reinforcement learning
Learning through consequences
Reinforcement learning is a type of machine learning where an algorithm learns by trial and error, receiving rewards for desirable outcomes and penalties for undesirable ones.
Like training a dog with treats, the system discovers which actions maximize cumulative reward over time: The AI learned to play chess through reinforcement learning, improving by playing millions of games against itself.
One who acts independently
- environmentExternal conditions affecting life
- stateCondition or mode
- actionProcess of doing
- rewardPayment for service
policy · episode · trajectory · transition · observation · return · horizon
Traveling through unknown regions
- exploitationSelfish use of others
- intrinsic motivationDesire from within oneself
- optimismTendency to expect good
Go-playing artificial intelligence
- AtariSeventies gaming pioneer
- roboticsAutomaton design field
- autonomous drivingVehicle operates independently
- resource managementAllocating available means efficiently
traffic control · drug discovery · energy management · supply chain · personalization · chatbots · chip design
Process of approaching sameness
- bootstrappingSelf-starting without external input
- overfittingExcess adaptation to samples
- generalizationDeriving principles from examples
- simulationImitation of real processes
agent
One who acts independently
- environmentExternal conditions affecting life
- stateCondition or mode
- actionProcess of doing
reward · policy · episode · trajectory · transition · observation · return · horizon
exploration
Traveling through unknown regions
- exploitationSelfish use of others
- intrinsic motivationDesire from within oneself
- optimismTendency to expect good
AlphaGo
Go-playing artificial intelligence
- AtariSeventies gaming pioneer
- roboticsAutomaton design field
- autonomous drivingVehicle operates independently
resource management · traffic control · drug discovery · energy management · supply chain · personalization · chatbots · chip design
convergence
Process of approaching sameness
- bootstrappingSelf-starting without external input
- overfittingExcess adaptation to samples
- generalizationDeriving principles from examples
learning, reinforced, enforcement, enforcing, reinforcement, enforces, enforcer, reinforcing, reinforcements