s4548663-GFNet-ADNI

[YifeiZhang233eV]

Description

Dear 3710 team, this PR introduces and builds the GFNet model to classify brain MRI scans as either Alzheimer's Disease (AD) or Cognitive Normal (CN), using the ADNI dataset. Significant improvements have been made to ensure model accuracy, robustness, and memory efficiency.

Implementation

1. dataset.py: Load and preprocess the ADNI data within PyTorch custom data loader;
2. modules.py: Build a GFNet model structure;
3. train.py: Train a best model using the structure in modules.py by the data in the train set;
4. predict.py: Get the result of the test dataset by the trained model;
5. README.md: Including the description of the dataset and model structure, the final result with figures, future work if possible, and dependencies.

Testing

1. Run the train.py to get the trained model in ./models;
2. Run the predict.py to the accuracy result of the test set.

Notes:

1. Increased training time: This may lead to increased training time due to the use of a large number of data augmentation methods and small batch sizes.
2. Memory requirements: Although memory usage is controlled by reducing the batch size, the model still requires large GPU memory. If running on a low-profile device, you may need to adjust the batch size further.
3. Model complexity: Using multi-layer Transformer blocks and MLPS may make the model more complex, and the generalization ability on different datasets needs further validation.

Others

I submitted this homework late because I applied for an extension. I'm sorry for everyone's extra workload. Thank you for your comments. I will make sure to correct any problems in time.

[wangzhaomxy]

<This is an initial inspection, no action is required at this point.>

File Organizing: Well-organized files.

Problem Solving:

• The algorithm solves the problem appropriately.
• Accuracy in testing dataset: 0.7485.

Model and functions:

• It correctly uses PyTorch to construct the GFNet models and functions.
• With data augmentation.

Code design: Good

Code comment and docstring:

• Good code comments
• Minmal function docstrings

Difficulty: Hard.

Additional Comments:

• Reasonable commits
• Good ReadMe design

[aniketgupta17]

Observational Feedback

Pull Request:
Correctly created the Pull request from Topic Recognition Branch .
The pull request includes a clear description about the file structure .

File Organizing: Well-organized files.

Commit Log:
Commit messages are progressive for the Recognition Problem solved using 4 files .
Commits are regularly made, showing incremental development rather than bulk updates.

Documentation:
The README file could be made more readable with use of bullet points and organised code blocks .
Code comments are included.
Can add about Conclusion and future improvements in the ReadMe.

[gayanku]

Marking

Good/OK/Fair Practice (Design/Commenting, TF/Torch Usage)
	Good design and implementation.
	Spacing and comments.
	No Header blocks.	-1
Recognition Problem
	OK solution to problem. Accuracy in testing dataset: 0.7485.	-1
	Driver Script present.
	File structure present.
	Good Usage & Demo & Visualisation & Data usage.
	Module present.
	Commenting missing. Minimal doc strings	-1
	No Data leakage found.
	Difficulty : Hard. GFNet (Hard Difficulty)
Commit Log
	Good Meaningful commit messages.
	Good Progressive commits.
Documentation
	Readme :Good.
	Model/technical explanation :Good.
	Description and Comments :Good.
	Markdown used and PDF submitted.
Pull Request
	Pull Request has problems: Late submission	-2
	Feedback action required: Feedback marks possible +2 if the requested changes are made. Remove model checkpoint, cache files.	-2
	Request Description is good.
TOTAL		-7

Marked as per the due date and changes after which aren't necessarily allowed to contribute to grade for fairness.
Subject to approval from Shakes

[shakes76]

Approved extension +2

[shakes76]

Cant merge because of conflicting changes to main repo files (model files). Please update for merge, doesn't affect grade.