The goal of the project is to develop a machine learning-based AI system and to evaluate its performance. You should use at least one of the machine learning models that are covered in this course.
The system should be implemented in a numerical computation or neural network library: PyTorch is recommended, but you may use Tensorflow, Keras or other appropriate libraries. You may use an existing codebase as starting point for your implementation (with appropriate acknowledgement), as long as you train, validate and test the model yourself. As a platform to train and run models, including to train neural networks, Google Colab is recommended. Most neural networks train more efficiently on GPUs (which you can do on colab, although there are time restrictions on training time), but you can also just use your own machine and you are not required to train a model on a GPU.
You may work in groups of 1 to 3 students. If you have questions about the project please ask them as early as possible.
The project will consist of the following steps:
Choose a dataset to work with, Copies of the first two datasets are available on Vula under Resources.
a) Covid-19 Daily Reports dataset https://github.com/CSSEGISandData/COVID-19Daily case reports.
b) TED Talks dataset https://www.kaggle.com/rounakbanik/ted-talks Transcripts of TED talks and metadata. Basic natural language processing will be required to process the dataset.
c) City of Cape Town Open Data Portal https://web1.capetown.gov.za/web1/OpenDataPortal Various datasets related to Cape Town. You will need to pick an appropriate dataset (or a combination of datasets).
d) Choose your own dataset (email the lecturer for approval).
2. Formulate the machine learning problem, i.e. the prediction task. You should formulate
the task as a multi-class classification problem that predicts some attribute in the dataset. You may aggregate a range of discrete or continuous values in the data into a single class if appropriate. Binary classification is allowed, but if there are multiple classes or a range of values in the data you should not reduce it to binary classification. Decide what the inputs and output of the model should be, and split the data input training, validation and test sets. Choose an appropriate evaluation metric. Here you also need to consider if there are any ethical issues associated with building an AI system for this task.
3. Pick a baseline. This should be the simplest possible approach to the problem.
4. Develop the model: Decide which type of model (or neural network architecture) to use
and which features or input representation to use. Implement and train the model. Tune hyperparameters and evaluate different choices of features or model architectures. Perform the final evaluation.
5. Analyse the model’s performance. If the model has a low accuracy, that could be a perfectly valid outcome and you won’t be penalized for that by itself. What is more important is that you show evidence that you carefully considered the various modelling choices, optimized the model on the chosen dataset, evaluated the model appropriately
and analyzed what its shortcomings might be.
The project has the following deliverables (submit as a single zip file):
Project report: Suggested length and format: 4 pages in ACM format. The report should be a comprehensive description of your work: Address everything entioned above. You need to include a problem description, explain the baseline, model design choices, experimental setup, and results on both the validation set (for tuning) and the test set (for final evaluation), and an analysis of your model’s performance.
Code and data: Submit the code for the project. If you use external code as part of the project that should be indicated clearly. Submit the test data, as well as your model’s final predictions on the test set. Include instructions or a script for reproducing your results.
Get solution of this project at affordable price. please send your query at contact@codersarts.com we'll share you price quote
How can you contact us for assignment Help.
Via Email: you can directly send your complete requirement files at email id contact@codersarts.com and our email team follow up there for complete discussion like deadline , budget, programming , payment details and expert meet if needed.
Website live chat: Chat with our live chat assistance for your basis queries and doubts for more information.
Contact Form: Fill up the contact form with complete details and we'll review and get back to you via email
Codersarts Dashboard : Register at Codersarts Dashboard, and track order progress
