>>12214801>the theory behind neural network are tensorsTensors are the structure of which neural networks are being calculated. But neural networks aren't very associated with tensors above the fact that they are using the data structure to hold large amount of data (for instance several instances of images (x, y) * n amount of images, becoming a tensor [pixel x][pixel y][image n]).
If you want to truly "understand" neural networks, you're at a "gravity" sort of pitfall. It doesn't really have an explanation, some ai models just happens to sort of predict some things which we find value in. "Why" these values has these collective properties, we don't know, it's just an approximatively generated model of reality.
In short, it's about curve fitting. That you know what you want a curve to look like, but you don't know how to express it. It's somewhat similar to Maclaurin series, in that regard. Go dig deeper into gradient descent to see if you can find intuition there, but in general, it just takes time and repetition.
