You can edit almost every page by Creating an account and confirming your email.

Neural state machine

From EverybodyWiki Bios & Wiki

A neural state machine combines a neural network with a state machine to create more powerful neural networks.

In contrast to Recurrent neural networks, which are working with feedback, a neural state machine is based on an internal state variable. This variable stores a sequence of steps in a timeline. The principle was introduced in 1993 and has much in common with a finite state machine.[1][2]

Even if the idea is relatively new, there are some applications available, for example Natural-language understanding and vision understanding, especially visual question answering.[3] Both applications have to do with interpreting a sequence of steps over a time span. Apart from a purely technical understanding, it is possible to describe neural state machines from a philosophical standpoint. The question is whether the human brain is working with this principle in mind to process information from the environment.[4]

Neural state machines are used for grounding a domain into a world model. This is equal to a graph neural network which consists of linguistic information.[5] The difference is that a graph neural network is using message passing as a computational model.

References

  1. John Robert Burger (31 July 2009). Human Memory Modeled with Standard Analog and Digital Circuits: Inspiration for Man-made Computers. John Wiley & Sons. p. 140. ISBN 978-0-470-46419-9. Search this book on
  2. Paul Mc Kevitt (30 June 1996). Integration of Natural Language and Vision Processing: Theory and Grounding Representations. Springer Science & Business Media. pp. 191–. ISBN 978-0-7923-3944-1. Search this book on
  3. Institution of Electrical Engineers (1995). Neural Network Applications in Control. IET. pp. 30–. ISBN 978-0-85296-852-9. Search this book on
  4. Nadia Magnenat-Thalmann; Daniel Thalmann (12 December 1994). Artificial Life and Virtual Reality. Wiley. ISBN 978-0-471-95146-9. Search this book on
  5. Hudson, Drew A.; Manning, Christopher D. (2019). "Learning by abstraction: The neural state machine". arXiv:1907.03950 [cs.AI].


This article "Neural state machine" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:Neural state machine. Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.