Wolfram Computation Meets Knowledge

Wolfram Summer School


Laura Sainz Villalba

Summer School

Class of 2014


Laura began to pursue a degree in physics in 2010 at the Complutense University of Madrid, Spain, while she was studying medicine at the medical school in the same university. She wished to improve her mathematical background to model natural phenomena of biological systems, specifically focused on the physiology of the nervous system. After graduating in medicine in 2012, she became a resident of neurosurgery at the Fundación Jiménez Díaz Hospital of Madrid. She is now about to graduate in physics. Her research interests lie in computational neuroscience, dynamics of complex systems and brain-machine interfaces. In her spare time, Laura enjoys playing the violin, making LEGO-Arduino constructions, and riding her unicycle. She also likes to construct hand-made wood ship models.

Project: Modeling Brain Diseases through Wrongly Coded Programs

The mistakes are the clues

The human brain is able to perform complicated tasks using high-level functions that require handling very subtle abstract features. For the past several decades, we have developed some sense of how the brain works through stimuli-controlled experiments in healthy people. Other things that give us a hint of the brain’s inside machinery are the kinds of problems people with brain damage might have performing certain tasks.

Whenever there is an injury on the brain tissue, it affects certain functions. This will actually be noticed as an increase of the percent of incorrect outputs and, thereafter, a decrease of performance. When trying to achieve a goal, the mistakes that might appear are not just random. We will try to do reverse engineering analyzing the abstract features of the subset of errors compared to the space of all possible mistakes. This will give a sense of the underlying function structure that is involved.

The goal of this project is to try to simulate brain damage, taking an analogy between the brain and a computer program. Given a classifier program, random wrong code will be introduced at certain relevant levels. Analyzing the outputs, we will try to debug the code to understand the underlying structure of the program. The relation between outputs and inputs will be given in a confusion matrix. Each element will be proportional to the frequency of the mistakes for each input and output association. Understanding what types of mistakes are more common with respect to other possible ones may be a good tool to predict which level has been “damaged” with wrong code. We can figure out what features are used to actually perform the task and the internal hierarchy of the information within the program.

This method could be applied to models that reproduce human brains. Comparing the types of mistakes with real patients with the ones produced in the computer lab might be helpful to understand the physiology and the areas involved in certain diseases like stroke or tumors.

Favorite Outer Totalistic r=1, k=2 2D Cellular Automaton

Rule 136516