https://github.com/harshilshah1804/cs203_assignment7

https://github.com/harshilshah1804/cs203_assignment7

Last synced: 2 months ago
JSON representation

• Host: GitHub
• URL: https://github.com/harshilshah1804/cs203_assignment7
• Owner: HarshilShah1804
• Created: 2025-03-15T10:07:53.000Z (3 months ago)
• Default Branch: main
• Last Pushed: 2025-03-15T12:32:30.000Z (3 months ago)
• Last Synced: 2025-03-15T13:20:36.870Z (3 months ago)
• Language: Jupyter Notebook
• Size: 80.3 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# CS 203 - Software Tools and Techniques for AI 

Assignment - 7 

Shah Harshil Hardik (23110132), Shah Tirth (23110295) 

Link to Github Repo: [Github](https://www.github.com/HarshilShah1804/CS203_Assignment7) 

1. Model Summary: 

   ![](readmeimages/Aspose.Words.c1f9843f-3953-4769-870f-e3a3a54a8aea.001.png)

2. Hyperparameters: 

1. Bag-of-words: 

- max\_features : 10,000 (CountVectorize) 

- Num\_layers = 5 

- Activation = Relu 

- Architecture: (Visible in the image above) 

- Dropout rate: 0.3 

- Learning rate (Training) = 1e-3 

- Learning rate (Fine-Tuning) = 1e-4 

- Epochs = 10 

- Loss\_fn = CrossEntropyLoss 

- Batch\_size (training + fine-tuning) = 64 

2. Embeddings: 

- Num\_layers = 5 

- Activation = Relu 

- Architecture: (Visible in the image above) 

- Dropout rate: 0.3 

- Learning rate (Training) = 1e-3 

- Learning rate (Fine-Tuning) = 1e-4 

- Epochs = 10 

- Loss\_fn = CrossEntropyLoss 

- Model = bert-base-uncased 

- Batch\_size (embeddings) = 16 

- Batch\_size (training + fine-tuning) = 32 

3. Logged Metrics: 

1. Accuracy: 

![](readmeimages/Aspose.Words.c1f9843f-3953-4769-870f-e3a3a54a8aea.002.jpeg)

2. Loss: 

![](readmeimages/Aspose.Words.c1f9843f-3953-4769-870f-e3a3a54a8aea.003.jpeg)

4. Final Evaluation Result:  

   1. Validation Accuracy (Embeddings) = 0.51 

   1. Validation Accuracy (Bag-of-words) = 0.90 

4. Confusion Matrix Visualization: 

1. Bag-of-words (STT2) 

![](readmeimages/Aspose.Words.c1f9843f-3953-4769-870f-e3a3a54a8aea.004.jpeg)

2. Bag-of-words (IMDB) 

   ![](readmeimages/Aspose.Words.c1f9843f-3953-4769-870f-e3a3a54a8aea.005.jpeg)

3. Embeddings (STT 2): 

![](readmeimages/Aspose.Words.c1f9843f-3953-4769-870f-e3a3a54a8aea.006.jpeg)

4. Embeddings (IMDB): 

   ![](readmeimages/Aspose.Words.c1f9843f-3953-4769-870f-e3a3a54a8aea.007.jpeg)

6. Training and Validation loss curves: 

![](readmeimages/Aspose.Words.c1f9843f-3953-4769-870f-e3a3a54a8aea.008.jpeg)

Legend: 

![](readmeimages/Aspose.Words.c1f9843f-3953-4769-870f-e3a3a54a8aea.009.png)