How can I use "loss" to determine how my model is doing?
When training my model in Keras after each epoch model outputs a values named "loss". I assume it somehow correlates with the accuracy of my model.
How can I use "loss" to determine how my model is doing?
xazshcdt
2020-10-15 11:04:18 UTC
report abuse
Answers
In short, loss is a sum of errors based on training and validation data. The less the sum of errors, the better your model is doing.
Thanks! Could you also explain how the loss is calculated?
Loss can be calculated using a variety of algorithms, but when we are talking neural networks, usually its negative log-likelihood or residual sum of squares. It also depends on what kind of model(regression or classification you are making).
Are you shure you want to close this post?
Are you shure you want to pin this post?
Report comment
Thank you for reporting this comment, the Imaginghub Community Management will verify your request.
Log in with Google