Online learning model of olfactory bulb external plexiform layer network (Imam & Cleland 2020)

 Download zip file 
Help downloading and running models
Accession:261864
This model illustrates the rapid online learning of odor representations, and their recognition despite high levels of interference (other competing odorants), in a model of the olfactory bulb external plexiform layer (EPL) network. The computational principles embedded in this model are based on the those developed in the biophysical models of Li and Cleland (2013, 2017). This is a standard Python version of a model written for Intel's Loihi neuromorphic hardware platform (The Loihi code is available at https://github.com/intel-nrc-ecosystem/models/tree/master/official/epl).
Reference:
1 . Imam N, Cleland TA (2020) Rapid learning and robust recall in a neuromorphic olfactory circuit Nature Machine Intelligence, in press
Model Information (Click on a link to find other models with that property)
Model Type: Neuron or other electrically excitable cell;
Brain Region(s)/Organism: Olfactory bulb;
Cell Type(s): Olfactory bulb main mitral GLU cell; Olfactory bulb main interneuron granule MC GABA cell;
Channel(s):
Gap Junctions:
Receptor(s):
Gene(s):
Transmitter(s):
Simulation Environment: Python;
Model Concept(s): Neurogenesis; Synaptic Plasticity; Temporal Pattern Generation; Activity Patterns; Learning; Gamma oscillations; STDP; Coincidence Detection; Delay; Hebbian plasticity; Memory; Olfaction; Oscillations; Pattern Recognition; Synchronization;
Implementer(s): Imam, Nabil ; Cleland, Thomas [tac29 at cornell.edu];
Search NeuronDB for information about:  Olfactory bulb main mitral GLU cell; Olfactory bulb main interneuron granule MC GABA cell;
README for OB_EPL model
Imam and Cleland 2020, Nature Machine Intelligence
Also:  https://arxiv.org/abs/1906.07067


1. System requirements
	a. Software dependencies 
		-- Python version 2.7 
		-- Python packages numpy and matplotlib
	b. Versions that the software has been tested on 
		-- Python version 2.7 running on Mac OS X 10.11 and Windows 10
	c. Non-standard hardware 
		None. The code provided here runs on a convetional desktop computer. For the full model, 
		this code runs three orders of magnitude slower than the model implemented in Intel's Loihi neuromorphic 
		system. 

2. Installation guide
	a. We recommend using the Anaconda Python Distribution. 
		-- Installation link: https://docs.anaconda.com/anaconda/install/
		-- Chooose your operating system and Python version 2.7
		-- Download and install
	b. Typical install time: Approximately 20 minutes

3. Demo
	a. Instructions to run on data
		-- Open the Anaconda Navigator
		-- Click "Launch" under "Spyder" (a Python development environment)    
		-- Click File->Open 
		-- Open the file "singleOdorTest.py"
		-- Click Run->Run
	b. Expected output
		-- Raster plots showing spiking activity for a test sample of Toluene. Autoassociative 
		   network dynamics across five gamma cycles can be observed across the five raster plots. 
		-- Bar plot showing the similarity between a test sample of Toluene and the learned representation 
		   of Toluene across five gamma cycles. 
	c. Expected run time on a conventional desktop computer: ~15 seconds. 

4. Instructions for use
	a. Files "singleOdorTest.py", "multiOdorTest.py", "plumeTest.py" can be run
	   through Spyder (see step 3a above). 
        b. Reproduction instructions 
		-- "singleOdorTest.py" generates main results of Figure 3. 
		-- "multiOdorTest.py" generates main results of Figure 4. 
		-- "plumeTest.py" generates main results of Figure 5. 
.

Loading data, please wait...