Skip to content

Learning Graph Quality Metrics Report

Overview

  • Total Concepts: 300
  • Foundational Concepts (no prerequisites, other concepts depend on them): 7
  • Terminal Nodes (nothing depends on them, but have prerequisites): 107
  • Orphaned Nodes (completely disconnected, no edges): 0
  • Concepts with Dependencies: 293
  • Average Dependencies per Concept: 2.27

Graph Structure Validation

  • Valid DAG Structure: ✅ Yes
  • Self-Dependencies: None detected ✅
  • Cycles Detected: 0

Foundational Concepts

These concepts have no prerequisites:

  • 1: Matrix Multiplication
  • 2: Matrix Transpose
  • 5: Symmetric Matrix
  • 11: Gradient Descent
  • 13: Chain Rule (Calculus)
  • 15: PyTorch Tensor
  • 21: Graph (Undirected)

Dependency Chain Analysis

  • Maximum Dependency Chain Length: 12

Longest Learning Path:

  1. Graph (Undirected) (ID: 21)
  2. Node (Vertex) (ID: 23)
  3. Edge (Link) (ID: 24)
  4. Heterogeneous Graph (ID: 38)
  5. Knowledge Graph (ID: 178)
  6. KG Entity (ID: 179)
  7. KG Triple (ID: 181)
  8. KG Completion (ID: 182)
  9. DistMult (ID: 186)
  10. ComplEx (ID: 187)
  11. RotatE (ID: 188)
  12. KG Embedding Geometry (ID: 189)

Terminal Nodes Analysis

Terminal nodes are concepts that nothing else depends on but have prerequisites. They represent natural endpoints of learning paths — culminating or specialized concepts.

  • Total Terminal Nodes: 107 (35.7% of all concepts)
  • Healthy Range: 5-40% of total concepts

Concepts at the end of learning paths:

  • 39: Multigraph
  • 48: Planar Graph
  • 54: Barabási–Albert Model
  • 64: HITS Algorithm
  • 68: Teleportation (PageRank)
  • 71: Louvain Algorithm
  • 72: Girvan-Newman Algorithm
  • 74: Normalized Cut
  • 76: BigCLAM Model
  • 79: Graphlet Degree Vector
  • 81: Weisfeiler-Lehman Kernel
  • 83: Belief Propagation
  • 84: Influence Maximization
  • 90: Katz Similarity
  • 92: Embedding Space
  • 94: Shallow Embedding
  • 99: BFS Strategy (node2vec)
  • 100: DFS Strategy (node2vec)
  • 103: LINE Embedding
  • 106: Structural Equivalence

...and 87 more

Orphaned Nodes Analysis

Orphaned nodes are completely disconnected concepts with no inbound AND no outbound edges. These indicate a quality problem — every concept should connect to the graph.

  • Total Orphaned Nodes: 0

✅ No orphaned nodes detected. All concepts are connected to the graph.

Connected Components

  • Number of Connected Components: 1

✅ All concepts are connected in a single graph.

Indegree Analysis

Top 10 concepts that are prerequisites for the most other concepts:

Rank Concept ID Concept Label Indegree
1 108 Graph Neural Network (GNN) 72
2 21 Graph (Undirected) 34
3 91 Node Embedding 20
4 26 Node Degree 18
5 25 Adjacency Matrix 17
6 17 Neural Network Layer 16
7 24 Edge (Link) 15
8 23 Node (Vertex) 12
9 178 Knowledge Graph 11
10 29 Degree Distribution 8

Outdegree Distribution

Dependencies Number of Concepts
0 7
1 46
2 159
3 64
4 17
5 5
6 1
10 1

Recommendations

  • Terminal node percentage (35.7%): Within healthy range (5-40%)
  • DAG structure verified: Graph supports valid learning progressions

Report generated by learning-graph-reports/analyze_graph.py