_i}$ is raised to the power of $1$ when $x_j$ is equal to the $k$-th value, otherwise $θ_{ky_i}$ is raised to the power of $0$. (Figure 12.12)

Thank you guys for creating this handbook. I truly appreciate your efforts, and I'm grateful for the excellent outcome.