Presentation Information
[17p-A33-10]Memory Capacity of Recurrent Neural Networks with Multi-dimensional Input
〇Aozora Higashi1, Nozomi Akashi1, Akihiro Yamamoto1 (1.Kyoto Univ.)
Keywords:
Reservoir computing,Memory capacity,Recurrent neural network
In recent years, a lot of research has been done on dealing with sequential data, such as sentences, in machine learning. Reservoir computing is one of the methods for machine learning and is known for its excellent computational efficiency and memory capabilities.
In reservoir computing, previous inputs' memory capacity (MC) is formally defined. There have been a lot of analyses of the behavior of MC on reservoir computing. This research theoretically and experimentally demonstrated that MC of multi-dimensional input RNNs equal the number of nodes in the network under certain conditions. the results are expected to be a good standard for analyzing the memory capabilities of machine learning that requires multi-dimensional input.
In reservoir computing, previous inputs' memory capacity (MC) is formally defined. There have been a lot of analyses of the behavior of MC on reservoir computing. This research theoretically and experimentally demonstrated that MC of multi-dimensional input RNNs equal the number of nodes in the network under certain conditions. the results are expected to be a good standard for analyzing the memory capabilities of machine learning that requires multi-dimensional input.
Comment
To browse or post comments, you must log in.Log in