🧠 [[Data]] [[knowledge graph]]

A living [[ontology]] of [[data]] science, ML, and AI built as an Obsidian vault with typed semantic relationships.

---

Not a glossary, a graph. Each concept carries typed relationships in YAML frontmatter (broader, uses, subclass of, …), forming a navigable semantic network. 73 root concepts branch into 8 nesting levels, connected by 7,800+ wikilinks.

📈 Growth

	Count
Rich (200+ words)	109
Medium (50–199 words)	112
Stubs (< 50 words)	359
Empty (link targets)	292

🏗️ Structure

Each concept is a markdown file with YAML frontmatter defining typed relationships and wiki-links connecting to related terms:

# Knowledge graph.md
---
uses:
  - "[[Ontology]]"
  - "[[Taxonomy]]"
subclass of:
  - "[[knowledge base]]"
  - "[[labeled directed graph]]"
studied in:
  - "[[Ontology engineering]]"
---
A knowledge graph represents structured domain knowledge
as entities connected by typed relationships...

🗺️ Topics

🤖 AI & LLMs Agents, RAG, transformers, prompt engineering, LLM evaluation	📊 Machine Learning Algorithms, training, evaluation metrics, feature engineering	🔧 Data Engineering Pipelines, data lineage, ETL, data quality
🧬 Deep Learning Neural networks, CNNs, transformers, attention, backpropagation	🔗 Knowledge Representation Ontologies, taxonomies, knowledge graphs, SKOS	⚖️ Fairness & Privacy Bias, FERPA, PII management, interpretability
🏛️ Infrastructure Cloud, containers, HTTP, authentication, databases	📐 Math & Statistics Probability, graph theory, vector spaces, combinatorics	💼 Process & Product Design thinking, CRISP-DM, agile, management

🎯 Use as a Data Ontology

Use case	How
GraphRAG	Navigate concepts through semantic connections instead of keyword matching
Data catalog enrichment	Import as standardized terminology for hierarchical classification
LLM grounding	Reduce hallucination through explicit relationship constraints
Onboarding	Interactive learning with concept definitions and relationship paths
Metadata schema design	Use partOf, uses, subclass of as schema blueprints
Knowledge graph construction	Export to Neo4j / RDF by parsing frontmatter into typed edges
Concept disambiguation	Canonical definitions that resolve cross-team terminology drift

🔬 Formal Reasoning

scripts/reason.py uses Semantica's Datalog engine for purely logical inference — no LLM involved.

pip install -r requirements.txt

python scripts/reason.py infer          # derived ancestor chains, symmetric related, inverses
python scripts/reason.py check          # cycles, broken links, S27 violations, orphans
python scripts/reason.py query "ancestor(AI agent, ?X)"   # query with real note titles
python scripts/reason.py report         # write stats/reasoning_report.md

9 Datalog rules encode SKOS semantics: transitive broader closure, broader↔narrower inverses, related symmetry, cycle detection, and SKOS S27 disjointness.

🌳 Taxonomy — SKOS Broader/Narrower Hierarchy

The graph encodes SKOS-style hierarchical relationships via broader: in YAML frontmatter. 701 of 866 concepts (80.9%) are classified — 73 root concepts, up to 8 nesting levels.

Browse full taxonomy (73 root concepts → 701 classified concepts)

• agent observability
  • agentOps
  • LLM Observability
    • Langfuse
• Agile approach
  • Daily stand-ups
  • Demo sessions
  • Product owner
  • Sprint planning
• AI
  • AI engineering
  • AI Hierarchy of needs
  • AI Proof detection model
  • AI stack
  • Artificial Intelligence (AI)
    • agentic System
      • action module · agency · agent harness · agent identity · agent role type · Autonomy · claude code · claude managed agents · comparag agent · context mode · Human-in-the-loop · ReAct data agent · ReAct mode · search module · Subagent · Tool arena · Tool Assignment Gate · Toolkits · tools
      • Model Context Protocol (MCP) → MCP host · MCP server
    • compound AI system
    • computer vision → Facial recognition
    • connexionism · Expert systems · Narrow AI · Symbolism
    • Natural Language Processing
      • bag of words · Sentiment analysis · token · Tokenizer
      • Embeddings → embedding · Embeddings models · latent space · vectorization · Word embeddings
      • Language modeling
        • autoregressive language models → GPT2
        • Large Language Model
          • AI prompt → chain-of-thought · Prompt Identity · Prompt Optimization Techniques · Prompt templates · system prompt · …
          • RAG (Retrieval-Augmented Generation) → Retriever · HyDE · …
          • chat engines · Hallucination · In-Context Learning · LLaMA · Ollama · …
        • masked language models → BERT → camemBERT
  • claude cowork · Man-Computer Symbiosis · The Shift from Models to Compound AI Systems · …
• algorithm
  • algorithm selection · Greediness · Non-parametric algorithms · selection
• automation
  • automation framework · office automation
• Business understanding
  • define success · Problem framing
• Centrality
  • betweenness centrality · closeness centrality · Degree centrality · Eigenvector centrality
• clause
  • clause classification heuristic · exclusion clause · force majeure · non-compete clause
• Clustering
  • Elbow method · K-Means Clustering
• Combinatorics · combinatorics
  • graph theory · probability · combination · Unary operation
• Confusion Matrix
  • False Negative (FN) · False Positive (FP) · True Negative (TN) · True Positive (TP)
• Containerization (computing)
  • container · Docker → Docker images
• contract
  • compliant privacy policy · liability · model contract · Rider
• control theory
  • Damping · Inertia · James Clerk Maxwell · state observer
• Deep learning
  • artificial neural network → artificial neuron · autoencoder · backpropagation · CNN · multilayer perceptron · ReLU · Sigmoid · Weights · …
  • foundation model → multimodal model · tabular foundation model → TabPFN
  • Keras · Transfer learning
  • Transformers → attention weights · Decoder-only · Encoder-only
• elicitation
  • Dialogue · elicitation interview · Socratic → Socrates
• evaluation
  • Cross validation · Evaluation metrics → accuracy · F1 Score · Precision · Recall · ROC · Regression Error Metrics · …
• Fairness
  • bias → Feedback loop bias · Historical bias
  • Data privacy → Data privacy laws · Privacy by design · Sensitive Information · …
  • Transparency → Interpretability → LIME · SHAP
• Graph theory
  • adjacency matrix · Edges · node · Heterogeneous graph · Network graph · …
• Infrastructure
  • authentication protocol → OAuth · SSH Key
  • cloud computing · Datawarehouse · Deploy · Docker · HTTP · NoSQL database · Virtualization · …
• knowledge graph
  • CYPHER · context graph · graph hop · Properties · Text2Cypher · …
• knowledge organization system
  • concept · controlled vocabulary · Ontology · taxonomy · thesaurus · …
• machine learning
  • supervised learning → Decision tree · KNN · Naive Bayes · SVM
  • Unsupervised Learning · Reinforcement Learning · Federated learning
  • Hyperparameters · loss function · Regularization
  • Optimizer algorithm → Gradient descent
  • ML system → cold start problem · Reproducibility · …
• machine learning projects
  • CRISP-DM Process → Data preparation · Deployment · modeling · Problem definition · …
• ML models
  • Linear models · Neural network models · feature · prediction · …
• Model fit → Overfitting · Underfitting
• Model maintenance → concept drift · Data drift · Model registry · Retraining · …
• Model tuning → Fine-tuning · LoRA · Pruning · Quantization
• Probability → bayes' theorem · confidence interval · variance · …
• Process
  • application development → backend · Frontend · Programming language · DevOps · …
  • data → Data quality · data type · metadata · …
  • Design thinking · knowledge · Management · …
• Shareholders agreement · contract · negotiation
• spreadsheet → excel · google sheet · named-ranges · …
• Vector space → Distance · Cosine similarity · Tensors · vector · …
• version control → Git · Repository · …
• + 20 more root concepts: Data augmentation, Data science team, Decision support, Dividual, Graph algorithm, heuristic, Kernel, Metric, Modeling, pattern, Pipelines, Product, Prototyping, regression, risk, Schema design, Search, semantics, Stochastic processes, systems theory, …

🚀 Getting Started

Explore — Open the graph/ folder in Obsidian and use Graph View to navigate visually.

Export — Convert to other formats:

Target	Method
Neo4j	Parse frontmatter into Cypher CREATE statements
RDF/OWL	Map broader: → skos:broader, relationship types → predicates
JSON-LD	Export as linked data for web interoperability

---

_{Built with Obsidian · Reasoned with Semantica · Stats updated daily via [[GitHub]] Actions}