The paper consists of 2,000 words plus an abstract, references, figures, mathematical equations, code, and optional tables/appendices, as many as needed (within reason).

The format of the paper is flexible but:

• It should contain an abstract and selected references (being 10 the ideal number)
• Besides the abstract and references (task 5 below), it should contain the following or similar four sections and in this order: Introduction, Methods, Results & Discussion (with 3 subsections corresponding to tasks 2, 3, 4 below), and Concluding Remarks.
• The tasks to address in the paper are the following:
  1. Introduction and Methodology sections: Explain how spikes and their firing rates are generated, both in real neurons and in models, specifying details. [20 marks]
  2. First subsection of results: Generate artificial spike trains using any model(s) of your choice. Artificial spikes should resemble as much as possible the real ones emitted by at least one neuron in your specific dataset (ideally more) and for at least one trial of your choice. Briefly discuss the results. Evaluate and succinctly discuss the similarity between real spikes and ‘in silico’, computationally generated spikes. [20 marks]
  3. Second subsection of results: Estimate the firing rates using any method of your choice from the real neurons in your dataset and for at least one trial. Ideally, in addition, using spikes generated from at least a neuron model. Evaluate and briefly discuss the similarity between real and artificial firing rates. [20 marks]
  4. Third subsection of results: Decode the types of behaviour from the real firing rates using any decoder or decoders of your choice. Evaluate the decoding performance for at least one trial, ideally for more. If not possible, you are free to design the classification problem by using fewer than three or four classes. However, addressing a multi-class classification in addition would be preferable. Succinctly discuss the results. [20 marks]
  5. Write an abstract (typically 150-300 words) and provide 10 selected references. [20 marks]

For tasks 2-4, you are free to use the MATLAB proprietary implementations provided in Brightspace.

However, using your own code awards up to 4 marks per task (2, 3 and 4), for a total of 12 marks. If you use your own code, please put it as an appendix (the appendix does not count toward the word limit). You can develop the code in any language, but Python or Julia are suggested. Please note the marking criteria outlined below.

For tasks 2 and 3 it is not allowed to code harvested in the wild from the internet, an LLM or a package. Thus, the code should be a commented, fully explainable, transparent and original re-implementation of the spiking models and rate methods discussed in the lectures.

For task 4, you can use any package of your choice. Please note the marking criteria outlined below.

ORIGINALITY REQUIREMENT

The following originality requirements will apply to this assignment:

You are allowed to use any Generative AI or other AI powered tools, such as ChatGPT, for specific aspects as directed by the Unit Leader. Where any part of your assessment is sourced, or partially sourced from a generative AI tool, this requires a reference in the BU Harvard style.