AI-ordlista
Den kompletta ordlistan över AI
SAT Solver
Algorithm that determines whether a propositional boolean formula has a variable assignment that makes it true. Fundamental for solving decision problems in program synthesis.
SMT Solver
Extension of SAT solvers that integrates mathematical theories like arithmetic, arrays, and datatypes. Allows for solving more complex constraints in program synthesis.
Logical Constraints
Mathematical formulas expressing the properties and behaviors the generated program must respect. Serve as a bridge between user specifications and the automatic search for solutions.
Constraint-Driven Synthesis
Paradigm where specifications are transformed into a system of constraints solved by automatic solvers. Generates programs that formally guarantee the fulfillment of requirements.
Problem Encoding
Process of translating high-level specifications into logical formulas understandable by solvers. Determines the efficiency of the constraint solving phase.
Inductive Synthesis
Technique that automatically infers programs from examples of desired input-output pairs. Combines inductive learning and constraint solving to generalize behaviors.
SKETCH
Programming language that allows specifying partial programs with holes to be automatically filled. Uses SMT solvers to find the optimal values for the holes.
SyGuS
Syntax-Guided Synthesis, a standard that formalizes synthesis problems with a specified solution grammar. Allows for controlling the structure and complexity of the generated programs.
Abstraction-refinement
Iterative strategy initially simplifying the problem then progressively refining constraints. Balances efficiency and precision in complex program synthesis.
Constraint-based learning
Hybrid method combining machine learning techniques and logical constraint solving. Accelerates synthesis by intelligently guiding the search space.
Conditional verification
Formal process validating that the synthesized program satisfies all initial specifications. Essential for ensuring correctness of automatically generated programs.
DPLL Algorithm
Davis-Putnam-Logemann-Loveland, fundamental algorithm for boolean formula satisfiability. Basis of many modern SAT solvers used in program synthesis.
Theory of combinations
Mechanism enabling SMT solvers to simultaneously handle multiple heterogeneous mathematical theories. Indispensable for modeling realistic synthesis problems.
SAT modulo theories
Formalism unifying propositional logic and mathematical theories in a single solving framework. Enables handling complex constraints in program synthesis.
Symmetry constraints
Formal properties eliminating equivalent solutions to reduce the search space. Significantly accelerate synthesis by avoiding redundant exploration.
Incremental synthesis
Approach progressively building the program by iteratively adding features and constraints. Manages complexity through problem decomposition.
Hoare Models
Formal {Precondition} Program {Postcondition} triplets for verification and synthesis of correct programs. Theoretical foundation ensuring the validity of generated programs.