Who Wrote the Book of Life?

Phase 2 of the project will expand the basic database by using a more powerful neural network. Funds from the NASA Advanced Concepts Office provided scientists with a Beowulf-class (A Beowulf-class computer is a cluster of personal computers running LINUX, connected by its own private Local Area Network) parallel computer, to address scientific problems associated with large data sets.

Scientists have named the new parallel computer "Leibniz," after the German mathematician whose lifelong goal was to organise all human knowledge. This computer system will expand the image database by acquiring and classifying new and ambiguous images. To discriminate organic life forms from inorganic shapes, microbiologists often use the vague criteria, "Does it look alive to you?" A parallel computer using pattern recognition can make this task easier and more exact by breaking the starting image down into identifiable parts.

"Human judgement is still very much depended upon for identifying microbial life forms," says Dr. David Noever of NASA's Marshall Space Flight Centre. "Automated filters would be much like the filters commonly used to sort out useful e-mail's from useless ones. The user of the neural network would get a morning menu of microbial candidates for further detective work."

Although the trained human eye is better at recognising microbial life forms, using a computer "filter" to check for lifelike patterns could help cut the immense scale of the Book of Life project down to a more manageable size.

By Phase 3 of the project, the neural network will be so advanced in its learning that it will be able to acquire and classify new images with minimal human supervision. This network would then be equipped for future search scenarios, including the examination of meteorites found on Earth and samples retrieved from lunar or interplanetary space missions. This advanced neural network will be a fast and efficient classifier of the vast amount of microbial images that will need to be catalogued.

A Big Problem
This speed and efficiency are extremely important due to the detail with which the samples must be analysed. Not only are there a lot of samples to study, but there are multiple dimensions to consider. D'Arcy Thompson used mostly linear and quadratic maps to compare different life forms. Linear maps between two shapes require four coefficient variables, while quadratic maps use 10 variables.

Thompson wrote in "On Growth and Form," "I know that in the study of material things number, order, and position are the threefold clue to exact knowledge: and that these three, in the mathematician's hands, furnish the first outlines for a sketch of the Universe."