AI / ML / DL — Algorithms/Methods map

Author

E. Valderrama-Araya, Ph.D.

Venn diagram

AI/ML/DL Diagram.

Hierarchy of AI / ML / DL methods

  • Artificial Intelligence (AI)
    • Machine Learning (ML)
      • Deep Learning (DL)
        • CNN
        • RNN
        • Transformer
        • LLM
        • GAN
      • (ML, but not DL)
        • SVM
        • Random Forest
        • k-Means
        • PCA
        • Reinforcement Learning
    • (AI, but not ML)
      • Expert Systems
      • A* Search
      • Minimax
      • Fuzzy Logic
      • RAG (Retrieval-Augmented Generation)

Flowchart

flowchart TD
    AI[Artificial Intelligence]
    AI --> ML[Machine Learning]
    ML --> DL[Deep Learning]

    %% --- AI but not ML
    AI --> Expert[Expert Systems]
    AI --> Astar[A* Search]
    AI --> Fuzzy[Fuzzy Logic]
    AI --> Minimax[Minimax]
    AI --> RAG[RAG]

    %% --- ML but not DL
    ML --> SVM[SVM]
    ML --> RF[Random Forest]
    ML --> KMeans[k-Means]
    ML --> PCA[PCA]
    ML --> RL[Reinforcement Learning]

    %% --- DL items
    DL --> LLM[LLM]
    DL --> CNN[CNN]
    DL --> RNN[RNN]
    DL --> Transformer[Transformer]
    DL --> GAN[GAN]

    %% --- Styles
    style SVM fill:#fff3b0,stroke:#333,stroke-width:1px
    style RF fill:#fff3b0,stroke:#333,stroke-width:1px
    style KMeans fill:#fff3b0,stroke:#333,stroke-width:1px
    style PCA fill:#fff3b0,stroke:#333,stroke-width:1px
    style RL fill:#fff3b0,stroke:#333,stroke-width:1px

    style LLM fill:#b7e4c7,stroke:#333,stroke-width:1px
    style CNN fill:#b7e4c7,stroke:#333,stroke-width:1px
    style RNN fill:#b7e4c7,stroke:#333,stroke-width:1px
    style Transformer fill:#b7e4c7,stroke:#333,stroke-width:1px
    style GAN fill:#b7e4c7,stroke:#333,stroke-width:1px

Definitions

Deep Learning (DL)

  • CNN (Convolutional Neural Network): Neural network specialized for grid-like data such as images, using convolutions to capture spatial patterns.
  • RNN (Recurrent Neural Network): Neural network for sequences (text, time series) with feedback connections to retain context over time.
  • Transformer: Sequence model using self-attention (not recurrence) to capture long-range dependencies efficiently.
  • LLM (Large Language Model): Very large Transformer-based model trained on massive text corpora to predict and generate text, perform reasoning, and follow instructions.
  • GAN (Generative Adversarial Network): Generator and discriminator trained in opposition to create realistic synthetic data (images, audio, etc.).

Machine Learning (non-DL)

  • SVM (Support Vector Machine): Margin-maximizing classifier/regressor using kernels to separate classes.
  • Random Forest: Ensemble of decision trees averaged/voted to boost accuracy and reduce overfitting.
  • k-Means: Clustering algorithm that partitions data into (k) groups by minimizing within-cluster variance.
  • PCA (Principal Component Analysis): Linear dimensionality reduction projecting data onto directions of maximum variance.
  • Reinforcement Learning: Agents learn policies by maximizing cumulative reward through interaction with an environment.

AI (non-ML)

  • Expert Systems: Rule-based systems with a knowledge base and inference engine to emulate expert decision-making.
  • A* Search: Heuristic shortest-path search combining path cost and heuristic estimate.
  • Minimax: Game-tree search assuming optimal play, often paired with pruning (α-β).
  • Fuzzy Logic: Reasoning with degrees of truth rather than crisp true/false values.
  • RAG (Retrieval-Augmented Generation): System pattern that retrieves relevant external documents and feeds them to a generator (usually an LLM) to ground answers in up-to-date or domain-specific knowledge.