search

paul-tqh-nguyen / reuters_topic_labelling

Deep learning to automatically label Reuter's articles with the relevant topics.
1 stars 0 forks source link

Track the best F1 score we've gotten so far. #11

Closed paul-tqh-nguyen closed 4 years ago

paul-tqh-nguyen commented 4 years ago

Since we've gotten https://github.com/paul-tqh-nguyen/reuters_topic_labelling/commit/45bf67c559808163278a26c334fc657411a71a1a to implement support for F1 score gathering for https://github.com/paul-tqh-nguyen/reuters_topic_labelling/issues/10, we are now able to accurately measure performance.

Since searching the hyperparameter space and determining the best neural network architecture for our goals are tasks that are yet to be complete, it makes sense that we create useful tools to aid in that endeavor.

One such useful tool is a global tracking of the hyperparameters and scores of the best performing models so far.

We can globally track progress by logging to a revision-backed file a new score and a new set of hyperparameters if that new score beats the score currently logged in the file. This updating can happen during training.

Let us implement this idea.

paul-tqh-nguyen commented 4 years ago

https://github.com/paul-tqh-nguyen/reuters_topic_labelling/commit/1cf0c24e657ed017a52f881453637b8273fb2884 accomplishes this task. Let us mark this task as complete!