Neural network reinforcement learning
AssigneeInternational Business Machines Corporation
InventorsSakyasingha Dasgupta, Takayuki Osogami
Life of this patent
Estimating the term…
Abstract
A computer-implement method and an apparatus are provided for neural network reinforcement learning. The method includes obtaining, by a processor, an action and observation sequence. The method further includes inputting, by the processor, each of a plurality of time frames of the action and observation sequence sequentially into a plurality of input nodes of a neural network. The method also includes updating, by the processor, a plurality of parameters of the neural network by using the neural network to approximate an action-value function of the action and observation sequence.
Claims16
A computer program product for neural network reinforcement learning, the computer program product comprising a non-transitory computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to perform a method comprising:
- obtaining, by a processor, an action and observation sequence;
- inputting, by the processor, each of a plurality of time frames of the action and observation sequence sequentially into a plurality of input nodes of a First-In First-Out (FIFO)-based neural network;
- and
- updating, by the processor, a plurality of parameters of the neural network by using the FIFO-based neural network to approximate an action-value function of the action and observation sequence,
- wherein values from nodes for a same action are sequentially stored in a FIFO queue between neurons of the FIFO-based neural network.
An apparatus, comprising:
- a processor, configured to:
- obtain an action and observation sequence;
- input each of a plurality of time frames of the action and observation sequence sequentially into a plurality of input nodes of a First-In First-Out (FIFO)-based neural network;
- and
- update a plurality of parameters of the neural network by using the FIFO-based neural network to approximate a function for determining a subsequent action based on the action and observation sequence,
- wherein values from nodes for a same action are sequentially stored in a FIFO queue between neurons of the FIFO-based neural network.
Dependencies are inferred from the claim text.