Markovian Bias of Neural-based Architectures With Feedback Connections

Abstract

Dynamic neural network architectures can deal naturally with sequential data through recursive processing enabled by feedback connections. We show how such architectures are predisposed for suffix-based Markovian input sequence representations in both supervised and unsupervised learning scenarios. In particular, in the context of such architectural… (More)
DOI: 10.1007/978-3-540-73954-8_5

7 Figures and Tables

Topics

Statistics

0102030201520162017
Citations per Year

Citation Velocity: 8

Averaging 8 citations per year over the last 3 years.

Learn more about how we calculate this metric in our FAQ.
  • Presentations referencing similar topics