Chaos seems to aid learning

By Kimberly Patch, Technology Research News

Although it's clear that the cerebellum is the part of the human brain involved in coordinating movements in ways that allow people to learn skills like riding a bike, there are mysteries about how the learning process works.

Researchers from Core Research for Evolutional Science and Technology (CREST) in Japan have built a computer simulation of the inferior olive, a portion of the brain that probably relays errors in movement to the cerebellum. It has been difficult to explain the mechanics of this relationship because inferior olive cells that connect to the cerebellum fire slowly, and this does not fit well with the common hypothesis that high-fidelity error signals are needed for efficient learning.

The researchers got the idea for the simulation after initial research showed that if neurons were electrically coupled, or linked, a certain type of chaotic signal could emerge.

The researchers' simulation shows that moderate electrical coupling between nerve cells in the inferior olive could produce a type of chaotic firing that effectively recodes the high-frequency information into slower signals by imparting information within the rhythm rather than just the frequency of nerve firing. "The chaotic firing was more robust than we expected," said Nicholas Schweighofer, a researcher at Core Research for Evolutional Science and Technology. The model shows that "chaos can be useful in the brain," he said.

In addition to allowing researchers to better understand the mechanics of the brain, the researchers' theory of chaotic resonance could speed electronic communications and improve robotics. "In communications, our work [could] maximize the information transmitted in networks," he said. "In robotics, chaos could be used to explore the environment to optimize learning," he said.

Electrical signals carry information from one end of a nerve cell to the other, while a chemical reaction is responsible for passing signals from one cell to another through their interconnected dendrites, or nerve cell fibers.

The researchers' results explain some unusual properties of the inferior olive cell input to the Purkinje cells of the cerebral cortex, according to Schweighofer. Each Purkinje cell contains two types of nerve synapse inputs, or connections to other nerve cells. The cells connect to about 100,000 other nerve cells via parallel connections, but have only a single connection to an inferior olive neuron.

The parallel connections generate simple nerve spikes, or on signals, but the inferior olive connection generates a more complicated signal. Experiments have also uncovered apparently random firing, and chaotic subthreshold activity, or signals that are not strong enough to trip the chemical reaction that ordinarily passes a signal to neighboring cells. It is also known that the inferior olive neurons are electrically coupled.

It was a challenge to make a realistic model of the inferior olive, said Schweighofer. "Finally showing the existence of chaos... necessitated very lengthy computations," he said.

The researchers' inferior olive cell models included the known location of the ionic currents that carry signals between nerve cells, the gap junctions between the cells and the synaptic inputs.

The researchers modeled two types of networks of a few simulated inferior olive cells: chain networks, and grid networks. In chain networks, each neuron is electrically coupled to its one or two neighboring cells depending on its position in the chain. In grid networks of 2 by 2, 3 by 3, and 9 by 3 cells, cells are connected to two, three or four neighbors depending on their grid positions.

When the researchers removed to the connections between cells, each cell generated plain periodic spikes, or signals at an average rate of 3.1 spikes per second. When the researchers connected cells within a network using just an intermediate coupling strength, the firing pattern of individual cells appeared chaotic and the average firing rate was reduced to 1.8 spikes per second. When the researchers used a strong coupling strength, the cells generated regular, synchronized spikes at a firing rate of 3.5 spikes per second. These results are consistent with experiments on actual nerve cells.

The simulation showed that moderate electrical coupling speeds information transfer, according to Schweighofer. The mutual information per spike for a single cell at the center of the 3 by 3 networks was 48 percent greater than the same network without coupling and 37 percent greater for the coupled network as a whole despite the lower spike-per-second rate. The researchers found similar results for the other types of networks.

Now that they have proved computationally that chaotic signals are capable of carrying extra information, the researchers are aiming to show empirically that this is what happens. The next step is "doing in vivo work showing that chaos actually exists in the inferior olive," said Schweighofer.

Schweighofer's research colleagues were Kenji Doya, Hidekazu Fukai, Jean Vianney Chiron, Tetsuya Furukawa and Mitsuo Kawato. The work appeared in the March 30, 2004 issue of Proceedings of the National Academy of Sciences. The research was funded by the Telecommunications Advancement Organization and the Human Frontier Science Program.

Timeline:   5 years
Funding:   Private
TRN Categories:  Chaotic Systems, Fuzzy Logic and Probabilistic Reasoning; Logic; Artificial Intelligence; Robotics; Computer and Machine Learning
Story Type:   News
Related Elements:  Technical paper, "Chaos May Enhance Information Transmission in the Inferior Olive," Proceedings of the National Academy Of Sciences, March 30, 2004




Advertisements:



May 5/12, 2004

Page One

DNA bot targets cancer

Memory stores three bits in one

Chaos seems to aid learning

Y switches set up low-power logic

Briefs:
Color display blocks prying eyes
Net lets hand-helds view 3D data
Speed limits could slow viruses
Nano test tubes fabricated
Nanowires make tiny compasses
Upbeat computers boost users

News:

Research News Roundup
Research Watch blog

Features:
View from the High Ground Q&A
How It Works

RSS Feeds:
News  | Blog  | Books 



Ad links:
Buy an ad link

Advertisements:







Ad links: Clear History

Buy an ad link

 
Home     Archive     Resources    Feeds     Offline Publications     Glossary
TRN Finder     Research Dir.    Events Dir.      Researchers     Bookshelf
   Contribute      Under Development     T-shirts etc.     Classifieds
Forum    Comments    Feedback     About TRN


© Copyright Technology Research News, LLC 2000-2006. All rights reserved.