5 Topics for Final Projects

5.1 Agent Based Modeling

Provide an overview of the idea behind agent based models in psychology. Give a demonstration by implementing in racket a simple version (including graphics) of the classic work of Schelling  (Schelling 1971).

5.2 Genetic Algorithms

Genetic algorithms are algorithms where the "learning" takes place via selection and recombination in analogy to evolution. It has been used to select the parameters of otherwise conventional neural networks, but can also be used on its own. If you select this project you will need to explain what a genetic algorithm is, make the case that it is relevant to some aspect of neural or psychological modeling, and provide some racket code implementing some examples. A tutorial paper (pdf) with MATLAB® code is available. A more academic treatment (and much deeper treatment) is available as a review  (Stanley et al. 2019). A recent podcast features a discussion with one of the authors, Joel Lehman.

5.3 Quantum Probability

Developed for quantum mechanics the formalism of quantum probability has been suggested to be a better approach to human decision making than conventional, classical probability (Bruza et al. 2015);  (Pothos, Emmanuel M. and Busemeyer, Jerome R. 2022). It seems that one of the authors has some MATLAB® programs available for some of the published models. For this project you will need to provide one example of a different prediction between classical and quantum probability, review the empirical results in favor of the latter, and show a racket program that computes some of the key values. Some additional reviews are available  (Khrennikov 2023), (Lewis 2021).

5.4 Vector Symbolic Architectures

Provide an overview of vector symbolic architectures and provide a short racket implementation of Kanerva’s "what is Mexico’s dollar?" example  (Kanerva 2010).

5.5 Linear Ballistic Accumulators

Many models of human decision making envision the process as one of evidence accumulation that drifts us towards a threshold. Hit one border and you decide "no"; the other direction and you decide "yes". The amount of time it takes is a proxy for reaction time. The proportion of times you hit one border is a proxy for accuracy. The full drift diffusion models have some complex mathematics so the Linear Ballistic Accumulator Model (Brown and Heathcote 2008) was proposed as a much simpler alternative. For this project you will explain the LBA model and demonstrate an implementation in Racket. It would be nice to be able to show how to fit data to this model, but that will be more challenging. There are some statistics functions in Racket for simulation, but this will take more work to get working right.

5.6 Fitzhugh-Nagamo Neuron Model

This is a model of neuronal firing that is very commonly used as a simpler, but informative, example of neuronal dynamics. For this project you will implement this model and demonstrate how to visualize the effects of parameter manipulations, e.g. by using vector field plots.

Project References