Programs and Family

Rowley Home


Current Programs


Alternate Energy

Based on the 7 patents devloped by Tom Rowley and Gerry Woodall, Rowley Energy Technologies provides consulting services and patent licensing



A startup machine learning, proof-of-concept, development program based on Neuro-Evolution and Genetic algorithmns

Thin Laptop

Safe, Renewable and On-Demand Hydrogen Energy


The patented Rowley-Woodall process converts water and aluminum into fuel-grade hydrogen using a simple chemical reactor. The process is completely non-toxic, safe and scaleable to any size production. The aluminum hydroxide byproduct of the reaction is ultra-pure and can be readily smelted back to aluminum fuel, essentially storing the output of, for example, hydroelelectric plants to be safely and conveniently transported. The reactor operates completely standalone and requires no supporting power infrastructure

A simple example of this technology in action would be a Combined Heat and Power system (illustrated above) for a home, apartment building or even a campus or military base. The reactor produces hydrogen, for both local power as well as vehicle fuel. Additionally, the reaction heat can be used for space heating (or cooling) and hot water. The reaction by-product is periodically collected and recycled when new fuel aluminum ingorts are delivered.

Neuro and Genetic Evolution


How NeuroEvolution Works

The Neural Network is trained to perform some Task by interacting with an Environment. The Environment produces "observations", a camera image, for example, and performs "actions" sent to it by the Neural Network (NN) , for example, turning a steering wheel or moving a robot arm. The NN starts with just simple direct connections between the input observation neurons and the output action neurons, so the intial actions the NN generates are random.

To train the NN to perfrom a Task, the NeuroEvolution Engine (NEE) makes hundreds of "chromesomes" which encode the structure of the NN as well as the weights between each Neuron. Typically there are hundreds of chromesomes, each with slightly different descriptions of the NN and different weights. In the training loop, each chomesome is used to configure the NN and then run against the Environment for many cycles. In each cycle, the Environment gives back an observation which is processed by the NN to produce actions. After the cycles complete, the Environment generates a "Fitness score" measuring how good that chromesome's NN did in performing the task. This process is repeated for all on the chromesomes.

When all of the Chromesomes have a Fitness score, they are ranked from best to worst and the worst are dropped from the population of chromesomes. The survivors act as parents for the new population which then is 'mutated', that is, small variations are made in their chromesomes, modifying the NN structure or some of the neuron weights. Now the new generation repeats the fitness trials again and again until some chromesome evolves to a good enough Fitness score in solving the Task.

In the last decade. Machine Learning (ML) has made great advances, in large part due to the use of gradient descent algorithms and, for visual applications, the use of Convolutional Neural Networks. However, the high computational costs and the restriction that this technique only works for differentiable objective functions have shown the limitations of this "mainstream" approach.

During roughly the same time period (from early in this century), a class of machine learning technologies inspired by evolutionary processes, such as fitness selection and mutation, has arisen which are rapidly overcoming the limits of "traditional" gradient descent algorithms. This approach is generally referred to "Genetic or Evolutionary" solutions.

A specific subset of the paradigm is NeuroEvolution in which both the parameters of the Neural network as well as its internal structure are simulaneously evolved. This technology,coupled with another recent innovation called "Attention", is the core methodology behind our current proof-of-concept program.

A Nationally Distributed Family

Ted's Family

Ted, Genna, Finn and Parker

Richmond, VA

Oldest son and family are in Virginia

Tom and Cathie


Based in San Jose,CA, Tom and Cathie are the parents, grand and great-grand parents of this clan

Gramps and Grandma Cathie
Doug's Family

Doug, Jen and Drew

Limrick, PA

Younger son, wife - both schoolteachers - and grandson live near famous Valley Forge

Michelle,Ron,Sierra, Cody

Michelle, Ron, Sierra & Cody

San Antonio, TX

Daughter and husband live just outside San Antonio

COdu's famile

Cody, Courtney & Bryce

Nacogdoches, TX

Grandson Cody Palm, wife and stepson are in the famous Texas town of Nachedoches

Mom  and Sis

Mom and Teri

Saratoga, NY

Gloria (Mom) and sister Teri live in upstate New York

Say hello to us