Method for adaptive exploration to accelerate deep reinforcement learning
AssigneeInternational Business Machines Corporation
InventorSakyasingha Dasgupta
Life of this patent
Estimating the term…
Abstract
Adaptive exploration in deep reinforcement learning may be performed by inputting a current time frame of an action and observation sequence sequentially into a function approximator, such as a deep neural network, including a plurality of parameters, the action and observation sequence including a plurality of time frames, each time frame including action values and observation values, approximating a value function using the function approximator based on the current time frame to acquire a current value, updating an action selection policy through exploration based on an ε-greedy strategy using the current value, and updating the plurality of parameters.
Claims20
A computer program product including one or more computer readable storage mediums collectively storing program instructions that are executable by a computer to cause the computer to perform operations comprising:
- inputting a current time frame of an action and observation sequence sequentially into a function approximator including a plurality of parameters, the action and observation sequence including a plurality of time frames, each time frame including action values and observation values;
- approximating a value function using the function approximator based on the current time frame to acquire a current value;
- updating an action selection policy through exploration based on an ε-greedy strategy using the current value;
- and
- updating the plurality of parameters.
A method comprising:
- inputting a current time frame of an action and observation sequence sequentially into a function approximator including a plurality of parameters, the action and observation sequence including a plurality of time frames, each time frame including action values and observation values;
- approximating a value function using the function approximator based on the current time frame to acquire a current value;
- updating an action selection policy through exploration based on an ε-greedy strategy using the current value;
- and
- updating the plurality of parameters.
An apparatus comprising:
- an inputting section configured to input a current time frame of an action and observation sequence sequentially into a function approximator including a plurality of parameters, the action and observation sequence including a plurality of time frames, each time frame including action values and observation values;
- an approximating section configured to approximate a value function using the function approximator based on the current time frame to acquire a current value;
- an action selection policy updating section configured to update an action selection policy through exploration based on an ε-greedy strategy using the current value;
- and
- a parameter updating section configured to update the plurality of parameters.
Dependencies are inferred from the claim text.