Skip to main content

ML Fundamentals

What You'll Build Here​

This module introduces the core concepts behind machine learning. You'll understand how a single neuron thinks, how it learns from mistakes (gradients), and how classical ML models work under the hood.

💭 Real Talk

You can invoke model.fit() without knowing this. But you cannot debug a model that refuses to learn, or optimize an architecture, without understanding these fundamentals.

Estimated Time Impact

20-25 hours total

Classical ML Models(8h)
Deep Learning Primitives(10h)
Evaluation Metrics(5h)

1. Classical ML Models ("Old School Cool")​

Not every problem needs a billion-parameter Transformer. In fact, for tabular data, these "old school" models often win on speed and interpretability.

article
beginner

IBM: What is Machine Learning?

1-2 hours

Linear & Logistic Regression​

video
beginner

StatQuest: Linear Models

2 hours

Support Vector Machines & Decision Trees​

video
intermediate

StatQuest: Decision Trees & Random Forests

3 hours
video
intermediate

StatQuest: Support Vector Machines

4-6 hours
Why These Matter

Production systems love Random Forests. They are robust, don't need scaled data, and are easy to explain to a boss. "The model failed because feature X was > 5" is easier to sell than "Matrix multiplication said so."

2. Deep Learning Primitives​

course
intermediate

Andrew Ng: Neural Networks and Deep Learning (Course 1 & 2)

8-10 hours

Focus:

  • How a neuron computes (Weights + Bias + Activation)
  • Forward Propagation (Prediction) vs. Backward Propagation (Learning)
  • Loss Functions (How wrong are we?)
  • Gradient Descent and how models learn

3. Evaluation Metrics​

How do you know if your model works? Accuracy is often a liar.

article
beginner

12 Important Model Evaluation Metrics

2 hours

Action Items:

  • Completed the entire Classification metrics section (Precision, Recall, F1-Score, ROC-AUC, Confusion Matrix)