📘 ML From Scratch Tutorials

A curated collection of Machine Learning tutorials created from scratch, combining theory, intuition, math, and hands-on implementation.
These Jupyter notebooks walk through ML concepts step-by-step — perfect for beginners, students, and anyone who wants to understand ML deeply rather than treating it as a black box.

Created by Deepti Jethwani.

---

📂 Repository Contents (Detailed Explanation of Each File)

1. Key concepts of ML.ipynb

This notebook introduces the fundamental concepts of Machine Learning:

• What ML is and how it works
• Types of learning: Supervised, Unsupervised, Reinforcement
• Key terms: features, labels, models, training, testing
• Bias–variance tradeoff
• Underfitting vs Overfitting
• Train/test split basics
• Introduction to evaluation metrics

Goal: To build a strong conceptual foundation before exploring algorithms.

---

2. MNIST - Classification Problem.ipynb

This notebook walks through digit classification using the MNIST dataset:

• Loading and exploring the MNIST dataset
• Preprocessing and normalization
• Building a classification model
• Training the model on 60,000 images
• Testing on 10,000 images
• Visualizing predictions and errors
• Evaluating accuracy

Goal: Learn end-to-end ML workflow on a real-world dataset.

---

3. Linear Regression ( Math ).ipynb

A mathematical and practical overview of Linear Regression:

• Intuition of linear relationships
• Hypothesis function
• Cost function (MSE)
• Gradient Descent explained step-by-step
• Formula derivation
• Implementing Linear Regression from scratch
• Plotting regression lines and errors

Goal: Understand the math and mechanics behind regression models.

---

4. Logistic Regression.ipynb

A complete guide to Logistic Regression for classification:

• Logistic function and sigmoid
• Why linear regression fails at classification
• Decision boundary
• Log loss / Cross-entropy
• Gradient descent for logistic regression
• Implementing logistic regression from scratch
• Evaluating model accuracy

Goal: Learn binary classification deeply.

---

5. SVM.ipynb

Support Vector Machines explained with intuition and implementation:

• Maximum-margin concept
• Hyperplanes and support vectors
• Kernel trick (Linear, Polynomial, RBF)
• Training an SVM classifier
• Visualizing decision boundaries
• Understanding when SVMs perform best

Goal: Build intuition for one of the strongest classical ML models.

---

6. Decision Trees.ipynb

A full tutorial on Decision Trees:

• Gini impurity
• Entropy and information gain
• How trees split data
• Classification and regression trees
• Implementing decision trees
• Training on datasets
• Tree visualization using .dot files

Goal: Understand how tree-based models make decisions.

---

📢 Future Additions (Planned)

• KNN from scratch
• Naive Bayes
• PCA
• Random Forest
• Gradient Boosting
• Notes