Machine Learning for Everyone - Lecture Notes

Introduction to Kylie Ying

• Physical scientist and engineer worked at notable places like MIT, CERN, Free Code Camp.
• Aimed at teaching machine learning to beginners.

Overview of Machine Learning

• Machine learning (ML) is a sub-domain of computer science that focuses on algorithms allowing computers to learn from data without explicit programming.
• Difference between AI, ML, and Data Science:
  • AI: Enabling computers to perform human-like tasks.
  • ML: Subset of AI focused on making predictions using data.
  • Data Science: Field that finds patterns and insights in data.

Types of Machine Learning

1. Supervised Learning:
   • Uses labeled inputs to train models.
   • Example: Predicting classes based on features (e.g., pictures of animals).
2. Unsupervised Learning:
   • Uses unlabeled data to find patterns or groupings.
   • Example: Clustering data points based on similarities.
3. Reinforcement Learning:
   • An agent learns in an interactive environment through rewards and penalties.

Supervised Learning

• Key Concepts:
  • Features: Inputs used by the model to make predictions.
  • Labels: Output to predict based on features.
• Types of Supervised Tasks:
  • Classification: Predicting discrete classes (e.g., spam/not spam).
    • Binary Classification: Two classes (e.g., cat vs not cat).
    • Multi-Class Classification: More than two classes (e.g., cat, dog, lizard).
  • Regression: Predicting continuous values (e.g., price of a house).

Data Set Example: Magic Gamma Telescope Dataset

• Uses properties of light hitting a gamma telescope to predict types of particles.
• Process:
  1. Download dataset from UCI machine learning repository.
  2. Set up Google CoLab.
  3. Import libraries: NumPy, Pandas, Matplotlib, etc.
  4. Read and prepare the dataset (e.g., assign column names).

Data Preparation and Exploration

• Data Exploration Techniques:
  • Viewing the first few entries of the dataset (using .head()).
  • Checking unique labels in the dataset.
  • Converting labels of gamma and hadron particles from letters to numbers.

Supervised Learning Implementation Steps

1. Data Loading:
   • Load the data into a DataFrame.
2. Data Cleaning:
   • Assign column names.
   • Convert categorical labels into numeric.
3. Feature Selection:
   • Select relevant features for the model.

Example of Implementation in Google CoLab

• Import necessary libraries.
• Load the data using Pandas.
• Clean the data and handle missing values.

Machine Learning Models Implemented

1. K-Nearest Neighbors (KNN):

   • Classifies points based on majority class of neighbors.
   • Distance function used: Euclidean distance.

2. Naive Bayes:

   • Uses probability to classify data based on prior probabilities.
   • Assumes independence of features.

3. Logistic Regression:

   • Predicts probability of class membership using a logistic function.
   • Models the relationship between dependent and independent variables.

4. Support Vector Machines (SVM):

   • Finds hyperplane that best separates different classes in high-dimensional space.

5. Neural Networks:

   • Models with layers of interconnected nodes (neurons).
   • Trains on labeled data to predict outputs.
   • Uses backpropagation to adjust weights based on loss.

Evaluation of Machine Learning Models

• Performance metrics include:
  • Accuracy
  • Precision
  • Recall
  • F1 Score
  • Mean Absolute Error (MAE)
  • Mean Squared Error (MSE)
  • R-Squared (R²)

Unsupervised Learning - K-Means Clustering

• K-means clustering algorithm:
  1. Choose number of clusters (k).
  2. Randomly assign initial centroids.
  3. Assign points to nearest centroid.
  4. Recalculate centroids based on assignments.
  5. Repeat steps until convergence.

Unsupervised Learning - Principal Component Analysis (PCA)

• Reduces dimensionality of dataset while preserving variance:
  • Projects data into lower dimensions.
  • Finds the direction with the most variance.
  • Enables visualization of high-dimensional data.

Conclusion

• Summarized the key concepts and implementations in supervised and unsupervised learning.
• Encouraged community engagement for further learning.