Asymmetric variate generation via a parameterless dual neural learning algorithm

Area of Science:

• Computational neuroscience
• Information theory
• Applied mathematics

Background:

• Previous work introduced a random number generator (RNG) utilizing a tunable non-linear neural system.
• The initial design employed a learning rule based on statistical principles and look-up table (LUT) implementation.

Purpose of the Study:

• To enhance the previously developed neural system-based random number generation method.
• To improve ease of implementation and address limitations of the prior approach.

Main Methods:

• Modification of the learning principle within the non-linear neural system.
• Retention of the efficient look-up table (LUT) based implementation strategy.

Main Results:

• The revised method demonstrates increased ease of implementation compared to the previous version.
• The new approach successfully relaxes certain limitations inherent in the earlier random number generation technique.

Conclusions:

• The updated random number generation method offers a more practical and versatile solution.
• Further development in neural network-based RNGs can lead to more robust and accessible systems.