anytime algorithm


Also found in: Wikipedia.

anytime algorithm

(algorithm)
An algorithm that returns a sequence of approximations to the correct answer such that each approximation is no worse than the previous one, i.e. the algorithm can be stopped at _any time_.

Newton-Raphson iteration applied to finding the square root of a number b is another example:

x = (x + b / x) / 2

Each new x is closer to the square root than the previous one.

Applications might include a real-time control system or a chess program that is allowed a fixed thinking time.
References in periodicals archive ?
He coauthored the papers that coined the widely used terms anytime algorithm, performance profile, and conformant planning.
This approach includes an anytime algorithm, which permits to access to solutions in real time.
Key Word: decision-making, anytime algorithm, network transportation, disturbed urban transportation network.
The basic definition of an anytime algorithm is: "an anytime algorithm is an algorithm which allows, in exchange of a longer execution time, to give results of a better quality.
an automata (ATN) allows to model the stopping point of the anytime algorithm,
an anytime agent is an agent with an anytime algorithm and which will be able to:
Solutions suggested by MASDAT: for this example, we interrupted the anytime algorithm three times with 10 seconds regular intervals.
The MASDAT under consideration includes an anytime algorithm that proposes partial solution to dysfunction that became precise in time.
1995], and specialized query languages that can selectively reorder operations for higher efficiency [Imielinski and Mannila 1996]; and (iii) efficient online reformulations of the basic technique also exist [Hidber 1999] that can terminate early, once results of the desired quality are achieved, and can be viewed as anytime algorithms [Ramakrishnan and Grama 1999] due to their interruptibility and the monotonic improvement of the quality of the answer with time.
The trade-off at the center of the article is to maximize the controller builder's response value by ensuring its timeliness, at the expense of some degradation in the quality of the controller produced, in the manner of anytime algorithms.
Optimal Schedules for Monitoring Anytime Algorithms.
His research interests include anytime algorithms, decision theory, design of autonomous agents, heuristic search, information gathering, principles of metareasoning, planning and scheduling, reinforcement learning, and resource-bounded reasoning.