Which objects reference my ViewController strongly?

Is there a more efficient way of building artificial neural nets other than using objects as nodes which reference other node objects?

• The problem that I'm having is that once the neural net gets to a certain size my program crashes due to stack overflow.

• Answer:

  Free connectivity recurrent neural networks can be represented by the matrix of their synaptic weights. If you have N neurons in your network, the network can be represented by a NxN matrix where the (i,j) element is the connection weight from neuron i to neuron j (or 0 if there is no connection). (If your network has a layered structure, you should probably instead reflect that structure into the data and store layer-to-layer connectivities as individual matrices.) If your network is densely connected, the matrix representation will be vastly more memory-efficient than chained lists (which is what you describe in the title). If you have a sparse connectivity using a "naive" matrix representation will be a waste of space but then your matrix will be sparse too, and there are some specific tools and libraries to work with sparse matrices. Doing that also allows you to treat most of the oprtations on a neural network as a linear algebra problem. The only nonlinear part in a neural network is usually the activation function, and you can parallelize its application. All that is already implemented in many math libraries in many language, and is likely to result in much quicker code too.