Opening up the black box of artificial neural networks

Journal Article · Sun May 01 04:00:00 EDT 1994 · Journal of Chemical Education

DOI:https://doi.org/10.1021/ed071p406· OSTI ID:131612

Spining, M T ^[1]; Darsey, J A ^[2]; Sumpter, B G; Noid, D W ^[3]

---

1. Univ. of Tennessee, Knoxville, TN (United States)
2. Univ. of Arkansas, Little Rock, AR (United States)
3. Oak Ridge National Lab., TN (United States). Chemistry Div.

In this paper, neural networks are divided according to training methods--supervised and unsupervised. Supervised training is used when a training set consisting of inputs and outputs is available. The network uses the training set to determine an error and then adjusts itself with respect to that error. Unsupervised networks are used when training sets with known outputs are not available, for example, for realtime learning. These networks use the inputs to adjust themselves so that similar input gives similar output. Another classification that will be used is feedforward and feedback networks. In a feedforward network, information is propagated through the network in one direction until it emerges as the network`s output. However, in a feedback (recurrent) network, the input information is propagated through the network but can also cycle back into the network (the signal is recurrent). In the present paper, the authors give a fundamental overview of feedforward neural networks, present some applications using them in chemical physics, and comment on the potential for future uses in chemistry. They begin by discussing some specific types of neural networks that provide the generality needed to pursue applications in the chemical sciences.

• Sponsoring Organization: USDOE
• DOE Contract Number: AC05-84OR21400
• OSTI ID: 131612
• Journal Information: Journal of Chemical Education, Journal Name: Journal of Chemical Education Vol. 71; ISSN 0021-9584; ISSN JCEDA8
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
ARTIFICIAL INTELLIGENCE
COMPUTER ARCHITECTURE
NEURAL NETWORKS
TRAINING
TRANSFER FUNCTIONS
USES