Add radix-ness classifier as first step of the radix lower-bound path#85
Open
bc1cindy wants to merge 2 commits into
Open
Add radix-ness classifier as first step of the radix lower-bound path#85bc1cindy wants to merge 2 commits into
bc1cindy wants to merge 2 commits into
Conversation
Contributor
Author
|
also related with #5 |
arminsabouri
suggested changes
May 28, 2026
arminsabouri
left a comment
arminsabouri
left a comment
Contributor
There was a problem hiding this comment.
Couple questions to start with. Im curious how this will be used in the simulations. I may take a stab at integrating some of this
bc1cindy
added a commit
to bc1cindy/tx-indexer
that referenced
this pull request
Jun 1, 2026
Wires the dense-subset-sum 4-path subset-sum/coinjoin-mapping counts (brute, radix, sparse, Sasamoto) into btsim two ways: - As agent cost: WLowerBoundMetric (W lower bound via brute/sparse) and RadixMappingMetric (k×m! denomination mappings, gated by radix density) plug into the existing PrivacyBundle. With a subset_sum_threshold / radix_threshold set in ScorerConfig, each candidate plan is scored by a linear deficit to the threshold, so agents favor coinjoins with more subset-sum ambiguity. - As measurement: four_counts runs all four paths over the sim's confirmed txs recording CPU, peak RAM (alloc-probe feature), and density regime (kappa via the L bracket). Plus a correctness benchmark against the dep's oracle, a kappa/L sweep, a switch_sweep locating the sparse->Sasamoto handoff, and a CJA comparison (W vs Maurer sub-transaction mappings). Adds denominated funding (standard-denomination UTXOs in a band, via mine_denominated) so the sim can produce dense coinjoins, and fixes a latent over-spend in coin selection that the dense funding exposed (select_all now returns None unless inputs cover the target; session contribution is gated on coverage). Exploration on the fork (review-only) - distinct from the native radix metric of payjoin#84/payjoin#85. rustfmt + clippy clean.
bc1cindy
commented
Jun 1, 2026
bc1cindy
left a comment
bc1cindy
left a comment
Contributor
Author
There was a problem hiding this comment.
thanks for the review @arminsabouri
Im curious how this will be used in the simulations.
test and measure benchmarks from the simulations of the four primitives: brute force / radix / sparse convolution / Sasamoto , who are competing/complementary, with two asymmetries worth stating:
brute,radixandsparseare guaranteed lower bounds whileSasamotois an approximation- radix is the cheap, pre computable shortcut for the standard-denomination special case (it avoids solving subset-sum at all), whereas brute and sparse are the ones with real cost/limits. So in the sim we run them in parallel with guards(brute aborts before the exponential blow-up; sparse stops once the sumset table exceeds a RAM budget) and characterize each on two axes - correctness (matches the exact brute/DP count? lower bound holds? Sasamoto error?) and cost (CPU time, peak RAM). That's what lets us later compose them into per-actor profiles (a maker can spend minutes/GB; a mobile client can't) and pick the regime-switch points.
please take a look at #86, it will be better to see it
I may take a stab at integrating some of this
sounds great!
arminsabouri
suggested changes
Jun 12, 2026
bc1cindy
commented
Jun 16, 2026
bc1cindy
left a comment
bc1cindy
left a comment
Contributor
Author
There was a problem hiding this comment.
removed max_gap/max_combination_value/sumset from this PR, keeping just the per-series multiplicities. the gap/density check can come back with the lower-bound consumer
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
the radix path applies when amounts decompose into Hamming-weight-1 denominations in complementary bases (2, 3, 10). geometric progressions with non-trivial multiplicity in those bases generate dense subset-sum instances, which let the number of valid sub-transaction mappings be lower-bounded cheaply
this PR delivers a classifier that, given a tx's input and output amounts, returns per-series multiplicity counts
primitives vendored from small hamming weight denominations for coinjoins. Consts and generators match the notebook
the gap/density check is deferred to the lower-bound consumer that uses it
part of #84