In this challenge, you will explore why the LeNet model in our example consistently achieves around 90% accuracy on the MNIST dataset. Your goal is to hypothesize, experiment, and explain the reasons behind this plateau in accuracy, and possibly suggest improvements.
Deliverables
Report Your Findings:
Summarize your observations about why the model is limited to 90% accuracy.
Include screenshots or plots (e.g., loss curves, accuracy comparisons, or visualized data) to support your conclusions.
Experiment Results:
Share the changes you made (if any) to improve accuracy.
Provide before-and-after accuracy metrics and explain why your adjustments worked (or didn’t work).
Propose Solutions:
Improvement of the current implementation that could help achieve higher accuracy.