Multi-layer network architectures, Computer Engineering

Multi-Layer Network Architectures:

As we considered we saw in the previous lecture that perceptrons have limited scope in the type of concepts that they can learn - but they can only learn linearly separable functions. Moreover, we can think of constructing larger networks by building them out of perceptrons. So in generally such larger networks that we call the step function units the perceptron units are in multi-layer networks.

However as with individual perceptrons there multi-layer networks can be used for learning tasks. In fact, the learning algorithm in which we look at as the backpropagation routine is derived mathematically by utilising differential calculus. Hence the derivation relies on having a differentiable threshold function that effectively rules out using perceptron units so if we want to be sure that backpropagation works correctly. Conversely the step function in perceptrons is not continuous, thus non-differentiable. If there an alternative unit was therefore chosen that had similar properties to the step function in perceptron units but that was differentiable.

Posted Date: 1/10/2013 7:39:48 AM | Location : United States







Related Discussions:- Multi-layer network architectures, Assignment Help, Ask Question on Multi-layer network architectures, Get Answer, Expert's Help, Multi-layer network architectures Discussions

Write discussion on Multi-layer network architectures
Your posts are moderated
Related Questions
Which method do you invoke on the DataAdapter control to load  your generated dataset with data? dataAdapter.Fill(ds). The beauty of this method is  it automatically implicitly

Expand CPI-C. CPI-C= Common Program Interface Communication.

A device that forwards data packet from one network to another is known as a  Switch device.

What are the specifications/characteristics used by the manufacturers to describe a digital to analog converter? Explain each one briefly. Ans: D/A converter characteristics

Explain the principles of operation of centralized SPC and distributed SPC and compare their performance. All the control equipment is replaced with a single processor that mus

Q. Describe about full adder? Let's take full adder. For this other variable carry from previous bit addition is added let'us call it 'p'. Truth table and K-Map for this is dis

In binary representation there are some states that are believed to be never occurred because of some particular functionality of a given circuit. If zero output is assigned to suc

Q. Explain about distributed systems? A Distributed System in that Data, Process and Interface component of information System are distributed to many locations in a computer n

Defines a macro Defines a macro with the given name, having as its value the given replacement text. After that (for the rest of the current source file), wherever the preproce

Q. Principles for developing successful system? (1)  Both developers and customers must be involved for accuracy in the information. (2)  A problem solving approach must be