Decentralized online multi-agent visual question answering

A computer system comprising:

• one or more computer processors;
• one or more computer readable storage media;
• program instructions stored on the one or more computer readable storage media for execution by at least one of the one or more processors, the program instructions comprising:
• program instructions to train a plurality of agents based on data associated with images and questions about the images;
• program instructions to determine a pair of agents of the plurality of agents based on a similarity of tasks performed by the pair of agents,
• wherein the pair of agents include processors and controllers;
• program instructions to provide an image and a question, associated with the image, to the pair of agents;
• program instructions to generate, using controllers of the pair of agents, different programs to answer the question;
• and
• program instructions to determine a particular program, of the different programs, to answer the question based on a performance of the particular program;
• and
• program instructions to execute the particular program to generate an answer to the question,
• wherein the answer is provided to a computing device via a network.

A computer program product comprising:

• one or more non-transitory computer readable storage media and program instructions stored on the one or more non-transitory computer readable storage media, the program instructions comprising:
• program instructions to train a plurality of agents based on data associated with images and questions about the images;
• program instructions to provide an image and a question, associated with the image, to a pair of agents of the plurality of agents,
• wherein the pair of agents include processors and controllers;
• program instructions to generate, using a first controller of a first agent of the pair of agents, a first program to answer the question;
• program instructions to generate, using a second controller of a second agent of the pair of agents, a second program to answer the question,
• wherein the first program and the second program are generated by the pair of agents during an inference phase;
• program instructions to determine the first program or the second program to answer the question based on a performance of the first program and a performance of the second program;
• and
• program instructions to execute the first program or the second program to generate an answer to the question,
• wherein the answer is provided to a computing device via a network.