1 / 12

Deep Blue

Deep Blue. background. chess: “the touchstone of the intellect” machine would model thinking, some say chess problem “sharply defined” First chess-playing machine: 1760’s Maezal Chess Automaton. minimax algoritm. generate all possible moves by player and opponent a number of steps ahead

leal
Download Presentation

Deep Blue

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Deep Blue

  2. background • chess: “the touchstone of the intellect” • machine would model thinking, some say • chess problem “sharply defined” • First chess-playing machine: 1760’s Maezal Chess Automaton

  3. minimax algoritm • generate all possible moves by player and opponent a number of steps ahead • outcomes generally reside in outermost leaves of trees • depth of search -> program rating

  4. limits • initially tried to emulate humans • human skills: pattern recognition and associative memory • engineering approach, relies on computer’s strengths

  5. Tuning Mechanisms • hill climbing perform a lookahead search then adjusts the parameter • best fit function of machine’s evaluation of positions and the true values

  6. logistics • 6 move lookahead: 38^12 = 9 billion billion moves • each move lookahead worth about 400 rating points (world champions around 2,900)

  7. Deep Blue features • Hsu’s single chip move generator • Anantharaman’s singular extension algorithm

  8. Hsu’s “chess machines” • evaluates 2 million positions per second • 256 teamed together: 100 million positions per second • 70% of chip devoted to evaluating positions

  9. position evaluation • separate value for pieces on different squares • stored endgame positions • position features from grandmasters

  10. alpha-beta search • eliminates bad moves from further consideration • move ‘refuted’ if opponent can force a worst outcome than the previous estimation

  11. singular extensions • horizon problem • looks 30 or 60 moves in certain situations, checks or piece exchanges • fewer choices, deeper search

  12. Kasparov 2.5 Deep Blue 3.5 • May 3 – 11, 1997 • blunder costs previously undefeated World Champion

More Related