Dynamic-clamp-information-protocol

About

---

Uses an artificial neural network (ANN) to generate an excitatory and inhibitory conductance based on the absence or presence of a hidden state. This input can be utilized to estimate the mutual information.

The dynamic clamp input (conductance) generation method is described in:
Schutte, M. and Zeldenrust, F. (2021) Increased neural information transfer for a conductance input: a dynamic clamp approach to study information flow. Msc. University of Amsterdam. Available at: https://scripties.uba.uva.nl

The mutual information estimation protocol is described in:
Zeldenrust, F., de Knecht, S., Wadman, W. J., Denève, S., Gutkin, B., Knecht, S. De, Denève, S. (2017). Estimating the Information Extracted by a Single Spiking Neuron from a Continuous Input Time Series. Frontiers in Computational Neuroscience, 11(June), 49. doi:10.3389/FNCOM.2017.00049

Please cite these references when using this method.

Setup

---

The code depends on the following packages (version):

• numpy (1.20.3)
• pandas (1.2.4)
• scipy (1.6.3)
• matplotlib (3.4.2)
• seaborn (0.11.1)
• Brian2 (2.3)

A more detailed installation instruction is provided at the link.

Usage

---

To run a mutual information simulation the code has to go through 4 steps:

1. Input generation
   make_dynamic_experiments generates a binary stimulus and the respons (input theory) of the artificial neural network to this stimulus. The input theory can both be in the form of a flutuating current or a fluctuating conductance (dynamic clamp).
2. Input scaling
   scale_input_theory scales the input (current or conductance) with a given scaling factor. This will result in a Brian2.TimedArray with the correct unit to be injected into a (model) neuron.
3. Model initiation & Input injection
   Barrel_PC & Barrel_IN will initialize a model neuron using the Brian2 package. Barrel_PC.run(input) simulates the response of the model neuron to the injected input. Model fitting is described in Sterl & Zeldenrust (2020)
4. Mutual information estimation analyze_exp calculates the mutual information between the binary stimulus, injected input and output spike train as described in Zeldenrust et al (2017)

A run through steps 1-3 are provided in the big_sim.py file. Schutte & Zeldenrust (2021) gives a theoretical walk-through of the protocol.

License

---

This repository has licensed under the MIT licence. Read LICENCE.txt for the full terms and conditions.