Machine Learning Fundamentals

After this lesson, you will be able to: Understand how machines learn patterns from data, supervised, unsupervised, reinforcement.

Machine learning is the engine under modern AI. Instead of hand-writing rules, you give the machine data and a goal, it finds the pattern. This lesson covers the three flavors and the lifecycle of a real ML project.

Prerequisites:What is Artificial Intelligence?

Three types of ML

Supervised, labeled data. (Photos labeled 'cat'/'dog' → predict new photo.) Unsupervised, no labels. (Cluster customers by purchase patterns.) Reinforcement, agent acts in environment, gets rewards. (Chess, robot control.)

Your first ML model (Python + scikit-learn)

Predicting iris flower species, the 'hello world' of ML:

python

from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.tree import DecisionTreeClassifier
from sklearn.metrics import accuracy_score

X, y = load_iris(return_X_y=True)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3)

model = DecisionTreeClassifier()
model.fit(X_train, y_train)
predictions = model.predict(X_test)

print(f'Accuracy: {accuracy_score(y_test, predictions):.2%}')

ML project lifecycle

1
1. Define problem + success metric.
2
2. Collect + clean data.
3
3. Split train/val/test.
4
4. Pick a model family.
5
5. Train + tune.
6
6. Evaluate on test.
7
7. Deploy + monitor for drift.

Bias-variance tradeoff

Underfit (high bias), model too simple, misses pattern. Overfit (high variance), model memorizes training data, fails on new data. Most ML work is finding the sweet spot.

←What is Artificial Intelligence?

Back to Artificial Intelligence

Neural Networks and Deep Learning→