Tensorflow Algorithm Requirements

In order for the digital twins to use a TensorFlow algorithm, the algorithm must meet the following requirements:

  • The input should be a vector of float named input_vector.
    • The size of the input vector is limited to 10 float values.

  • The output is named output_0.
    • It is a float vector of size 1, whose value represents whether the prediction is an anomaly or not.

    • If the output value is lower than 0.5, it is considered not an anomaly. If it is greater or equal to 0.5, it is considered an anomaly.

  • The TensorFlow model has been converted to an ONNX file.

See Converting a TensorFlow model to an ONNX file for details on how to convert your TensorFlow model to ONNX.