Fruit Fly Decision-Making Simulation

Biomimetic computing simulation that models decision-making in fruit flies using Spiking Neural Networks (SNNs). By mimicking natural biological processes, the simulation provided insights into how these networks can be applied to study decision-making in living creatures.

Introduction to the Project

This project involved developing a biomimetic computing simulation to understand possible decision-making processes in fruit flies using Spiking Neural Networks (SNNs).

Biomimetic computing refers to the creation of systems that mimic natural biological processes.

Simulated fruit fly standing on a document showing made up code describing the inner workings of a fruit fly's decision making process

Why Fruit Flies?

The fruit fly was selected as the subject of this simulation due to its comparatively simple biological system relative to more complex organisms. This simplicity make it easier for me to focus on a small set of actions related to the decision-making process of a fruit fly. These actions include:

  1. Forage
  2. Mate
  3. Evade Predators

The goal was to create a type of decision-making pipeline that when fed sensory input, would output actions. Using SSNs (Spiking Neural Networks), I was able to successfully achieve this goal. The network in this simulation is simple and it would be worth doing further work to make it more closely represent the complexities found in a fruit fly. For example, this simulation could be expanded to simulate the GABAergic neurons as found in the mushroom body of a fruit flies brain. These neurons are required for the formation of long-term memory which could help improve the simulated decision-making process.

The potential for implementing similar studies on more complex subjects, is a current area of research.

Simulation Design and Methodology

The simulation was designed with the help of Python and the Nengo Brain Maker library, using SNNs to model the flow of information from sensory input to output decision. This allowed for the simulation of a fruit fly to sense food, a mate, or a predator (sensory input), and then conducting a rudimentary risk/reward analysis to decide whether to forage for food, mate, or evade the predator (action output).

Project Limitations and Scope

It is important to mention that this project has limitations. Developing a highly accurate model of a biological neural network presents considerable challenges due to the complexity of such systems. Furthermore, the simulation largely focused on binary outcomes, which may oversimplify the multi-faceted decision-making processes in real-world scenarios.

Future Directions

While this project has demonstrated an application of SNNs to model decision-making processes in a relatively simple organism, it also points to the larger challenges inherent in applying such techniques to more complex organisms. It should be seen as a starting point that can lead to future research in the field of biomimetic computing.

Google Colab Notebook Link