Fractional Raised Complex Pooling

Real or complex data to the power of fractional or complex number is used to raise the real to complex pooling for improved training.

Signals from 5G or MRI classification is considered a supervised machine learning aiming to assign a class of properties to a signal. Due to the high dimensions of such signal data, machine learning approaches may have low accuracy. Especially when facing both low Signal to Noise Ratio (SNR) and high SNR situations at the same time. In this research, we use Convolutional and/or Temporal Transformer networks for classification. The fractional complex pooling shown here can improve the training and classification significantly with properly selected fractional biased rasier.