# Difference between MLP(Multi-layer Perceptron) and Neural Networks?

I am wondering about the differences. Based on my understanding, MLP is one kind of neural networks, where the activation function is sigmoid, and error term is cross-entropy(logistics) error. Looking for help, thanks!

• i'm not 100% sure but the definition of MLP seems a bit vague, I've seen those two terms being used interchangeably, personally I always use NN to avoid ambiguity. Commented Apr 24, 2016 at 4:21

You are right, MLP is one kind of neural network.

There are several kinds of NN, you can have a NN based on Radial Basis Function with a Soft gating strategy, for example. You can use a committee machine strategy to form a NN...

• Thanks! And I am wondering if MLP is the NN whose activation function is sigmoid, and error term is cross-entropy(logistics) error? Commented Apr 24, 2016 at 0:11
• Yes, you need to use the sigmoid as activation functions, because there is no way to use gradient descendent in the hidden layers if you use a linear function as the activation function. You need to use derivations and once you use it in a linear function the result will always be 0... I didn't understand your cross-entropy question Commented Apr 24, 2016 at 14:59

MLP is fully connected feed-forward network. In particular CNN which is partially connected, RNN which has feedback loop are not MLPs.

The main difference is that MLP is one way. Thus, it's a feedforward network without any loop. Whereas, Neural networks such as DNN can contain loops. See more here

Multi-Layer Perceptron is a model of neural networks (NN). There are several other models including recurrent NN and radial basis networks. For an introduction to different models and to get a sense of how they are different, check this link out.

• Please explain what the link says and quote the most relevant parts. Where possible, try to give a proper reference -- or at least enough information that the information could be found again if the link moved. Commented Apr 24, 2017 at 4:15