Adaptive Transfer Function Networks

Conference · Fri Jan 01 00:00:00 EST 1993

OSTI ID:6421035

Goulding, J R ^[1]

---

1. Los Alamos National Lab., NM (United States) Portland State Univ., OR (United States). Dept. of Electrical Engineering

Real-time pattern classification and time-series forecasting applications continue to drive artificial neural network (ANN) technology. As ANNs increase in complexity, the throughput of digital computer simulations decreases. A novel ANN, the Adaptive Transfer Function Network (ATF-Net), directly addresses the issue of throughput. ATF-Nets are global mapping equations generated by the superposition of ensembles of neurodes having arbitrary continuous functions receiving encoded input data. ATF-Nets may be implemented on parallel digital computers. An example is presented which illustrates a four-fold increase in computational throughput.

• Research Organization: Los Alamos National Laboratory (LANL), Los Alamos, NM (United States)
• Sponsoring Organization: USDOE; USDOE, Washington, DC (United States)
• DOE Contract Number: W-7405-ENG-36
• OSTI ID: 6421035
• Report Number(s): LA-UR-93-1983; CONF-930741-3; ON: DE93014344
• Resource Relation: Conference: World congress on neural networks, Portland, OR (United States), 11-15 Jul 1993
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS//MATHEMATICS, COMPUTING, AND INFORMATION SCIENCE
NEURAL NETWORKS
TRANSFER FUNCTIONS
COMPUTERIZED SIMULATION
FORECASTING
TIME-SERIES ANALYSIS
TRAINING
EDUCATION
FUNCTIONS
MATHEMATICS
SIMULATION
990200* - Mathematics & Computers