Shifting Focus
Taariq R -
This week, I shifted my focus towards running initial tests on my AI model after successfully setting up the development environment. With all necessary packages installed, I began training the model on a small dataset to evaluate its preliminary performance. This helped identify key areas for improvement, such as feature extraction and classification accuracy.
One major development was implementing a better loss function, which stabilized the training a lot. The model had unstable loss fluctuations before, but after I changed the loss function and other settings, I saw more stable curves. I also used data augmentation techniques like rotating images and changing contrast to allow the model to learn more from different versions of the scans. I was also able to fix the packaging issue, where my python processing version was not up to date, so I redownloaded the package with the latest version.
Despite these advances, I have been struggling with overfitting, where the model has performed ideally on training sets but not generalized to validation samples. To fix this, I might try regularization techniques. Another challenge has been trying to balance computational speed and accuracy to ensure the model is working as efficiently as it can be.
Soon, I plan to enlarge the dataset and perform more accurate testing with more labeled scans. I also would like to tune hyperparameters and try differing structures for better performance.
Overall, this week has been a milestone in moving from setup to actual testing, and I am happy to continue to refine the model for better early diagnosis.
Comments:
All viewpoints are welcome but profane, threatening, disrespectful, or harassing comments will not be tolerated and are subject to moderation up to, and including, full deletion.