ml-deeplearning-notebooks

This is a series of Jupyter notebooks introducing machine learning and deep learning concepts, developed for the workshop Introduction to Machine Learning in Python, offered by the Carpentry @ UCSB Library in Winter 2026.

These notebooks take you from ML fundamentals through convolutional neural networks.

Authors: Tian Qiu, Jose Nino Muriel

Notebooks

1. Introduction to Machine Learning

Covers the foundations of machine learning, including:

• ML vs rule-based programming; classification, regression, and clustering
• 10-step ML workflow and problem formulation
• Project: Penguin species classification (Seaborn penguins dataset)
  • Logistic regression and hyperplane-based decision boundaries
  • Confusion matrices and accuracy evaluation
• Introduction to neural networks: single neurons, ReLU/GELU activation functions, one-hot encoding

2. Neural Network Classification

Builds a full neural network classifier using Keras/TensorFlow:

• Project: Penguin species classification (continued)
• Keras Sequential API: Input, Dense, softmax output layer
• Categorical crossentropy loss, Adam optimizer
• Training for 100 epochs, loss curve visualization
• Predictions, confusion matrix heatmap

3. Monitoring Training & Regression

Extends to regression tasks and introduces training best practices:

• Project: Predicting Basel sunshine hours from European weather data (Zenodo dataset, 11 cities, 3654 days)
• 70/15/15 train/validation/test split strategy
• Gradient descent, batch training, and epoch concepts
• RMSE metric, baseline model comparison
• Overfitting detection using validation loss
• EarlyStopping callback

4. Image Classification with CNNs

Introduces convolutional neural networks for image data:

• Project: MNIST handwritten digit classification (60,000 training images, 28×28 grayscale)
• Image normalization and channel dimensions
• Conv2D filters, kernel size, stride, padding
• MaxPooling2D for dimensionality reduction
• Full CNN architecture: Conv → Pool → Conv → Pool → Flatten → Dense → Softmax
• Visualizing learned filters and intermediate layer outputs

Key Learning Objectives

Notebook	Task	Key Concepts
1	Classification	ML fundamentals, neurons, activations
2	Classification	Keras model building, training loop
3	Regression	Train/val/test split, overfitting, early stopping
4	Classification	CNNs, filters, pooling

Technologies

• TensorFlow / Keras — model building and training
• scikit-learn — classical ML, metrics, train/test split
• pandas / NumPy — data manipulation
• Matplotlib / Seaborn — visualization