Book chapter
One-Shot Ontogenetic Learning in Biomedical Datastreams
Artificial General Intelligence, pp.143-153
Lecture Notes in Computer Science, Springer International Publishing
07/15/2017
DOI: 10.1007/978-3-319-63703-7_14
Abstract
Recent technological advances in the biological and physical sciences have allowed for the generation of large quantity datasets necessary for applying deep neural networks. Despite the demonstrable success of these methods in a variety of tasks including image classification, machine translation, and query-answering, among others, their widespread adoption in biomedical research has been tempered due to issues inherent to modeling complex biological systems not readily addressed by traditional gradient-based neural networks. We consider the problem of unsupervised, general-purpose learning in biological sequence data, wherein variable-order temporal dependencies, multi-dimensionality and uncertainty in model structure and data are the norm. To successfully model and learn these dependencies in an intuitive and holistic manner, we have utilized the data abstraction of Simplicial Grammar within a Bayesian learning framework. We demonstrate that this framework offers the ability to quickly encode and integrate new information, and perform prediction tasks without extensive, iterative training.
Details
- Title: Subtitle
- One-Shot Ontogenetic Learning in Biomedical Datastreams
- Creators
- John Kalantari - University of IowaMichael A Mackey - University of Iowa
- Resource Type
- Book chapter
- Publication Details
- Artificial General Intelligence, pp.143-153
- Publisher
- Springer International Publishing; Cham
- Series
- Lecture Notes in Computer Science
- DOI
- 10.1007/978-3-319-63703-7_14
- eISSN
- 1611-3349
- ISSN
- 0302-9743
- Language
- English
- Date published
- 07/15/2017
- Academic Unit
- Roy J. Carver Department of Biomedical Engineering; Radiology
- Record Identifier
- 9984197132402771
Metrics
20 Record Views