Nowadays we see a lot of research papers cluttered with complex
equations in neuroscience and neural networks. But they contain no
theorems that have some value in mathematics. Worse is that they use
mathematical terms such as "attractors" without definition.
When we increase the number of parameters we get a model that is
uglier but closer to experimental data. Do you know that this type of
research is called parameter physics? Threshold transformations in
neural systems are too hard to handle. But what is most needed is new
mathematics that resolve the difficulties.