wiki_research

personal research wiki
git clone https://a3nm.net/git/wiki_research/
Log | Files | Refs

commit cf818e2b20cc7577655eeeb7cfddab8d6764b957
parent c61667b82897f81b10278e480e56b17b0540a27a
Author: Antoine Amarilli <a3nm@a3nm.net>
Date:   Tue, 29 Jul 2025 15:41:27 +0200

commit with codex

Diffstat:
context_free_grammar_equivalence | 2+-
context_free_grammar_finite | 4++--
dsdnnf | 2+-
first_order_model_checking | 7++++---
language_equivalence | 2+-
mengel2025lower | 7+++++++
obdd | 3++-
unambiguous_finite_cfg | 7+++++++
8 files changed, 25 insertions(+), 9 deletions(-)

diff --git a/context_free_grammar_equivalence b/context_free_grammar_equivalence @@ -6,4 +6,4 @@ Two [CFGs] are *equivalent* if they recognize the same [formal_language] Up: [context_free_grammar], [language_equivalence] -Aliases: CFG equivalence, CFG equivalent +Aliases: CFG equivalence, CFG equivalent, equivalent CFG, grammar equivalent diff --git a/context_free_grammar_finite b/context_free_grammar_finite @@ -2,10 +2,10 @@ A *finite CFG* is a [context_free_grammar] which represents a [finite_language] -Unlike general [CFGs], for finite CFGs, there always exists a [grammar_equivalent] [unambiguous_CFG] +Unlike general [CFGs], for finite CFGs, there always exists a [grammar_equivalent] [unambiguous_finite_CFG], but doubly exponential blowup ([mengel2025lower]) Up: [context_free_grammar], [language_finite] See also: [slp], [factorized_database] -Aliases: finite context free grammar +Aliases: finite context free grammar, finite context free grammars, finite CFG, finite CFGs diff --git a/dsdnnf b/dsdnnf @@ -6,4 +6,4 @@ not closed under [complementation], cf [vinallsmeeth2024structured] by [harry] Up: [knowledge_compilation_classes], [ddnnf], [sdnnf] -Aliases: dSDNNFs +Aliases: dSDNNFs, d_SDNNFs, d_SDNNF diff --git a/first_order_model_checking b/first_order_model_checking @@ -9,14 +9,15 @@ [first_order_property_conjecture] on the hardness of FO model checking, cf [gao2018completeness] - it is a weaker version of [strong_exponential_time_hypothesis] -## Bounded variables +## On specific [logical_fragments] - [first_order_model_checking_fok] - [first_order_model_checking_fo2] -## [parameterized_complexity] +## [Computational_complexity] -[fo_model_checking_parameterized_complexity] +[FO_model_checking_complexity] +- [fo_model_checking_parameterized_complexity] Up: [model_checking] for [first_order_logic] diff --git a/language_equivalence b/language_equivalence @@ -10,6 +10,6 @@ test if two languages are the same Up: [formal_language_computational_problem], [equivalence] -See also: [query_equivalence], [language_inclusion], [regular_expression_equivalence], [automaton_equivalence] +See also: [query_equivalence], [language_inclusion], [regular_expression_equivalence], [automaton_equivalence], [CFG_equivalent] Aliases: language equivalent, language equivalence problem diff --git a/mengel2025lower b/mengel2025lower @@ -0,0 +1,7 @@ +# mengel2025lower + +by [stefan_mengel] and [harry], at [PODS_2025] + +Up: [academic_paper] + +Aliases: mengel2024lower diff --git a/obdd b/obdd @@ -5,7 +5,8 @@ A [deterministic] [bdd] with a [variable_order] - [uobdd] - [nobdd] -[synthesis]: we can take [conjunction] and [disjunction] of two OBDDs where the [variable_order] is the same +We can tractably take the [conjunction] and [disjunction] of two OBDDs where the [variable_order] is the same +- [OBDD_conjunction] Up: [bdd] diff --git a/unambiguous_finite_cfg b/unambiguous_finite_cfg @@ -0,0 +1,7 @@ +# Unambiguous finite CFG + +A [finite_CFG] which is an [unambiguous_CFG] + +Doubly exponential separation bounds in [mengel2025lower] + +Up: [context_free_grammar_finite], [unambiguous_CFG]