By seeking, you will discover... 2005-11-30T22:17:30Z Paul Kulchenko http://notebook.kulchenko.com/modeling/neural-modeling Blooki/0.4808 <p>After attending our local <a href="http://groups.yahoo.com/group/seattleai/">SeattleAI</a> meeting and discussing <a href="http://groups.yahoo.com/group/seattleai/message/74">neural network frameworks</a> I decided to put together my notes on this subject. </p> <h2>What would one need to know to do neural modeling?</h2> <p>This paper (<a href="http://www.cs.colostate.edu/~anderson/res/neural/techreport2004.pdf">Introduction to Computational Neural Modeling for Computer Scientists and Mathematicians</a>) provides a list of principles that govern well-practiced neural modeling:</p> <ul> <li>Knowledge of neuroscience principles</li> <li>Knowledge of computational time and space complexities</li> <li>Awareness of computational resources</li> <li>Well-researched neurological parameters</li> <li>Well-planned and implemented model design</li> <li>Documentation discipline (both parameter sources and coding)</li> <li>Knowledge of statistical analysis techniques</li> <li>Patience</li> </ul> <h3>Neuron</h3> <p>See <a href="modeling/neural-modeling-neuron">Neural Modeling: Neuron</a></p> <h3>Synapse</h3> <ul> <li>See <a href="modeling/neural-modeling-synaptic-plasticity">Neural Modeling: Synaptic Plasticity</a></li> <li>See <a href="modeling/neural-modeling-synaptic-connectivity">Neural Modeling: Synaptic Connectivity</a></li> <li>Post synaptic potential (PSP) models</li> <li>Neurotransmitters (only four most significant types out of hundreds discovered): <ul> <li><span class="caps">AMPA </span>-- fast excitatory</li> <li><span class="caps">NMDA </span>-- slow excitatory</li> <li><span class="caps">GABAA </span>-- fast inhibitory</li> <li><span class="caps">GABAB </span>-- slow inhibitory</li> </ul> </li> <li>Transmission delays</li> <li>Plasticity: presynaptic/postsynaptic</li> <li>Operations: <span class="caps">AND, OR, XOR, NOT, SUM, MUL, MAX</span></li> </ul> <h3>Learning</h3> <ul> <li>Learning algorithms: <ul> <li>Synaptic weight changes: <ul> <li>Hebb's rule: if S sends a pulse at time <em>t</em> and R fires at time <em>t+1</em>, then that synapse becomes <em>more</em> effective at firing R in the future</li> <li>Milner's modification: if S sends a pulse at time <em>t</em> and R does not fire at time <em>t+1</em>, then that synapse becomes <em>less</em> effective in firing R in the future</li> </ul> </li> <li>Connection changes (more/less connections with a particular neuron)</li> <li>Neuron changes (birth/death)</li> </ul> </li> <li>Learning related problems: <ul> <li>Under-/over-learning; under-/over-fitting</li> <li>Catastrophic interference</li> </ul> </li> <li>Sensitization/habituation</li> <li>Types of learning: autonomous, continuous, on-line, supervised/unsupervised</li> </ul> <h3>Initial configuration</h3> <h3>Architecture</h3> <ul> <li>Input/output <ul> <li>Static/dynamic</li> <li>Various types of sensory input and motor output</li> </ul> </li> <li>Neural circuits</li> <li>Modular (<a href="http://scholar.lib.vt.edu/theses/available/etd-06092000-12150028/unrestricted/etd.pdf">Biologically Inspired Modular Neural Networks</a>)</li> </ul> <h3>Simulation</h3> <ul> <li>Time-driven vs. event-driven</li> <li>See <a href="modeling/modeling-getting-visual">Visualization</a></li> </ul> <h3>Overall models</h3> <ul> <li>Rodney Cotterill: <a href="http://dx.doi.org/10.1016/S0301-0082(00)00058-7">Co-operation of the basal ganglia, cerebellum sensory cerebrum and hippocampus: possible implications for cognition, consciousness, intelligence and creativity</a> (or <a href="http://www.cogsci.bme.hu/csaba/downloads/basalganglia.pdf">here</a>) and <a href="references/articles/mind-a-moving-story"><span class="caps">MIND</span>-A-MOVING-STORY</a></li> <li>Peter F. C. Gilbert: <a href="http://dx.doi.org/10.1016/S0926-6410(01)00035-0">An outline of brain function</a></li> <li>Yuri I. Arshavsky: <a href="http://dx.doi.org/10.1016/S0165-0173(02)00249-7">Cellular and network properties in the functioning of the nervous system: from central pattern generators to cognition</a></li> </ul> <h3>Modeling tools and simulators</h3> <ul> <li>See <a href="modeling/neural-modeling-tools-and-simulators">Tools and Simulators</a></li> </ul> http://notebook.kulchenko.com/modeling/neural-modeling 2005-11-30T22:17:30Z 2005-11-30T22:17:30Z