Page Actions
Wiki Actions
User Actions
Submit This Story
Writing /alumni/bweber/.html/dokuwiki/data/cache/4/4c3cdfd686da78cff30e1731b1d663e2.i failed
Unable to save cache file. Hint: disk full; file permissions; safe_mode setting.
Writing /alumni/bweber/.html/dokuwiki/data/cache/4/4c3cdfd686da78cff30e1731b1d663e2.i failed
Unable to save cache file. Hint: disk full; file permissions; safe_mode setting.
Writing /alumni/bweber/.html/dokuwiki/data/cache/4/4c3cdfd686da78cff30e1731b1d663e2.xhtml failed

ML techniques for learning the Objective Function

Performance (RSME):

  • Linear Regression: 0.322
  • Neural Nets: 0.37
  • K-Nearest Neighbor: 0.356

Attributes

Objective function Attributes:

  • numAllies
  • numEnemies
  • health
  • healthRoot
  • healthSquare
  • attackPotential
  • distance
  • TV
  • enemyHealth
  • enemyHealthRoot
  • enemyHealthSquare
  • enemyAttackPotential
  • enemyDistance
  • enemyTV


TV(u) = (0.001*health)*(0.5 + 0.1*armor(u))*(min(1.0), damage(u))*(0.01*MaxHealth)

Linear Regression

Attribute weights

  • + 0.079 * numAllies
  • - 0.060 * numEnemies
  • + 0.791 * health
  • - 0.232 * healthRoot
  • - 0.372 * healthSquare
  • + 0.030 * attackPotential
  • - 0.083 * distance
  • + 0.383 * TV
  • - 0.731 * enemyHealth
  • + 0.139 * enemyHealthRoot
  • + 0.386 * enemyHealthSquare
  • - 0.031 * enemyAttackPotential
  • + 0.082 * enemyDistance
  • - 0.359 * enemyTV
  • + 0.012

Unit Values

K-Nearest Neighbors

Too slow!


Configuration

  • k: 15
  • measure: CamberraNumericalDistance
  • weighted vote
 
comparison_of_techniques_for_learning_objective_function_rating.txt · Last modified: 2008/05/12 22:44 by bweber     Back to top