Nanoscience - Using Neural Networks
A new approach to crunching massive volumes of data uses neural networks, an architecture of multiple elements that is figuratively taught to pool imbedded information into results, like an artificial brain. Researchers at ORNL's Center for Nanophase Materials Sciences developed and trained a neural net to recognize significant patterns in data from analytical experiments using methods such as atomic force microscopy. The work was recently published in Physical Review Letters. The neural-network approach provides a bridge between theoretical and experimental research by quickly extracting physical parameters from voluminous and complex data. The work is funded by ORNL's Laboratory Directed Research and Development program.
O. S. Ovchinnikov, S. Jesse, P. Bintacchit, S. Trolier-McKinstry, and S. V. Kalinin, PRL 103, 157203 (2009)