Resources

Notebooks from the club sessions: regression, networks, CNNs.

The fastest way to pick up Python syntax.

Clear tutorials on data structures, kNN, k-means.

Broadcasting and axis operations.

groupby, merge and pivot_table, used in every tabular problem.

NumPy, Pandas, Matplotlib and scikit-learn in one place, with runnable code.

How to show your data: axes, legends, subplots, in a short tutorial.

Statistical plots on top of matplotlib, straight from a DataFrame.

The starting point for metrics, regression and thresholds.

Short explainers: loss, confusion matrix, feature engineering, fine-tuning.

Short articles on every syllabus topic, with code.

Hands-on micro-courses: feature engineering, intro ML, pandas.

Scaling, encoding, imputation, with short examples.

Cross-validation, metrics, learning curves.

Every statistics or ML concept, explained visually, step by step.

Probability and statistics, with graphs you nudge with your own hand.

Free PDF, the most approachable entry into classic supervised models.

Andrew Ng's notes: the math behind logistic regression and SVM.

The API you use at the contest, with theory on each page.

Decision trees and random forests, explained visually.

The course half the search material comes from.

The denser sibling of ISL, with the math behind every method.

K-Means, DBSCAN, hierarchical, with a visual comparison of when to use each.

K-means from scratch in Python: choosing k and the silhouette score.

PCA, LDA, UMAP at the intuition level, without heavy math.

Random Forest, gradient boosting and voting, with the hyperparameters explained.

Check the rules: it's not allowed at every stage.

The NLP book, free: embeddings and n-grams cover the text modules.

Word2Vec, RNN, LSTM, with slides and assignments on embeddings and sequences.

TF-IDF plus a classifier, step by step: the classic NLP pipeline.

A clean map of modern NLP, useful before you dig in.

From tokenization to fine-tuning, with transformers, free and no account needed.

The best public vision course; the convolution and backprop notes are worth reading in full.

Filters, thresholds, morphology: see what a kernel does before you write it.

Feature extraction vs fine-tuning, with code on torchvision.

A gentle intro to vision, free to audit.

Classification, detection, segmentation, with a Colab notebook for every unit.

Builds backprop and a network from scratch, in code.

The visual intuition behind networks and gradient descent.

Start with "Learn the Basics". You work in PyTorch at nationals.

The theory reference; the optimization and regularization chapters help most.

How to debug a network that won't learn.

Deep learning top-down: you train real models from the first lesson.

An interactive book: the theory right next to the code that implements it, in PyTorch.

The RL book, free: MDP, Bellman and Q-learning cover the module.

The sources of bias and fairness metrics, for the ethics part.

The classic UCL/DeepMind course: MDPs, Bellman, control, in ten lectures.

The practical half: you train real agents with Stable Baselines3.

Explained solutions from ONIA Winter Warmup 2025.

The official platform: submit and watch the leaderboard.

The judge that grades submissions. Make an account early.

Your first submission, from reading the data to the file you send in.

The official site: rules, calendar, results, resources.

What's required at each stage, with per-topic links.

A Romanian-language NLP hackathon, in teams.

Problems from the international olympiad.

The real local-stage problems. Solve them against the clock.

The county-stage problems, for calibration.