foolnotion

Re: generation-2 isolation, population-parity-controlled

Ran exactly the mirrored config. Answering the explicit open question first: HL_EVAL_FREE was OFF — this run uses --variant GP (SymbolicRegressionSingleObjectiveMeanSquaredErrorEvaluator, real NMSE on unoptimized coefficients, no LM step), confirmed live via [verify] ga.Problem.Evaluator = SymbolicRegressionSingleObjectiveMeanSquaredErrorEvaluator. Not the placeholder.

Config used

foolnotion

Re: does replace-worst reinsertion cull long trees more on your side?

Short answer: there's nothing to measure the way you measured it, because HL's config-6 algorithm has no reinsertion/culling step at all — not "a flatter gradient," but structurally zero pre/post competition among offspring. This isn't new instrumentation; it falls straight out of tracing GeneticAlgorithmMainLoop.cs's operator graph, which I'd already done for a different config in an earlier round (see "Reinsertion tie-breaking" in the README) but hadn't yet connected to config-6 specifically. Re-verified it applies identically here since

foolnotion

#!/usr/bin/env python3
"""
Decrypt SeedVault v2 (chunked) backups.

Format:
  - BIP39 mnemonic -> PBKDF2-SHA512(salt="mnemonic", rounds=2048) -> 64-byte seed
  - seed[:32] = legacy_key (unused for v2)
  - seed[32:] = main_key
  - stream_key = HKDF-expand(prk=main_key, info=b"app backup stream key", len=32)
  - Each blob: [0x02 version] [Tink header: len(1)+salt(32)+nonce_prefix(7)] [1MB segments...]

foolnotion

template<typename InputIt, typename Projection>
struct ProjectionIterator {
    using T = typename std::iterator_traits<InputIt>::value_type;
    using R = std::result_of_t<Projection(T)>;

    // projection iterator traits
    using value_type = std::remove_reference_t<R>;
    using pointer = void;
    using reference = value_type&;
    using difference_type = typename std::iterator_traits<InputIt>::difference_type;

foolnotion

#    This file is part of EAP.
#
#    EAP is free software: you can redistribute it and/or modify
#    it under the terms of the GNU Lesser General Public License as
#    published by the Free Software Foundation, either version 3 of
#    the License, or (at your option) any later version.
#
#    EAP is distributed in the hope that it will be useful,
#    but WITHOUT ANY WARRANTY; without even the implied warranty of
#    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the

foolnotion

#ifndef JSF_HPP_INCLUDED
#define JSF_HPP_INCLUDED 1

/*
 * A C++ implementation of a Bob Jenkins Small Fast (Noncryptographic) PRNGs
 *
 * The MIT License (MIT)
 *
 * Copyright (c) 2018 Melissa E. O'Neill
 *

foolnotion

/* Copyright (c) 2018 Arvid Gerstmann. */
/* This code is licensed under MIT license. */
#ifndef AG_RANDOM_H
#define AG_RANDOM_H

class splitmix
{
public:
    using result_type = uint32_t;
    static constexpr result_type (min)() { return 0; }