For those who had a hard time to study and understand classical estimation and detection algorithms, and unfortunately realized that these algorithms are simply ignored by many packages that have the same focus, e.g., scikit-learn, scipy, Keras/tensorflow, etc...

Just to list some of such "classical estimation and detection algorithms":

  • Detection:

    • Minimax
    • Neyman-Peason
    • Generalized likelihood ratio test (GLRT)
  • Estimation:

    • BLUE (Best Linear Unbiased Estimator)
    • Method of moments (MoM)
    • Kalman filtering

Such a classic detector as the Neyman Pearson detector cannot be found in any of these packages. Hence, my unique way out is to implement my own codes or, at most, compare my codes with a random low-rated package that is used by almost nobody.

On the other hand, algorithms from fields such as statistical machine learning, pattern recognition, and deep learning are readily implemented. For instance,

enter image description here

My unique goal is to understand why these methods aren't implemented in popular packages. I wonder whether such methods are worth it, after all.

  • $\begingroup$ Signal Processing Stack Exchange may be another relevant site for the question (but please avoid posting the same question on multiple Stack Exchange sites). $\endgroup$ Commented Jun 18, 2023 at 12:08