Scribd
Upload
38 views

WIX3001 Soft Computing Assignment 2

The document describes Assignment 2 for the WIX3001 Soft Computing course. Students will work in groups of 2-5 members to collect a handwritten character dataset where each character is labeled with the name of the person who wrote it. Each group member must contribute at least 50 handwritten characters. Students will then use soft computing methods programmed in Python like fuzzy logic, neural networks, or genetic algorithms to identify which group member wrote each character. Results will be analyzed and explained in a report with code in a Jupyter Notebook due by June 23rd. The report will be graded based on explanations of the dataset, methodology, results, and any explainable information generated from the model.

Uploaded by

Mr Moon
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
38 views

WIX3001 Soft Computing Assignment 2

The document describes Assignment 2 for the WIX3001 Soft Computing course. Students will work in groups of 2-5 members to collect a handwritten character dataset where each character is labeled with the name of the person who wrote it. Each group member must contribute at least 50 handwritten characters. Students will then use soft computing methods programmed in Python like fuzzy logic, neural networks, or genetic algorithms to identify which group member wrote each character. Results will be analyzed and explained in a report with code in a Jupyter Notebook due by June 23rd. The report will be graded based on explanations of the dataset, methodology, results, and any explainable information generated from the model.

Uploaded by

Mr Moon
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

WIX3001 Soft Computing

Assignment 2: Python Programming

1. Group Assignment: 2 – 5 members per group.


2. For this assignment, each group has to collect their own dataset similar to MNIST. Use this as a guideline:
https://www.kaggle.com/code/mikalaichaly/diy-mnist-dataset
3. Instead of labeling the characters as digits or alphabets, the characters are labeled with the name of the
person writing them. The objective of this assignment is to identify the user from their handwriting.
Example:
First set of characters: 0 1 2 3 4 5 6 7 8 9 A B C D … -> Labeled as Member1
Second set of characters: 0 1 2 3 4 5 6 7 8 9 A B C D … -> Labeled as Member2
…
4. Each member in the group must contribute 50+ handwritten characters, either digit, alphabet, or both.
Ideally each member should write each character multiple times.
5. When completed, the dataset should have X class labels, where X is the number of members in your group.
The dataset should look similar to the MNIST dataset.
6. Use soft computing methods to design a method to identify the user from their handwriting. Examples of
methods:
• Fuzzy logic.
• Neural networks.
• Genetic algorithms to optimize (fuzzy logic / neural network) parameters.
• Fuzzy extreme learning machine.
• Others…
7. The methods should be programmed using Python, and using Jupyter Notebooks to show your steps.
8. Write a report. It should contain the following:
• Details of your group members.
• Details of your dataset. How is it collected? What preprocessing methods are used? Show some
examples of handwritten characters from each member of your group.
• Details of the soft computing methods used. Example:
• Fuzzy logic: how many membership functions are used for input and output? How many
fuzzy rules are in the rule set? Which membership function is used and why? How is the
fuzzification and defuzzification performed?
• Neural networks: which NN model is used? What parameters are used (i.e. hidden layers,
number of nodes per layer)?
• What results are obtained? What metric did you use (i.e. Accuracy, Sensitivity) and what score did
you get?
• If you are using fuzzy logic, extract the fuzzy ruleset from your best model and show it using IF-
THEN format.
• If you are using neural networks, how do you extract explainable information from your model?
You can use methods like SHAP or LIME to generate explainable information.
• Or if you are using decision tree methods, then you can plot your model using some of the methods
listed here: https://mljar.com/blog/visualize-decision-tree/
• If you are using genetic algorithms, show the graph of the fitness score vs generations, similar to
what is done in Assignment 1.
• Analyze the results. Which digit or alphabet has the best and worst accuracy? Which member’s
handwriting has the best and worst accuracy? Show some examples of correct predictions and
wrong predictions.

9. Zip your report and the Jupyter Notebook file and upload to UM Spectrum by 23rd June 2023. Only the
group leader needs to submit the files.

Marking scheme (10%):

• Explained the dataset in detail: 1%


• Explained the soft computing methodology in detail: 5%
• Explained the results in detail: 2%.
• Generated explainable information: 2%.

You might also like