Artificial Neural Networks: Algorithms & Hardware for Implementation By: Nathan Hower CSC 3990 – Computing Research Topics
What are Neural Networks? -Biological vs. Artificial -Computational models Components of a Neural Network - Inputs - Outputs - Transition Function (weights)
What are Artificial Networks Good For Anyway? - classification: pattern recognition - function approximation - clustering: data mining - association: restoring noisy data
Types of Artificial Neural Networks - Feedforward - Recurrent
Learning Paradigms and Algorithms -Supervised learning: sample results and actual results are compared - Backpropagation: error is corrected backwards -Unsupervised learning: unlabeled samples given, data organization is desired -Reinforcement learning: maximize 'reward' value
Parallel Computing for Processing Artificial Neural Networks - Multiprocessor computers: fast data communication - Heterogeneous clusters: indefinitely expandable - Beowulf clusters: best cost/performance ratio
Problematic Areas - Scalability Algorithms are designed for particular scope and/or limited hardware. - Computational power/cost Some problems are so complex that they require expensive specially designed hardware. - Lack of standardization The use of artificial neural networks is recent; alternate naming conventions and multiple equally viable approaches occur.
Current Areas of Application - Neurology & Neurobiology - Economics: stock market prediction - Image compression - NP-complete problems - EBAI – Studying Eclipsing Binaries with Artificial Intelligence
Future Work - Brain-Computer Interface (BCI): neural network learning algorithms moves burden of learning to the computer - Mind uploading/brain simulation