Feed aggregator

The End of Normal : Identity in a Biocultural Era / Lennard J. Davis

Catalog: New Book Titles - Sat, 09/20/2014 - 02:02
Davis, Lennard J., 1949- author
Categories: Books

Achieving Consistent Near-Optimal Pattern Recognition Accuracy Using Particle Swarm Optimization to Pre-Train Artificial Neural Networks

Similar to mammalian brains, Artificial Neural Networks (ANN) are universal approximators, capable of yielding near-optimal solutions to a wide assortment of problems. ANNs are used in many fields including medicine, internet security, engineering, retail, robotics, warfare, intelligence control, and finance. ANNs have a tendency to get trapped at sub-optimal solutions (local optimum) , and therefore trial and error is commonly used to select the network topology and train the network, which is prohibitively time consuming and costly.^ Recent advances in our understanding of the biological brain, hardware, algorithms, and potential for novel applications renewed interest in ANNs. Evolutionary Artificial Neural Networks (EANN) are among the more successful paradigms explored to improve ANNs’ performance. EANNs employ evolutionary computation techniques such as Genetic Algorithms (GA) or Particle Swarm Optimization (PSO) to train ANNs, or to generate ANNs’ topologies. Still, these improvements are not consistent and usually problem-specific. ANN performance depends in part on the number of neurons in hidden layer(s). The more neurons in hidden layer(s) the better the network’s ability to recognize specific samples it had seen during the training phase; however, the network becomes incapable of learning general patterns and recognizing these patterns in novel data. Performance on training data improves with training, while performance on testing data (samples the network had not seen previously) degrades (overfitting). This work rigorously investigated using PSO to pre-train ANNs with varying number of neurons in the hidden layer.^ It was shown that using PSO algorithm to pre-train ANNs improves classification accuracy for diverse problems, and, most notably, a PSO parameter configuration was developed that consistently yielded near-optimal solutions for pattern recognition problems from different domains. It was also shown how an automated algorithm for efficiently evolving optimal ANN size and topology can be designed to take advantage of the study’s findings. Furthermore, a novel biologically inspired hybrid of a visual ventral stream object recognition model was investigated – Hierarchical Model and X (HMAX), PSO, and feed-forward back-propagation ANN. This or similar biologically plausible models may be useful in vivo studies as well as in vitro simulations and applications.^

Categories: Books

Keystroke Biometrics Studies on Short Password and Numeric Passcode Input, and on Long Spreadsheet, Browser, and Text Application Input

A keystroke biometric system was enhanced to capture raw keystroke data directly from an individual’s computer system using an open source key logger originally designed for software testing. The key logger runs in the background capturing keystrokes directly through the operating system requiring no additional capture software, text entry window, or edit box for input. This allows the user freedom to generate unrestricted keystroke entry from any application installed on their system. ^ Long input data were collected from 20 participants using spreadsheet, browser, and text applications. Participants were free to type whatever they desired without using copy tasks in any of these experimental scenarios. Verification experiments were run on these samples using two classifiers. The newer Multi Match was far superior to the older Single Match classifier yielding EER performance of 8.1%, 15.7%, and 5.8% for spreadsheet, browser, and text entry in comparison to 13.6%, 27.5%, and 12.8%, respectively, for the older Single Match. ^ Short input data simulating a ten digit passcode were collected from 30 users entering the digits from the numeric keypad section of the keyboard. Using the feature set from the previous experiments, results were obtained from both classifiers - the EER performance using the Multi Match, varying the participants from 10 to 20 to 30, were 5.5%, 5.7%, and 6.1% compared to 15.6%, 15.7%, and 15.0%, respectively, from the Single Match classifier. Additional short-input experiments were run using data and features from Carnegie Melon University (CMU). The first was another keypad experiment using Pace University data with the CMU feature set and the second was a password experiment using both data and features from CMU. ^ The experiments conducted in this study had various independent variables, including participant count, classifier, feature set, and content type. Additionally, the data samples from the long input experiments were analyzed to get a better understanding of the performance variances and how they relate to keystroke lengths by calculating keystroke densities as the number of keystrokes divided by data capture elapse time.^

Categories: Books

The Taliban : ascent to power / Gohari M.J

Catalog: New Book Titles - Wed, 09/17/2014 - 01:01
Gohari, M. J
Categories: Books
Syndicate content