Tensor Flow

TensorFlow is an open source software library for high performance numerical computation. Its flexible architecture allows easy deployment of computation Originally developed by researchers and engineers from the Google Brain team within Google's AI organization, it comes with strong support for machine learning and deep learning and the flexible numerical computation core is used across many other scientific domains.

Pyramid uses TensorFlow python library for analysis as part of its ETL process. The Pyramid ETL node uses the Deep Neural Net (DNN) functionality for modeling and predicting classification problems. Configure the following options:

  • Input Columns: must be a numeric column
  • Output Classifier: must be a nominal column (string)
  • Hidden Layers: the number of layers
  • Number of epoch: the number of times the learning set should be run
  • Dropout: a percentage of randomly selected neurons that will be ignored during training (to prevent overfitting). Select a number between 0 and 1.

Pyramid also supports model scaling by training and predicting the model with data streaming and thereby enabling training of the model with large scales of data.