Home
Discussions
Community Q&A
performance of algorithm in train set
Lefteris
Good evening
I would like to ask if I can get the performance (average error) of the algorithm in the train set to compare with the performance in validation set;
Thanks in advance
Find more posts tagged with
AI Studio
Accepted answers
jacobcybulski
Yes, it is done commonly, especially when optimising models, to check that the algorithm still had the capacity to learn (performance on a training set) vs its practical use (performance on a validation set). Simply apply the model to both the training and validation sets, and then run each result via a performance operator. Note that you can always utilise copies of the training data which emerges on output from the trained model to immediately apply the model to it, and then a copy of the model which emerged on output from Apply Model to be applied the second time to the validation data set. If this is done in an optimisation loop, which may be iterating over some model parameter (e. g. k of knn), then typically we would also use a Log operator to log the model parameter vs the performance on training and validation.
Lefteris
thank you so much
jacobcybulski
for the useful information
you are awesome
All comments
jacobcybulski
Yes, it is done commonly, especially when optimising models, to check that the algorithm still had the capacity to learn (performance on a training set) vs its practical use (performance on a validation set). Simply apply the model to both the training and validation sets, and then run each result via a performance operator. Note that you can always utilise copies of the training data which emerges on output from the trained model to immediately apply the model to it, and then a copy of the model which emerged on output from Apply Model to be applied the second time to the validation data set. If this is done in an optimisation loop, which may be iterating over some model parameter (e. g. k of knn), then typically we would also use a Log operator to log the model parameter vs the performance on training and validation.
Lefteris
thank you so much
jacobcybulski
for the useful information
you are awesome
Quick Links
All Categories
Recent Discussions
Activity
Unanswered
日本語 (Japanese)
한국어(Korean)