Neo Cortical Repository and Reports Database and Reports
Neo. Cortical Repository and Reports: Database and Reports for NCS Edson O. Almachar, Alexander M. Falconi, Katie A. Gilgen, Devyani Tanna, Nathan M. Jordan, Roger V. Hoang, Sergiu M. Dascalu, Laurence C. Jayet Bray, Frederick C Harris, Jr. Brain Computation Lab Department of Computer Science and Engineering University of Nevada, Reno
Outline Introduction n Background n Design Overview n Conclusion and Future Work n
Human Brain Neurons : ~ 8. 6 x 10^10 (86 Billion) n Synapses: ~ 1 x 10^14 (100 Trillion) n
Brain Background Neuron ( C ) - cell that uses electrical signals to send information, as well as process it n Axon ( A) - the nerve fiber that a neuron’s electric pulse flows through n
Brain Background Synapse - the transmission of information from one neuron to another n Network - a computational model of a cluster of neurons sending information n
Neural Simulators Allow users to create systems of neurons with parameterized cell data and connection information n Simulate brain activity using biological and mathematical models n Build a foundation for more research on the processes of the brain n
Levels of Organization of Modeling
What is NCS? n n Developed and maintained by the UNR Brain Computation Laboratory The Neo. Cortical Simulator is designed for modeling large-scale neural networks and systems n n Can model millions of neurons in real time Open source Runs on a heterogeneous cluster of CPUs and NVIDIA GPUs First simulator to support real-time neurorobotics application
Building Better Solutions Users are usually researchers in the neuroscience field. n User Inconveniences for Neural Simulators n Learning to code brain models n Time spent organizing output data n Generally Low Usability n
Building Better Solutions
Building Better Solutions n The Primary Users n Neuroscientists n Design Goals n n n The Primary Usage n Research n n n Simplicity Usability Learnability Easy Collaboration Fast
Brain Model Database Design n Three Neuron Model Types n n Necessary Capabilities n n Izhikevich, Leaky-Integrate-And-Fire, Hodgkin Huxley Storage, Searching, Updating Storage Structure n JSON format, Using Mongo. Kit
Brain Model Database Design
Reports Design n Graph Types n Raster Plot, Line Graph Understandable Real Time Reporting n Customization n Color, Size, Type, Neuron Selection Ability to Easily Save Reports
Framework FLASK : python microframework n Mongo. DB : nonrelational database n D 3. Js : Graphing Library n j. Query. UI. JS : javascript UI library n
NCR Database Goals Increased Collaboration n Simple Layout n Easy Searching n
Database Tab Components n Database Model Preview Headers Sorting Feature for Quick Searching n Listed in Ascending or Descending Order n Simple Preview Information n
Database Tab Components n Left Search Panel Collapsable Grouping Structure n Can Select Entire Types n Specify Parameter Values n n As Value or Range of Values
Database Tab
Database Tab Components n Detailed View n Opens when model preview is selected
Report Tab Goals Management Control Panel n Dynamic Creation & Deletion n Ability to Save Reports n
Reports Tab Components n Raster Plots
Reports Tab Components n Line Graphs
Reports Tab Components n Customizations Color Picker n Drag and Drop n Scale Axis n
Reports Tab Components n Customizations Cell Selection n Pause and Playback n
Reports Tab Components n Saving Reports Image: GIF or SVG n Animation: Animated GIF n
Conclusion n Web Application aims to make using NCS easy, Leading to more time spent on research
Future Work Complete full front end application by merging NCB with NCR and Virtual Robot n NCB n n Brain Builder Simulation Builder NCR n Reports n Model Database n Virtual Robot
Neo. Cortical Repository and Reports: Database and Reports for NCS Edson O. Almachar, Alexander M. Falconi, Katie A. Gilgen, Devyani Tanna, Nathan M. Jordan, Roger V. Hoang, Sergiu M. Dascalu, Laurence C. Jayet Bray, Frederick C Harris, Jr. Brain Computation Lab Department of Computer Science and Engineering University of Nevada, Reno
30
Hodgkin-Huxley Neurons (Added in NCS 7. 0) Biologically accurate n Developed in 1952 by Alan Hodgkin and Andrew Huxley from their experiments on the giant axon of a squid n Set of four differential equations n Three variables n, m, h n
Hodgkin-Huxley (cont)
Leaky Integrate-and-Fire n Comprised of n n n Sub-threshold leaky-integrator dynamic Firing threshold Reset mechanism Leakage Channels Drive the neuron to higher voltage Let the voltage decay to its resting potential
Izhikevich Created by Eugene M. Izhikevich n Published in 2003 n Most Simplistic n Computationally efficient and captures large variety of response properties of real neurons n Only 6 variables! n
Izhikevich (Added in NCS 6. 0) Image Source:
Izhikevich Output
- Slides: 36