Share via

Bug report: can't parametrize # hidden node or momentum in NN regression model

Adam Stewart 51 Reputation points
2020-07-14T20:22:22.25+00:00

I'm using the NN regression model in Azure ML Designer, but it appears to have a bug. If I switch trainer mode to ParameterRange, it allows me to use a semicolon-separated list of hyperparameters for learning rate and epochs. However, I'm unable to use a semicolon-separated list of hyperparameters for number of hidden nodes or momentum. Since the number of hidden nodes is the most commonly tuned hyperparameter for a MLP, this module doesn't seem particularly useful if I want to use it to tune hyperparameters.

Azure Machine Learning
Azure Machine Learning
An Azure machine learning service for building and deploying models.
2,657 questions
0 comments

2 answers

Sort by: Most helpful
Most helpful Newest Oldest
  1. YutongTie-MSFT 47,601 Reputation points
    2020-07-14T20:29:26.033+00:00

    Thanks for reaching out to us. We will investigate this deeper and let you know any update.

    Regards,
    Yutong

    0 comments
  2. Dave Patrick 426.3K Reputation points MVP
    2020-07-14T20:33:46.64+00:00

    You can provide this feedback over here on uservoice
    https://feedback.azure.com/forums/257792-machine-learning

    0 comments