Network analysis of Heraclitus fragments
Analýza Hérakleitových zlomků pomocí teorie sití
Work in gradual slow progress, eventually we might wrap it as an paper. / Práce na síťové analýze probíhá šnečím tempem, jakkoli – jednoho dne doufáme sesbíraný materiál publikovat v méně divoké podobě.
1 Heraclitus
1.1 Fragments Visualization
1.1.1 Ordering fragments by fragment degree
Link between nodes indicates existence of common keyword.
Color labeling by nests
Each nest defined manually by 2006 Book, links as in previous section. Session file with colors in he_frags.cys.
Nests grouping by contracting nests & edge bundling
Sorted by nests
Sorted by nests & edge bundling
1.1.2 Hierarchical layout (“direction of flow”)
1.1.3 Hierarchical mapping (yfiles)
The Hierarchic layout algorithm portraits the precedence relation of directed graphs. Use this algorithm to highlight the main direction or flow within a directed graph.
1.1.4 Nests grouping
Group attributes layout & bundle edges
Nests grouping by contracting nests
Nests grouping by contracting nests & edge bundle
Edge weighted spring embedded layout
Small local changes to improve readability of labels and extreme points were done, all points stays in global picture as they are.
1.2 Iterative decomposition of fragments
It's clear that two nodes 1 (degree 72) & 30 (degree 50) could be cut, where to make the threshold is way less clear. Head:
Count Nest Frag
72 1 1
50 8 30
41 5 45
39 15 129
Colors legens used below:
1.2.1 Glay
1.2.2 MCL
1.2.3 MCODE
1.3 Keywords visualization & clustering
1.3.1 Edge weighted spring embedded layout
Manual local (only) changes were done to improve labels readability (and moving extreme points) it won't change the global position wrt other labels.
1.3.2 Clustering via MCODE
1.3.3 Community cluster (GLay)
1.4 Iterative decomposition of keywords
Key Count
panta 192
logos 134
anthropoi 122
We cluster the network after removing three most connected keywords from the network.
1.4.1 Community Glay
1.4.2 MCode
1.4.3 MCL
1.5 Fragments Clustering
1.5.1 No keyword unification
Simple clustering
Affinity propagation
MCode
Network organic
Autosome
1.5.2 (Sub)keywords unification
MCode
Autosome - threshold 0.01
Community glay
Organic
Affinity propagation - threshold 1.51
Average shortest path length
Manual separation
- Manual group attributes (file)
- MCode group attributes - manual reorder (file)
2 Anaxagoras
2.1 Fragments Visualization
2.1.1 Ordering fragments by fragment degree
Link between nodes indicates existence of common keyword.
2.1.2 Hierarchical layout (“direction of flow”)
2.1.3 Hierarchical mapping (yfiles)
Edge weighted spring embedded layout
2.2 Keywords visualization & clustering
2.2.1 Edge weighted spring embedded layout
2.2.2 MCODE clustering