BENV 0115 Machine Learning
ASSIGNMENT BRIEF
You have been given a classification problem. The dataset is a collection of images of works by street artists, including Banksy, Blek le Rat and others. The dataset also includes numerical values that characterise each image, as well as a metadata file that describes what they mean.
Your task is to use this dataset to create a model that can identify art by Banksy. You need to implement the learning algorithm using Python, make the predictions and write a short technical report describing the solution method and your findings. You should submit the Python code as a Google Colab Notebook in addition to the report.
You can use any machine learning technique, either covered in this module or not. You must use the provided dataset, but you can modify it in any way that helps create a better model, and even extend the dataset with further images if you wish.
Using ChatGPT (or equivalent)
A unique element of this report is that you are encouraged to use ChatGPT for part of the codewriting. You are expected to create part of your code using ChatGPT or an equivalent AI tool. You need to indicate which part of the code using your annotations in the Colab Notebook. Later, you will have to provide a short critical reflection (see section 6).
Word length and document structure
The report should have a maximum of 15 pages, including images and bibliography, but excluding the appendices. The report should include the following sections. Please see the marking rubric for further details about what is expected in each section.
1. Introduction.
This part should explain the problem you are solving and justify the approach you are taking. Include:
• A definition of the problem
• A justification of the interest or importance of the problem.
• Your objectives
2. Related work and possible applications to heritage.
This part should connect your problem with a few examples of related work, in heritage and beyond.
• Similar or analogous examples in heritage.
• If these do not exist, use examples from other fields and speculate about their potential use in heritage.
• Justify well the relationship between the cases you cite and the problem you are solving
3. Problem definition and dataset.
This part should describe the problem in machine learning terms, including the dataset and how it is prepared for machine learning.
• Describe the dataset with summary statistics, indicating its main features.
• Identify the input features.
• Explain normalisation, nonlinear transforms, and any other preparatory steps.
• Describe the sets for training, testing and validation (if applicable)
4. Methods and algorithms.
This part should explain and justify the machine learning algorithm in a manner that would allow anyone to reproduce what you did. Include:
• A description of the implemented methods. Equations are optional and should be used only when necessary to communicate concepts.
• The regularisation method (if applicable)
• The validation method (if applicable)
• The selection of hyperparameters
5. Results and discussion.
This part should evaluate the output of the method, in comparison with actual data. Include:
• The method to evaluate the predictions
• Appropriate error measures
• The model is evaluated critically, not only statistically, but also analysing the results in relation to the objectives
6. Critical reflection on AI contributions
This part should reveal which part of the code has been written using ChatGTP or an equivalent AI. You should include the following points:
• Describe how the AI was prompted to produce the intended result
• Describe any modifications and corrections that were necessary on the AI output
• Reflect critically on the following point: what are the strengths and weaknesses of
ChatGPT (or equivalent) for academic work?
7. References
List any literature cited in your report here. You can use any reference system.
咨询 Alpha 小助手，获取更多课业帮助