Skip to content

feat: finding more intercontinental data - #14

Merged
a1exxd0 merged 2 commits into
mainfrom
alex/more-intercontinental-data
Jun 7, 2026
Merged

feat: finding more intercontinental data#14
a1exxd0 merged 2 commits into
mainfrom
alex/more-intercontinental-data

Conversation

@a1exxd0

@a1exxd0 a1exxd0 commented Jun 7, 2026

Copy link
Copy Markdown
Collaborator

No description provided.

a1exxd0 and others added 2 commits June 7, 2026 04:52
Add 10 listed banks (ICBC, DBS, ICICI, KB Financial, CommBank, Westpac,
Al Rajhi, Qatar National Bank, Standard Bank, FirstRand) to the real
G-SIB roster, taking it from 38 -> 48 entities and adding four
previously-missing regions. Regenerate the committed equity-correlation
snapshot (2021-06..2024-06, 556 obs) to cover all 48 tickers so the
data_network pipeline stays offline and reproducible.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…(k=4)

Re-run the full stressed-spec hardware pipeline over the extended
48-entity real network on ibm_boston, 200k shots/cluster:

  Stage 1  derive_real_network_clusters        -> k=4 [15,13,10,10], mod 0.132
  Stage 3  run_real_cluster_mixture_stress_hw   -> 4 clusters x 200k shots (8 jobs)
  Stage 4  analyze_real_cluster_mixture_stress_hw
  Stage 5  export_joint_counts + entity_scenario_vs_hardware

Result (qualified hardware demo, consistent with the documented story):
all 48 stressed PDs clear the 2.7% floor (SNR mean 5.6); global marginal
Spearman +0.35, within-cluster |corr| recovery 72%; magnitudes stay
decoherence-biased (RMSE 0.19) and the cascade tail is overstated
(P(severe) 0.26 HW vs 0.18 faithful-sim vs 0.09 ref).

Also rebuild data/scenario_dataset/real_gsib at 48 entities (synthetics
reproduce identically) and regenerate the stress sim-preview anchor.

Generalize the downstream scripts off the hardcoded n=38/k=3/severe=19/
3-bloc assumptions: analysis is now parametric and baseline-optional
(size-matched baseline-HW absent here, so the before/after section is
omitted); export derives blocs and the severe threshold from the data;
entity-scenario figure sizes its cluster palette to k.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
@a1exxd0
a1exxd0 merged commit 26b8cde into main Jun 7, 2026
1 check failed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant