neural networks cover rule based in zero order logic (Was: Higher Order Logic Programming and Autograd)

A storm of symbolic differentiation libraries
was posted. But what can these Prolog code
fossils do?
Does one of these libraries support Python symbolic
Pieceweise ? For example one can define rectified
linear unit (ReLU) with it:
                 /   x      x  >= 0
     ReLU(x) := <
                 \   0      otherwise
With the above one can already translate a
propositional logic program, that uses negation
as failure, into a neural network:
NOT     \+ p             1 - x
AND     p1, ..., pn      ReLU(x1 + ... + xn - (n-1))
OR      p1; ...; pn      1 - ReLU(-x1 - .. - xn + 1)
For clauses just use Clark Completion, it makes
the defined predicate a new neuron, dependent on
other predicate neurons,
through a network of intermediate neurons. Because
of the constant shift in AND and OR, the neurons
will have a bias b.
So rule based in zero order logic is a subset
of neural network.
Python symbolic Pieceweise
https://how-to-data.org/how-to-write-a-piecewise-defined-function-in-python-using-sympy/
rectified linear unit (ReLU)
https://en.wikipedia.org/wiki/Rectifier_(neural_networks)
Clark Completion
https://www.cs.utexas.edu/~vl/teaching/lbai/completion.pdf