Explore the algorithm family map
Follow a related lesson. Each card tells you what connects these methods.
Trees that learn
A forest combines tree votes. Boosting adds helpers that correct errors.
Decision TreeRandom ForestGradient BoostingNetworks and layers
A deep network uses more layers. Residual networks add shortcut paths.
Neural NetworkDeep LearningResNetMemory for sequences
RNN passes memory along a sequence. LSTM controls what to remember.
RNN (Recurrent Net)LSTM (Long Memory)Words and attention
These lessons share ideas about giving useful words more attention. They are related methods, not one required learning order.
TransformerBERTGPTT5 (Text-to-Text)Finding information
Word scores help rank documents. RAG combines retrieved information with an answer.
TF-IDFBM25RAG (Open Book AI)Learning and searching in games
Q-learning and DQN learn action scores. MCTS searches possible futures. AlphaGo and AlphaZero combine learned guidance with search.
Q-LearningDQNMCTS (Tree Search)AlphaGoAlphaZeroFinding objects
Object detection finds regions. Mask R-CNN also marks the object’s picture squares.
YOLO (You Only Look Once)Faster R-CNNMask R-CNNNoise and pictures
Diffusion removes noise. Stable Diffusion works using a compact picture representation.
DiffusionStable DiffusionMore methods to explore
These methods solve different kinds of problems. Being in this row does not mean one is built from another.
Linear RegressionLogistic RegressionKNN (Neighbors)Naive BayesSVM (The Divider)K-MeansDBSCANIsolation ForestCollaborative FilteringPCA (Data Shrinker)Apriori AlgorithmVision Transformer (ViT)Pose EstimationOCR (Text Reader)Word2VecWhisperWav2Vec 2.0CTC (Alignment)CLIPGANsCycleGANDeepfakesVAE (Autoencoders)U-NetAutoGPT (Agents)A* SearchARIMAProphetLDASurvival AnalysisGNN (Graph Networks)Simplex AlgorithmMulti-Armed Bandit