Compact LaTeX formula references for theory-focused machine learning and deep learning review.
This repository contains two standalone PDF review sheets designed for quick mathematical lookup. They emphasize formulas, notation, objectives, update rules, evaluation quantities, model decisions, tensor shapes where relevant, and compact derivations without requiring readers to search across full lecture notes or textbooks.
The content follows the study-material scope covered by each sheet. It is not intended to be an exhaustive textbook-level reference or a universal catalog of every machine learning and deep learning topic. Instead, it is a focused mathematical review resource for beginners and learners who want to strengthen their theoretical foundation.
These sheets are not tutorials and do not replace complete courses or textbooks. They are compact references for revision and formula checking.
| Sheet | Focus | |
|---|---|---|
| Deep Learning Formula Cheat Sheet | Neural networks, optimization, CNNs, sequence models, attention, Transformers, and tensor shapes | Download PDF |
| Machine Learning Formula & Decision Sheet | Regression, classification, evaluation, diagnosis, tree ensembles, clustering, recommenders, and reinforcement learning | Download PDF |
A mathematical reference for machine learning theory and model-based decisions. It combines objectives, update rules, compact derivations, metric interpretation, and judgment-style checkpoints for questions where selecting the correct method matters as much as recalling the formula.
A compact formula and tensor-shape reference for deep learning. It focuses on forward computations, losses, gradient flows, optimizer updates, architecture-specific objectives, and shape rules.
| Machine Learning Formula & Decision Sheet | Deep Learning Formula Cheat Sheet |
|---|---|
| Linear, polynomial, and logistic regression | Neural-network notation and forward propagation |
| Cost functions, gradient descent, regularization, Lasso, and Elastic Net | Loss functions, backpropagation, initialization, and optimization |
| Evaluation, ROC-AUC, PR-AUC, bias/variance, and error analysis | Regularization and batch normalization |
| Neural-network foundations, decision trees, bagging, and boosting | CNNs, classic architectures, object detection, and YOLO |
| K-means, anomaly detection, and recommender systems | Face recognition and neural style transfer |
| Reinforcement learning, Bellman equations, Q-learning, and Deep Q Networks | RNN/GRU/LSTM, embeddings, Seq2Seq, attention, and Transformers |
- Keep entries compact and formula-focused.
- Prefer display mathematics for central objectives and updates.
- Define notation close to the formulas that use it.
- Include shapes where dimensions clarify the computation.
- Include short derivations where they explain an update, metric, or decision rule.
- Prefer concise tables and notes over textbook-length prose.
| Artifact | Status |
|---|---|
machine-learning-formula-decision-sheet.pdf |
Available as the current Machine Learning review sheet. |
main.pdf |
Available as the current Deep Learning formula sheet. |
Deep Learning LaTeX source in main.tex and sections/ |
Included in this repository and built by GitHub Actions. |
The current tagged Deep Learning draft is v0.2.0 - Formula Hierarchy and Core Extensions.
The currently included LaTeX source builds the Deep Learning sheet:
make pdfManual fallback:
latexmk -pdf main.texGitHub Actions builds main.pdf and uploads it as the
deep-learning-formula-cheatsheet-pdf artifact.
Each sheet follows the scope of the study materials used to prepare it. External references may be used to verify standard formulas, notation, shapes, or mathematical correctness, but the sheets are not intended to silently expand into complete textbooks. The goal is a reliable, compact mathematical review resource within the covered topic range.