how the metrics R^2 (coefficient of determination) can be used to evaluate the model accracy?

[XinWang-HNU]

dear professer:

I have been paying much attention on the dimension reduction technique, and recently, I try the architecure of parametric_umap with autencoding based on the Tf2.3 library. I want to the monitor the model accracy in the training process. How can i introduce the R^2 into the architecture of parametric_umap.

Looking forward to your reply.Thanks a lot.

[jc-healy]

Have you tried changing the parametric_reconstruction_loss_fcn? It currently defaults to tf.keras.losses.BinaryCrossentropy(from_logits=True) Maybe you can get away with specifying whatever variant you are interested in there?

On Thu, Nov 11, 2021 at 1:50 AM Xin Wang @.***> wrote:

dear professer:

I have been paying much attention on the dimension reduction technique, and recently, I try the architecure of parametric_umap with autencoding based on the Tf2.3 library. I want to the monitor the model accracy in the training process. How can i introduce the R^2 into the architecture of parametric_umap.

Looking forward to your reply.Thanks a lot.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/lmcinnes/umap/issues/800, or unsubscribe https://github.com/notifications/unsubscribe-auth/AC3IUWXPCKORF3RR23CPJPLULNRSLANCNFSM5HZW62NQ . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

[XinWang-HNU]

Have you tried changing the parametric_reconstruction_loss_fcn? It currently defaults to tf.keras.losses.BinaryCrossentropy(from_logits=True) Maybe you can get away with specifying whatever variant you are interested in there? … On Thu, Nov 11, 2021 at 1:50 AM Xin Wang @.***> wrote: dear professer: I have been paying much attention on the dimension reduction technique, and recently, I try the architecure of parametric_umap with autencoding based on the Tf2.3 library. I want to the monitor the model accracy in the training process. How can i introduce the R^2 into the architecture of parametric_umap. Looking forward to your reply.Thanks a lot. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#800>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AC3IUWXPCKORF3RR23CPJPLULNRSLANCNFSM5HZW62NQ . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

thank you very much. pefessor.

i have tried this parameter setting as following:

"parametric_reconstruction_loss_fcn= tf.keras.losses.MeanSquaredError(),"

i can get the training loss and validation loss. but i still can not obtain the value of R^2 that lies in [0,1].

i don't find the parameter in the umap, which is used to be verify the model accuracy. Is it set in the callbacks fucntion? and how to define the parameter .

Looking forward to your reply. Thanks a lot.