Papers by Timothy Masters
Class Assignment: ULC 257: Ecosystem Restoration: Restoring our Waterways
Across our Country there are many Ecosystem restoration problems that we face, especially when it... more Across our Country there are many Ecosystem restoration problems that we face, especially when it comes to our waterways. Projects to restore these waterways, such as the Great Lakes and Stream restoration are constantly ongoing. These efforts and steps that are being taken to restore these waterways are greatly helping to improve the state of our environment and resources, but also the natural habitats that are also being changed. But have we been doing enough to restore our waterway ecosystem?
Modern Data Mining Algorithms in C++ and CUDA C: Recent Developments in Feature Extraction and Selection Algorithms for Data Science
Estimating Future Performance I: Unbiased Trade Simulation
The title of this chapter is optimistic, perhaps shamefully so. Financial markets are notoriously... more The title of this chapter is optimistic, perhaps shamefully so. Financial markets are notoriously fickle. They are nonstationary (their statistical properties change over time), vulnerable to unforeseeable outside shocks, polluted by occasional wild swings for no apparent reason, and generally uncooperative. The idea that we can actually estimate the future performance of a trading system to any great degree is ludicrous. But what we often accomplish is to identify trading systems that have a painfully small expected future return, so we can be wary. Naturally, what we would really prefer is the ability to identify systems that have a high likelihood of large future return. And we may occasionally get lucky and enjoy this rare reward. It doesn’t hurt to try. But the reader must understand that the real goal of this chapter is to use rigorous statistical methods to weed out those superficially promising systems that in reality should be discarded, or at least revised before being put to work with real money.
Information and Entropy
An effective model takes one or more predictor, processes the information that they contain, and ... more An effective model takes one or more predictor, processes the information that they contain, and estimates a predicteds variable that is ideally useful in some way. But even the most sophisticated model is helpless if it is not given the information it needs to make a good decision. In this chapter, we explore the concept of information content of a variable, and we present a variety of algorithms for assessing the amount and nature of this information.
Autoencoding
Apress eBooks, 2018
Data Mining Algorithms in C++ by Timothy Masters Computers
DEEP Operating Manual
Apress eBooks, 2018
This chapter presents a concise operating manual for DEEP 2.0. The first section lists every menu... more This chapter presents a concise operating manual for DEEP 2.0. The first section lists every menu option along with a short description of its purpose and the page number on which more details can be found if the short description is not sufficient.
Combining Numeric Predictions
In many applications, several competing models are developed. The instinctive plan is to compare ... more In many applications, several competing models are developed. The instinctive plan is to compare the performance of these models and choose the best for the final use. But there is usually a better approach: keep and use many or all of the models. In all likelihood, some models will have weaknesses that can be alleviated by the strengths of others. By intelligently combining the predictions made by multiple models, a consensus prediction can be made that is nearly always superior to that made by the single best model. This chapter discusses a variety of methods for combining numeric predictions.
Medical Imaging 2004: Image Processing, May 12, 2004
This research consisted of evaluating diagnostic performance results using SVM outputs previously... more This research consisted of evaluating diagnostic performance results using SVM outputs previously obtained from an integrated Duke/DDSM USF data set and the GRNN oracle. The SVM kernels used in this research included Additive, Multiplicative, S2000, and Spline kernels. GRNN results are presented for the following combinations of gate variables: age, mass margin (MM), age and MM, and all 6 BIRADS™ indicators plus age. For all experiments, Differential Evolution (DE) was used to train the GRNN. A summary of the DE process is described, independent of the software application. The experiments described in this paper show that the GRNN oracle, with all of the gate variable combinations, performed better than any of the individual SVM kernels alone at or below 98% sensitivity.
Optimizing the Decision Threshold
Elsevier eBooks, 1993
When a neural network is used to make a decision, a threshold of half-activation is traditionally... more When a neural network is used to make a decision, a threshold of half-activation is traditionally used. If it is possible to estimate reliably the distribution of activation levels under both the null and alternative hypotheses, we can often choose the threshold more intelligently.
Supervised Feedforward Networks
Deep belief nets are generally trained in stages. First, one or more (usually more) layers are tr... more Deep belief nets are generally trained in stages. First, one or more (usually more) layers are trained with unsupervised algorithms. Rather than seeking to learn class memberships or predicted values, the model simply tries to find consistent patterns within the independent variables. Only after such patterns have been found does training switch to supervised mode. However, because supervised training algorithms are easier to understand than the usual unsupervised algorithms, you will begin your study of deep belief nets with supervised training.
Data Mining Algorithms in C++
Apress eBooks, 2018
Foundations
Elsevier eBooks, 1993
New results in breast cancer classification obtained from an evolutionary computation/adaptive boosting hybrid using mammogram and history data
Multilayer Feedforward Networks
Elsevier eBooks, 1993
This standard workhorse family of networks is described in detail. Efficient algorithms for imple... more This standard workhorse family of networks is described in detail. Efficient algorithms for implementing them are presented. Several training methods are given, and comparisons are made with regard to alternative network architectures within this family.
A new training algorithm for the general regression neural network
CUDA Code
Apress eBooks, 2018
Gating Methods
Apress eBooks, Dec 21, 2017
Optimization Issues
Apress eBooks, 2018
<title>Application of a GRNN oracle to the intelligent combination of several breast cancer benign/malignant predictive paradigms</title>
Medical Imaging 2000: Image Processing, Jun 6, 2000
ABSTRACT The General Regression Neural Network (GRNN) is well known to be an extremely effective ... more ABSTRACT The General Regression Neural Network (GRNN) is well known to be an extremely effective prediction model in a wide variety of problems. It has been recently established that in many prediction problems, the results obtained by intelligently combining the outputs of several different prediction models are generally superior to the results obtained by using any one of the models. An overseer model that combines predictions from other independently trained prediction models is often called an oracle. This paper describes how the GRNN is modified to serve as a powerful oracle for combining decisions from four different breast cancer benign/malignant prediction models using mammogram data. Specifically, the GRNN oracle combines decisions from an evolutionary programming derived neural network, a probabilistic neural network, a fully- interconnected three-layer, feed-forward, error backpropagation network, and a linear discriminant analysis model. In all experiments conducted, the oracle consistently provided superior benign/malignant classification discrimination as measured by the receiver operator characteristic curve Az index values.
Uploads
Papers by Timothy Masters