Data Pattern Recognition for the Rest of Us — 33 ML Techniques with Colab Notebooks

33 Techniques.
One place to learn them all.

A non-linear ML learning hub built around techniques, not chapters. Content is modelled after ISLP (Introduction to Statistical Learning with Python) — with ▶ Stanford lectures — and fpppy (Forecasting: Principles & Practice, Pythonic) — with ▶ video lectures — for time series. 33 techniques, each with a Colab notebook, real ISLP data, runnable code, exercises, and a quiz with instant AI feedback.

Completed

Quizzes passed

Techniques

Topic groups

Create a GitHub account (free) at github.com/join if you don't have one. GitHub stores your notebooks and tracks your work across devices.

Fork the course repo. Go to the repo link in the header and click Fork (top-right). This creates your own copy of all 33 notebooks under your GitHub account.

Open any notebook in Colab. Click the orange Open in Colab button on any card below. Colab is Google's free Python environment — no install needed, free GPU included.

Work through the notebook. Run cells top-to-bottom with Shift+Enter. Fill in # YOUR CODE HERE sections, complete the quiz, then run the AI grading cell at the bottom for instant feedback.

Get AI feedback on your quiz. Run the grading cell — it prints a ready-made prompt. Copy it and click the ✨ Gemini button in the Colab toolbar to paste it directly into Gemini. Keep the conversation going — paste any output or chart and ask follow-up questions. No login, no API key required.

Save your work back to GitHub. In Colab: File → Save a copy in GitHub. Choose your forked repo. Your work is now versioned and accessible anywhere.

Track your progress here. Click any technique card → mark as complete or take the quiz. Progress is saved in this browser. Come back any time — no login required for this page.

Found a bug or have a suggestion? Open a GitHub Issue — describe the notebook name, cell number, and what went wrong. This is how real open-source projects work, and your reports directly improve the course.

Need help? First-time Colab users: go to colab.research.google.com and sign in with your Google account. First-time GitHub users: the GitHub fork guide takes about 5 minutes. AI feedback tip: The grading cell prints a prompt — find the Gemini button in the Colab toolbar, paste it in, and keep asking follow-up questions about any output or chart in the notebook.

Excel Spreadsheet	→	pandas DataFrame
Single column	→	pandas Series
Cell reference (A2)	→	df.iloc[1, 0]
Named range	→	df['column']

Filter rows	→	df[df['col'] > 5]
Sort by column	→	df.sort_values('col')
Pivot table	→	df.pivot_table()
VLOOKUP	→	df.merge()

SUM / AVERAGE	→	.sum() / .mean()
COUNTIF	→	df['col'].value_counts()
IF / IFS	→	np.where() / .map()
Charts	→	matplotlib / seaborn

Descriptive stats	→	df.describe()
Remove duplicates	→	df.drop_duplicates()
Find blanks	→	df.isnull().sum()
Trendline / Forecast	→	LinearRegression()

RMSE	Root Mean Squared Error — same units as target. Most interpretable.
MAE	Mean Absolute Error — robust to outliers.
MSE	Mean Squared Error — penalises large errors heavily.
MAPE	% error — scale-free. Avoid when target has zeros.

Accuracy	% correct. Misleading when classes are imbalanced.
Recall	Catches positives. Prioritise when missing one is costly.
Precision	Of predicted positives, how many correct. Prioritise to reduce false alarms.
F1 Score	Best for imbalanced classes. Harmonic mean of Precision + Recall.
AUC	Works across all thresholds. 0.5=random, 1.0=perfect.

33 Techniques.One place to learn them all.

33 Techniques.
One place to learn them all.