src/HOL/Tools/Sledgehammer/sledgehammer_mash.ML
Thu, 26 Jun 2014 13:33:08 +0200 blanchet adaptive k-NN
Tue, 24 Jun 2014 15:08:19 +0200 blanchet optimized traversal of proof terms by skipping bad apples (e.g. full_exhaustive_int'.pinduct)
Tue, 24 Jun 2014 14:56:08 +0200 blanchet minor table access optimization
Tue, 24 Jun 2014 12:36:45 +0200 blanchet optimize log
Tue, 24 Jun 2014 12:35:57 +0200 blanchet enable TF-IDF
Tue, 24 Jun 2014 12:35:49 +0200 blanchet added another experimental engine
Tue, 24 Jun 2014 12:35:43 +0200 blanchet tweaked experimental setup
Tue, 24 Jun 2014 08:19:58 +0200 blanchet use strings to communicate with external process, to ease debugging
Tue, 24 Jun 2014 08:19:55 +0200 blanchet added experimental MaSh engine
Fri, 20 Jun 2014 09:55:31 +0200 blanchet changed default MaSh parameters based on (in vitro) evaluation
Wed, 18 Jun 2014 17:42:24 +0200 blanchet more MaSh engine variations, for evaluations
Wed, 18 Jun 2014 17:42:24 +0200 blanchet split parameter into two
Wed, 18 Jun 2014 15:23:40 +0200 blanchet more generous formula -- there are lots of duplicates out there
Wed, 18 Jun 2014 14:19:42 +0200 blanchet automatically learn MaSh facts also in 'blocking' mode
Mon, 02 Jun 2014 11:59:50 +0200 blanchet add option to keep duplicates, for more precise evaluation of relevance filters
Fri, 30 May 2014 16:00:54 +0200 blanchet made 'Kuehlwein-style' be really like Python code, we now think
Fri, 30 May 2014 15:15:41 +0200 blanchet make SML code closer to Python code when 'nb_kuehlwein_style' is true
Fri, 30 May 2014 14:43:06 +0200 blanchet added sleep to give time for the server to shut down -- this is a hack, but it's only in experimental code that will hopefully soon go away
Fri, 30 May 2014 12:27:51 +0200 blanchet added another way of invoking Python code, for experiments
Fri, 30 May 2014 12:27:51 +0200 blanchet make SML naive Bayes closer to Python version
Fri, 30 May 2014 12:27:51 +0200 blanchet more work on exporter
Fri, 30 May 2014 12:27:51 +0200 blanchet extend exporter with new versions of MaSh
Wed, 28 May 2014 17:42:36 +0200 blanchet more generous max number of suggestions, for more safety
Wed, 28 May 2014 17:42:34 +0200 blanchet changed MaSh to use SML version instead of Python version of naive Bayes by default (i.e. if MASH=yes in the settings, or 'fact_filter=mash' with no other explicit setting)
Wed, 28 May 2014 17:42:33 +0200 blanchet export more ML functions, for experimentation
Wed, 28 May 2014 14:02:49 +0200 blanchet disabled IDF for now -- empirical evidence points the wrong way (as usual)
Wed, 28 May 2014 13:31:44 +0200 blanchet tuning
Wed, 28 May 2014 13:02:47 +0200 blanchet tuning
Wed, 28 May 2014 12:34:26 +0200 blanchet optimized computation
Wed, 28 May 2014 10:04:28 +0200 blanchet enabled IDF for naive Bayes ML
Wed, 28 May 2014 10:03:14 +0200 blanchet tuning
Wed, 28 May 2014 09:44:14 +0200 blanchet repaired subscript problem in SML kNN
Wed, 28 May 2014 09:38:39 +0200 blanchet tuning
Wed, 28 May 2014 03:10:30 +0200 blanchet always remove duplicates in meshing + use weights for Naive Bayes
Tue, 27 May 2014 17:48:11 +0200 blanchet updated naive Bayes
Mon, 26 May 2014 14:15:48 +0200 blanchet renamed 'MaSh' option
Fri, 23 May 2014 14:12:20 +0200 blanchet automatically reload state file when it changes on disk
Thu, 22 May 2014 14:27:43 +0200 blanchet avoid slow inspection of proof terms now that dependencies are stored in 'state'
Thu, 22 May 2014 13:46:49 +0200 blanchet properly mark relearns as dirty
Thu, 22 May 2014 13:07:53 +0200 blanchet disable weights that cause more harm than they help in kNN
Thu, 22 May 2014 13:07:52 +0200 blanchet add self dependency to naive Bayes
Thu, 22 May 2014 13:07:51 +0200 blanchet make MaSh Python the default when passing 'fact_filter = mash' without enabling the 'maSh' Isabelle system option
Thu, 22 May 2014 04:12:06 +0200 blanchet reverted '|' features in MaSh -- these sounded like a good idea but never really worked
Thu, 22 May 2014 03:29:35 +0200 blanchet until naive Bayes supports weights, don't incorporate 'extra' low-weight features
Wed, 21 May 2014 14:09:43 +0200 blanchet added comment
Tue, 20 May 2014 22:28:44 +0200 blanchet added naive Bayes ML implementation, due to Cezary Kaliszyk (like k-NN)
Tue, 20 May 2014 22:28:08 +0200 blanchet added Isabelle system option 'mash'
Tue, 20 May 2014 16:31:39 +0200 blanchet more flexible environment variable
Tue, 20 May 2014 16:11:37 +0200 blanchet tuning
Tue, 20 May 2014 09:57:10 +0200 blanchet implemented MaSh/SML hints
Tue, 20 May 2014 09:38:39 +0200 blanchet better way to take invisible facts into account than 'island' business
Tue, 20 May 2014 02:47:23 +0200 blanchet cleaner handling of learned proofs
Tue, 20 May 2014 00:13:31 +0200 blanchet implemented learning of single proofs in SML MaSh
Mon, 19 May 2014 23:43:53 +0200 blanchet take weights into consideration in knn
Mon, 19 May 2014 23:43:53 +0200 blanchet added SML implementation of MaSh
Mon, 19 May 2014 23:43:53 +0200 blanchet started work on MaSh/SML
Mon, 19 May 2014 23:43:53 +0200 blanchet tune
Mon, 19 May 2014 23:43:53 +0200 blanchet store all MaSh data on the Isabelle side, in preparation for replacing 'mash.py' with ML solution
Mon, 19 May 2014 13:53:58 +0200 blanchet hide more consts to beautify documentation
Thu, 27 Mar 2014 17:12:40 +0100 wenzelm clarified Isabelle/ML bootstrap, such that Execution does not require ML_Compiler;
Sat, 22 Mar 2014 18:19:57 +0100 wenzelm more antiquotations;
Fri, 21 Feb 2014 00:09:56 +0100 blanchet adapted to renaming of datatype 'cases' and 'recs' to 'case' and 'rec'
Mon, 03 Feb 2014 15:33:18 +0100 blanchet tuning
Fri, 31 Jan 2014 10:23:32 +0100 blanchet renamed many Sledgehammer ML files to clarify structure
Fri, 31 Jan 2014 10:23:32 +0100 blanchet renamed ML file
Thu, 19 Dec 2013 13:43:21 +0100 blanchet made timeouts in Sledgehammer not be 'option's -- simplified lots of code
Wed, 18 Dec 2013 16:50:14 +0100 blanchet tuning
Mon, 09 Dec 2013 04:44:59 +0100 blanchet disable generalization in MaSh until it is shown to help
Mon, 09 Dec 2013 04:03:30 +0100 blanchet generate problems with type classes
Mon, 09 Dec 2013 04:03:30 +0100 blanchet more reasonable default weight
Tue, 19 Nov 2013 18:34:04 +0100 blanchet tuning
Thu, 14 Nov 2013 15:57:48 +0100 blanchet have MaSh support nameless facts (i.e. proofs) and use that support
Fri, 18 Oct 2013 00:05:31 +0200 blanchet make sure add: doesn't add duplicates, and works for [no_atp] facts
Thu, 17 Oct 2013 20:49:19 +0200 blanchet generate a comment storing the goal nickname in "learn_prover"
Thu, 17 Oct 2013 20:20:53 +0200 blanchet clarified message
Thu, 17 Oct 2013 02:29:49 +0200 blanchet choose facts to reprove more randomly, to avoid getting stuck with impossible problems at first
Thu, 17 Oct 2013 01:04:00 +0200 blanchet if slicing is disabled, don't enforce last slice's "max_facts", but rather the maximum "max_facts"
Thu, 17 Oct 2013 01:03:59 +0200 blanchet remove overloading of "max_facts" -- it already controls the number of facts passed to ATPs for 'learn_prover'
Tue, 15 Oct 2013 16:14:52 +0200 blanchet improved duplicate detection in "build_name_tables" by ensuring that the earliest occurrence of a duplicate (if it exists) gets picked as the canonical instance
Sun, 13 Oct 2013 21:36:26 +0200 blanchet more prominent MaSh errors
Thu, 10 Oct 2013 08:23:57 +0200 blanchet repaired confusion between the stated and effective fact filter -- the mismatch could result in "Match" exceptions
Thu, 10 Oct 2013 01:17:37 +0200 blanchet simplify fudge factor code
Wed, 09 Oct 2013 16:07:33 +0200 blanchet parallelize MeSh
Wed, 09 Oct 2013 15:39:34 +0200 blanchet use same relevance filter for ATP and SMT solvers -- attempting to filter out certain ground instances of polymorphic symbols like + and 0 has unexpected side-effects that lead to incompletenesses (relevant facts not being selected)
Wed, 09 Oct 2013 09:47:59 +0200 blanchet optimized built-in const check
Fri, 04 Oct 2013 17:00:35 +0200 blanchet more tracing
Fri, 04 Oct 2013 16:51:26 +0200 blanchet more thorough spying
Fri, 04 Oct 2013 12:59:18 +0200 blanchet removed pointless special case
Tue, 01 Oct 2013 15:02:12 +0200 blanchet removed spurious save if nothing needs to bee learned
Tue, 24 Sep 2013 12:11:53 +0200 blanchet honor MaSh's zero-overhead policy -- no learning if the tool is disabled
Mon, 23 Sep 2013 09:48:06 +0200 blanchet provide a way to override MaSh's port from configuration file
Fri, 20 Sep 2013 22:39:30 +0200 blanchet reduce the number of emitted MaSh commands (among others to facilitate debugging)
Fri, 20 Sep 2013 22:39:30 +0200 blanchet MaSh tweaks to facilitate debugging
Thu, 12 Sep 2013 15:14:54 +0200 blanchet more robust approach to avoid Python byte code -- environment variables get inherited by subprocesses
Thu, 12 Sep 2013 11:05:19 +0200 blanchet when pouring in extra features into the goal, only consider facts from the current theory -- the bottom 10 facts of the last import might be completely unrelated
Thu, 12 Sep 2013 10:40:53 +0200 blanchet minor fixes
Thu, 12 Sep 2013 10:05:10 +0200 blanchet invoke Python with "no bytecode" option to avoid litering Isabelle source directory with ".pyc" files (which can be problematic for a number of reasons)
Mon, 26 Aug 2013 12:14:40 +0200 blanchet reverted 6c5e7143e1f6; took a better look at evaluation data this time
Mon, 26 Aug 2013 09:07:32 +0200 blanchet tuned fudge factor in light of evaluation
Fri, 23 Aug 2013 16:51:53 +0200 blanchet repaired num_extra_feature_facts + tuning
Fri, 23 Aug 2013 15:49:27 +0200 blanchet minor MaSh fix
Fri, 23 Aug 2013 15:07:32 +0200 blanchet eliminated some needless MaSh features
Fri, 23 Aug 2013 14:19:57 +0200 blanchet tuned output
Fri, 23 Aug 2013 14:04:08 +0200 blanchet better tracing + honor blocking better
Fri, 23 Aug 2013 13:30:25 +0200 blanchet learn new facts on query if there aren't too many of them in MaSh
Thu, 22 Aug 2013 23:03:23 +0200 blanchet increase relevance of unknown proximate facts
Thu, 22 Aug 2013 23:03:21 +0200 blanchet fixed subtle bug with "take" + thread overlord through
Thu, 22 Aug 2013 16:03:13 +0200 blanchet have kill_all also kill MaSh server + be paranoid about reloading after clear_state, to allow for easier experimentation
Thu, 22 Aug 2013 12:16:56 +0200 blanchet take chained and proximate facts into consideration when computing MaSh features
Thu, 22 Aug 2013 12:12:52 +0200 blanchet pour extra features from proximate facts into goal, in exporter
Thu, 22 Aug 2013 08:42:27 +0200 blanchet tuning
Wed, 21 Aug 2013 16:21:37 +0200 blanchet improve weight computation for complex terms
Wed, 21 Aug 2013 15:34:51 +0200 blanchet improved support for MaSh server
Wed, 21 Aug 2013 15:18:06 +0200 blanchet get rid of some silly MaSh features
Wed, 21 Aug 2013 14:54:25 +0200 blanchet weight MaSh constants by frequency
Wed, 21 Aug 2013 09:25:40 +0200 blanchet only generate feature weights for queries -- they're not used elsewhere
Wed, 21 Aug 2013 09:25:40 +0200 blanchet take out dangerous feature, now that all updates are permanent
Wed, 21 Aug 2013 09:25:40 +0200 blanchet use new MaSh command-line arguments
Wed, 21 Aug 2013 09:25:40 +0200 blanchet shutdown MaSh server
Tue, 20 Aug 2013 14:36:22 +0200 blanchet adapted ML code to new version of MaSh tool
less more (0) -120 tip