Skip to content
c96f48b
Compare
Choose a tag to compare

nle-lepton -- Find polynomial-like non-linear equations that solve for the three charged lepton masses, or any three solution masses

4.3.7.2

Bug Fixes:

  • Fixed bug in random number generatore where multiplier was initialized incorrectly
  • Literals are now explicitly created as long doubles in phase1 and phase2 functions
  • Phase 1 now reports correct solution precision in 3-term mode
  • Phase 1 solution detail now prints correct values for (C * term) fields

Known issues:

  • Phase 1 or Phase 2 will sometimes get stuck in some exponent combinations
  • Formulas in 1-smr mode can have more than three solutions. This mode should be deprecated or additional constraints added
8ae0e7e
Compare
Choose a tag to compare

nle-lepton -- Find polynomial-like non-linear equations that solve for the three charged lepton masses, or any three solution masses

4.3.7.1 Bug Fixes:

  • Fixed bug where phase 1 interesting two-term test http uploads were truncated to ~325 characters
  • Updated intended version numbers in example config files to 4.3.7
19b143a
Compare
Choose a tag to compare

nle-lepton -- Find polynomial-like non-linear equations that solve for the three charged lepton masses, or any three solution masses

4.3.7 New features:

  • Opposite sign exponents are now allowed in (1-smr) mode.
  • Added new config options 'smrfactor_gt_sm3' and 'smrfactor_lt_sm1' to restrict the reference mass / factors (if M/mr) or referenc mass * factors (if mr/M) to be greater than sm3 or less than sm1 in (1-smr) mode. Only one of these restrictions can be enabled at a time.
  • Significant performance improvement in phase 1 in (1-smr) mode for when coefficients are not close to 1.
  • Improved detection for when phase 1 needs to be reset due to coefficients diverging from the solution path

Bug fixes:

  • smrfactor_mass_user_random now filters on exact min and max limits instead of rounding to lowest exponents
  • smrfactor_mass_user_random mode now selects random values more evenly for graphs with a logaritmic axis
8e1242b
Compare
Choose a tag to compare

nle-lepton -- Find polynomial-like non-linear equations that solve for the three charged lepton masses, or any three solution masses

Bug fixes:

  • Added support for inverse solution mass ratios in (1-smr) mode. M/mr and/or mr/M are enabled with config file option 'smrfactor_mass_configuration'
  • Added config file option 'exp_inv_min' to require that at least one term has an inverse epxonent at least this large (+/-)
  • Added config file option 'exp_inv_include' to require that at least one term has this specific inverse exponent (can be + or - this value, 0 = random)
  • Added config file option 'exp_pos_enable' to independently allow positive and negative exponents
  • Added config file option 'phase1_run_continuous' to independently set run-once or run continuously. Previously this automatically followed set setting of 'phase1_random_samples_enable'
  • Added config file options 'smrfactor_mass_user_scan', '_mass_user_random', '_mass_user_min', '_mass_user_max', and '_mass_user_step' to allow scanning or random selection from a range of values instead of a fixed user-defined reference mass. This can be useful for graphing the behavior of two-term test over a wide range of reference masses in (1-smr) mode. 'phase2_enable=yes', and 'smrfactor_1minus_enable=no' are both unsupported when 'smrfactor_mass_user_random' or '_scan' is configured
  • Added 'term1 signature' to output when phase 1 is solved in (1-smr) mode. This shows whether (1-smr) is positive or negative for sm1, sm2, and sm3.
  • Added c3 to phase1 output status and results for (1-smr) mode
  • "Solving phase1..." status messages now also output for (1-smr) mode
  • Added scripts for graphing two-term test vs randomly generated reference mass in (1-smr) mode. The scripts are in the ./scripts/scan directory. Sample graphs are available at http://kevinloch.com/nle-lepton/scan/

Bug fixes:

  • Config file option 'phase1_mc_samples_limit' is now a long long integer to support > 2 billion samples
  • Removed confusing "sqrt()" label for reporting c1 and c2 when phase1 is solved in 2-term modes. The reported c1 and c2 still represent the unsquared term1 and term2 coefficients while phase1 internally works with the squared terms and coefficients. Phase 1 detail (deprecated) still shows the squared coefficients.
  • Fixed bug where same exponents in (1-smr) mode were not allowed when 'exp_neg_enable=no'
  • /aws/start.sh now uses the specified config file when launching program
  • Fixed bug where 'stuckprecision' was not set high enough to always cause it to be ignored in 2-term modes
  • Fixed bug with exponent sign checks in 3-term mode
54aed53
Compare
Choose a tag to compare

nle-lepton -- Find polynomial-like non-linear equations that solve for the three charged lepton masses, or any three solution masses

Bug fixes:

  • Fixed critical bug where some infactors and outfactors were being improperly cast to single precision. This was limiting precision of results in phase2 which could cause false negative matches against experimental uncertainties, especially when solving for v or sm2. This also caused many unnecessary duplicate formulas to be sent to phase 2.
74bd3b4
Compare
Choose a tag to compare

nle-lepton -- Find polynomial-like non-linear equations that solve for the three charged lepton masses, or any three solution masses

New features:

  • (1-smr) mode now allows even exponents on term 1, both terms having the same exponent, and negative exponents
  • Two new config file options allow user control of the two-term test "reasonable range"
  • New config file option for (1-smr) mode, 'phase1_unsolvable_checkpoint' sets the maximum samples to get phase 1 progress to increment before it is declared unsolvable. Lower values will find unsolvable formulas faster but risk missing slow to solve formulas
  • New config file option 'phase2_enable' to disable coefficient factoring and phase 2 processing. This can be used to speed up finding interesting two-term tests for later phase 2 targeted searches
  • drand48() has been completely replaced with 64-bit random number generator derived from https://www.pcg-random.org/posts/does-it-beat-the-minimal-standard.html.
  • Phase 2 core functions are now using long doubles (80-bit extended precision on x86) and solve down to <= 1.0E-16 precision

Bug fixes:

  • Fixed phase 2 result range checks. The relative difference of an output is now compared to the greater of it's uncertainty or the reference value's uncertainty
  • Widened allowed range of variables in phase2 from 20% to 50% to prevent phase 2 from getting stuck in certain cases
  • Fixed checks in phase 2 for (mw >= mz), (mz >= mh0), (mw >= mh0) conditions
  • Added checks in selectOutputs() for outfactor_rmr masses to prevent conditions where phase 2 will not converge
b8ed38c
Compare
Choose a tag to compare

nle-lepton -- Find polynomial-like non-linear equations that solve for the three charged lepton masses, or any three solution masses

New features:

  • Improved formatting of phase 1 status updates and removed/suppressed some unnecessary ones for (1-smr) mode
  • Tightened two-term test "reasonable" range for faster exclusion of uninteresting formulas in (1-smr) mode

Bug fixes:

  • Widened allowed range of variables in phase 2 from 10% to 20% to prevent phase 2 from getting stuck when using large (inverse) expnents or low phase1_match_filter
bcecc5f
Compare
Choose a tag to compare

nle-lepton -- Find polynomial-like non-linear equations that solve for the three charged lepton masses, or any three solution masses

New features:

  • New unsolvable test in (1-smr) mode that checks if any progress has been made on best ordering since last unsolvable_checkpoint. This allows config file option 'phase1_mc_samples_limit' to revert to a safety check with a much higher value
2cebf6b
Compare
Choose a tag to compare

nle-lepton -- Find polynomial-like non-linear equations that solve for the three charged lepton masses, or any three solution masses

New features:

  • srcipts/grep-logs.sh now detects and ignores duplicate results

Bug fixes:

  • Fixed critical bug where phase2 would fail to solve because nle_state->phase1_matches_count was not being reset between runs and phase2 was trying to solve stale formula terms from a previous run (often with the wrong exponent)
  • Removed duplicate upload of interesting two-term test finds
  • Updated titles in scripts/grep-logs.sh to reflect current program name
2ecb627
Compare
Choose a tag to compare

nle-lepton -- Find polynomial-like non-linear equations that solve for the three charged lepton masses, or any three solution masses

New features:

  • Support for reference mass ratios (rmr) outside radical
  • More options for controlling verbosity of status output
  • New configuration option nle_mixing_polarity for 2-term with (1-smr) mode. Status updates now indicate polarity
  • Setting phase1_random_samples_enable to 'no' now also causes phase1 to be run just once instead of continuously
  • phase1 now uses a 64-bit random number generator derived from https://www.pcg-random.org/posts/does-it-beat-the-minimal-standard.html. This may replace drand48() completely in future releases

Bug fixes:

  • Phase 2 valid result test now accounts for the relative uncertainty of the outputs when comparing their distance to the relative uncertainty of the experimental values
  • Fixed bug in srcipts/grep-logs.sh that was writing meta refresh line to wrong file
  • selectOutputs() rewritten and now checks that both the primary reference mass is floated in 2-term modes and highest uncertainty
    solution mass is also floated in (1-smr) mode
  • phase 1 has better unsolvable formula detection algorithms. Config option 'phase1_mc_samples_limit' can now be used to control one of these