Here I maintain a handful of basic Python 3 functions, that are useful in multiple projects of mine.
For my study being often theoretical, I am not conern too much about huge tensors and numpy is the only dependence for this repository.
I attempt to keep these functions up-to-date with the newest version of Python 3 and numpy,
but I cannot promise and thus I choose the MIT license.
collapse(probas): takes in probabilities and outputs binary states.duoramp(x, low, high): takes in numbers and changes any out-of-range numbers to the lower or upper limit.logistic(x, temperature): takes in activations and outputs probabilities, common for artificial neurons.softmax(x, temperature): takes in activations and outputs probabilities, when a neural layer has lateral inhibition.entropy(P, base): takes in probabilities distributions and outputs their entropy values.
To those who are more experienced, knowledgeable and/or intelligent in programming than me, I sincerely hope that you would leave me any comments for correction and/or improvement, or suggest me any existing packages/modules that outperforms mine. Although I do not care too much about the efficiency, several of my projects are dependent on them so I need them to be absolutely correct. In addition, as the functions are used a lot, any tiny improvement could lead to potentially a bigger one.