|
| 1 | +# `petersen_brd_right_cosets.py` documentation |
| 2 | + |
| 3 | +## Location |
| 4 | +- Script: `experiments/petersen_brd_right_cosets.py` |
| 5 | + |
| 6 | +## Purpose |
| 7 | +This script computes the exact mod-2 boundary-rank distribution (BRD) for the Petersen graph under the vertical simplicial boundary used in the rank-shifted bisimplicial DSTM / total-decalage model. |
| 8 | + |
| 9 | +For each relabeling of vertices, it computes the rank over `F_2` of the vertical boundary matrix and tabulates how often each rank occurs. |
| 10 | + |
| 11 | +## Mathematical Model |
| 12 | +- Graph: Petersen graph with `N = 10` vertices and `15` edges. |
| 13 | +- Vertex model: vertices are the 2-subsets of `{0,1,2,3,4}` (Kneser labeling). |
| 14 | +- Edge rule: two vertices are adjacent iff the two 2-subsets are disjoint. |
| 15 | + |
| 16 | +Adjacency matrices are encoded as integer row bitmasks. |
| 17 | + |
| 18 | +### Vertical Faces and Boundary |
| 19 | +For an `N x N` adjacency matrix `T`, the vertical face `d_k^v(T)` is principal deletion of row/column `k` with reindexing. |
| 20 | + |
| 21 | +The mod-2 vertical boundary is: |
| 22 | + |
| 23 | +`partial_v(T) = sum_{k=0}^{N-1} d_k^v(T)` over `F_2`. |
| 24 | + |
| 25 | +In code, this is implemented as XOR of all principal deletions. |
| 26 | + |
| 27 | +### Boundary Rank |
| 28 | +The script computes `rank_{F_2}(partial_v(T))` using bit-level Gaussian elimination on row bitmasks. |
| 29 | + |
| 30 | +## Group-Theoretic Reduction |
| 31 | +Let `A` be the Petersen adjacency matrix and let relabelings act by: |
| 32 | + |
| 33 | +`A^p = P_p A P_p^T`, with convention `p[i] = new label of old vertex i`. |
| 34 | + |
| 35 | +The script builds `Aut(P)` from the induced `S_5` action on 2-subsets and obtains `|Aut(P)| = 120`. |
| 36 | + |
| 37 | +Because `A^{p sigma} = A^p` for `sigma in Aut(P)`, the boundary-rank value is constant on right cosets of `S_10 / Aut(P)`. |
| 38 | + |
| 39 | +So it is exact to enumerate a right transversal only: |
| 40 | +- `|S_10 / Aut(P)| = 10! / 120 = 30240` representatives. |
| 41 | + |
| 42 | +SymPy is used for: |
| 43 | +- permutation groups, |
| 44 | +- right coset transversal generation, |
| 45 | +- induced permutation handling. |
| 46 | + |
| 47 | +## Algorithm |
| 48 | +1. Build Petersen adjacency row bitmasks from the Kneser model. |
| 49 | +2. Build `Aut(P)` as induced permutations of 2-subsets. |
| 50 | +3. Enumerate a right transversal of `S_10 / Aut(P)`. |
| 51 | +4. For each representative: |
| 52 | + - relabel adjacency matrix, |
| 53 | + - compute `partial_v`, |
| 54 | + - compute `rank_{F_2}`. |
| 55 | +5. Accumulate: |
| 56 | + - unweighted counts over coset representatives, |
| 57 | + - weighted counts over all `10!` labelings by multiplying each bucket by `|Aut(P)|`. |
| 58 | + |
| 59 | +## Built-In Checks |
| 60 | +Before BRD computation, `verify_basic_facts()` asserts: |
| 61 | +- `|Aut(P)| = 120`, |
| 62 | +- transversal size equals `10! / 120`. |
| 63 | + |
| 64 | +## Output |
| 65 | +The script prints: |
| 66 | +- graph size data, |
| 67 | +- automorphism-group size, |
| 68 | +- right-coset count, |
| 69 | +- BRD over right-coset representatives, |
| 70 | +- BRD over all `10!` labelings, |
| 71 | +- total-labelings consistency check. |
| 72 | + |
| 73 | +Current output: |
| 74 | +- Unweighted (right-coset transversal): |
| 75 | + - rank `4`: `376` |
| 76 | + - rank `6`: `8496` |
| 77 | + - rank `8`: `21368` |
| 78 | +- Weighted (all labelings): |
| 79 | + - rank `4`: `45120` |
| 80 | + - rank `6`: `1019520` |
| 81 | + - rank `8`: `2564160` |
| 82 | +- Total check: `3628800` (equals `10!`). |
| 83 | + |
| 84 | +## Logging |
| 85 | +This script does not configure file logging. |
| 86 | +All reporting is printed to standard output. |
| 87 | + |
| 88 | +## Run |
| 89 | +From repository root: |
| 90 | + |
| 91 | +```bash |
| 92 | +.venv/Scripts/python.exe experiments/petersen_brd_right_cosets.py |
| 93 | +``` |
0 commit comments