Improving Costs and Robustness of Machine Learning Classifiers Against Adversarial Attacks via Self Play of Repeated Bayesian Games.
Prithviraj DasguptaJoseph B. CollinsMichael McCarrickPublished in: FLAIRS Conference (2020)
Keyphrases
- machine learning
- machine learning algorithms
- game playing
- minimax search
- machine learning methods
- machine learning approaches
- decision trees
- supervised classification
- games played
- support vector
- imperfect information
- feature selection
- training data
- digital image watermarking
- online game
- text classification
- semi fragile watermarking
- learning algorithm
- multi player
- training set
- naive bayes
- bayesian networks
- evaluation function
- countermeasures
- learning classifier systems
- board game
- cost sensitive learning
- learning tasks
- total cost
- data mining
- image watermarking scheme
- computer games
- reinforcement learning
- majority vote
- supervised learning
- game tree
- monte carlo
- game theoretic
- video games
- nash equilibria
- educational games
- playing games
- game mechanics
- geometric attacks
- bayesian methods
- support vector machine
- information extraction
- perfect information
- model selection
- multiple classifier systems
- game development
- classifier ensemble