Comparing the performance of MLP and CNN on USPS dataset and visualizing it via TensorBoard.
Language used: Python
Libraries used: torch, torchvision, sklearn
Dataset: USPS dataset
Dataset Visualization:
![](https://private-user-images.githubusercontent.com/83514527/338695677-666201a7-c621-4516-bfc5-0e327a9f9938.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3MjAwMjcyNzksIm5iZiI6MTcyMDAyNjk3OSwicGF0aCI6Ii84MzUxNDUyNy8zMzg2OTU2NzctNjY2MjAxYTctYzYyMS00NTE2LWJmYzUtMGUzMjdhOWY5OTM4LnBuZz9YLUFtei1BbGdvcml0aG09QVdTNC1ITUFDLVNIQTI1NiZYLUFtei1DcmVkZW50aWFsPUFLSUFWQ09EWUxTQTUzUFFLNFpBJTJGMjAyNDA3MDMlMkZ1cy1lYXN0LTElMkZzMyUyRmF3czRfcmVxdWVzdCZYLUFtei1EYXRlPTIwMjQwNzAzVDE3MTYxOVomWC1BbXotRXhwaXJlcz0zMDAmWC1BbXotU2lnbmF0dXJlPTgyYjFhZmNiNjBjMGJmZjVmOWFhNjliMzVkMGE3OTI1MzY2MGNlMTY2NWZkZDQ5NjI0ZmJjMWIxYTk2YzZkZjQmWC1BbXotU2lnbmVkSGVhZGVycz1ob3N0JmFjdG9yX2lkPTAma2V5X2lkPTAmcmVwb19pZD0wIn0.sBNHQNRPszKWmzTG2yY18exEtt1aDmPTDuqUKktfO0U)
Observations:
Model | Epochs | Train Loss | Test Accuracy | Precision | Recall |
---|---|---|---|---|---|
MLP | 10 | 0.0373 | 0.9312 | 0.9265 | 0.9243 |
CNN | 10 | 0.0060 | 0.9497 | 0.9470 | 0.9449 |
Confusion Matrix for MLP and CNN:
![](https://private-user-images.githubusercontent.com/83514527/338697092-a557515c-1ad3-4e6b-9101-e7f07eb1571f.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.m24TgGTKBUlhpdGJ7aulNwE3a4ETVUF7WLO56nfLSx4)
![](https://private-user-images.githubusercontent.com/83514527/338697279-1077acb5-8306-4f02-8c57-6eaa6ebf27f7.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.S13_jXHpMYOPPwQCwfIboBDahCrQFoGHTubH6_NhaIs)
Precision and Recall:
![](https://private-user-images.githubusercontent.com/83514527/338697760-d6165378-da8f-480e-bea9-2a620f971a47.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.nswvuzz5v_Uyrs-h6ysO4Fzd9TnzWrW1cHru2v-LsRs)
![](https://private-user-images.githubusercontent.com/83514527/338697843-95892ce8-b76b-415f-a41a-df1137b04756.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.vP0AbkjDn78gnqWHDa1aWX9c-o3QnjuWkq_EJjPoc5I)