Hey guys I tried implementing Collaborative Filtering from Part 1 of FastAI course with Learners built from Part 2. But I am getting very high MSE(above 5), where as Jeremy’s implementation had very low MSE below 1. What am I doing wrong?
Hello,
Data Preprocessing: Ensure that your data preprocessing steps are consistent with those in the course. Any discrepancies in how the data is prepared can lead to significant differences in model performance.
Model Parameters: Verify that the parameters for your model (like n_factors, y_range, etc.) match those used by Jeremy. Even small differences can impact the Mean Squared Error (MSE).
Learning Rate: The learning rate can greatly affect the training process. Make sure you’re using the same learning rate and training schedule as in the course.
Validation Set: Ensure that your validation set is correctly set up and that it represents the data well. An improperly configured validation set can lead to misleading performance metrics.
Random Seed: If you’re using random splits for training and validation, setting a random seed can help ensure reproducibility and consistency with the course results.
Regularization: Check if you’re using the same regularization techniques (like weight decay) as in the course. Regularization can help prevent overfitting and improve generalization.
Best Regards
Esther598
Hello,
You double check that your model parameters (such as n_factors, y_range, etc.) are consistent with those used in the course materials. Even minor differences in these parameters can affect the Mean Squared Error. The learning rate is crucial for the training myfordbenefits process. Confirm that you are using the same learning rate and training schedule as specified in the course.
