How will you differentiate between feed-forward neural networks as opposed to recursive neural networks?

Selecting which neurons to connect to which other neurons in the next layer in a neural network is important. What will be the effect of allowing maximal connectivity?

Demonstrate with an example, how, by engineering a clever training set, you can make your neural net a lot more effective?

Let’s say we want to minimize the squared error over all of the training examples that we encounter. we have a bunch of variables (weights) and we have a set of equations (one for each training example)? Can we just solve this problem by setting up a system of linear system of equations? Whatâ€™s the flaw in this approach?

Why linear neurons are not used much in practice?