One of the main goals of this course is to provide you with the opportunity to tackle complex problems that require you to model the underlying science and engineering at hand, understand the background, and frame it as a machine learning problem.

While you can choose any topic for this course, try to select a project where you can combine existing analytical methods with machine learning techniques; or where you can study the problem at hand in a scientific manner. That is, you should aim to apply the scientific method to your project, which includes formulating a hypothesis, collecting and analyzing data, and training multiple models to compare the performance of the different methods, analyzing the results, going back to the data, and refining the model, etc. The number of times you go through this iterative process will show how much effort and understanding you’ve gained of the problem.

For this reason, working with spatio-temporal data where some theory for the problem you’re tackling is available is encouraged. In some cases, there may be temporal models that are themselves empirically discovered from data and in others there may be some theory that you can use to frame your problem and compare with your results.

The following will provide you with a detailed overview of the requirements for the project proposal.

1. Motivation and Question

  • Description: Clearly define the problem you intend to solve or the question you want to answer. Explain why this problem/question is significant and relevant. Typically, this section should include a brief introduction to the problem, its context, and its importance. While you need to choose something you can achieve in a month and a half, don’t be afraid to be creative and think outside the box!
  • Goal: Ideally, you should state the specific goal of your project, such as building a model to predict a certain outcome, identifying patterns in a dataset, or solving a specific problem. The goal should be well-defined and measurable, and it should be directly related to the problem or question you’ve identified. If you don’t have a clear idea of the goal at this stage, don’t worry; you can refine it as you go along, but you don’t have much time so the earlier you start, the better.
  • Background: Review at least 5 papers on the topic you choose and provide a summary of the key results and methods used. Frame your problem in the context of this existing literature. To find 5 papers that are relevant to your project, you’ll have to go through much more than that. I recommend using Google Scholar to find relevant papers and only reading the abstracts for most of them, unless you find them very relevant to your project.

2. Data Description

  • Sources: Detail where and how you will collect your data. Include sources, methods of data collection, and any challenges you anticipate. It is perfectly acceptable to generate your own data if you plan to develop novel algorithms.
  • Data Type: Describe the type of data (e.g., images, text, numerical) and its relevance to your problem.
  • Resources: List the resources you will need to collect the data, and run your models. If you need to use Octopus (AUB’s high performance computing facility), please list the resources you need (e.g., number of hours, number of GPUs, etc.) and apply for them as soon as possible.

3. Process and Timeline

Break down the project into key steps (e.g., data collection, preprocessing, model building) with a concise timeline. These steps will give you an idea on how long your project will take and what resources you will need. I recommend starting as soon as possible!

4. Methodology

  • Approach: Describe the models, methods, or algorithms you plan to use (e.g., deep learning, regression analysis).
  • Justification: Explain why these methods are suitable for your problem.

5. Evaluation Strategy

  • Metrics: Define how you will evaluate the performance of your model (e.g., accuracy, precision, recall).
  • Validation: Explain how you will validate your results (e.g., cross-validation, test data).

6. Project Deliverables

  • Progress Report: A progress report on your project will be due on Tuesday April 15.
  • Final Presentation: A final presentation on your project will be due on the final week of classes April 29 - May 1.
  • Final Report: A final report on your project will be due on Monday May 5.

Submission Guidelines:

  • Length: Proposals should be 2 pages max.
  • Format: Please submit a PDF document to Moodle.
  • Due Date: The proposal is due on Tuesday March 18.