Deep Feed forward Neural Networks

Also known as feedforward neural networks, or multilayer perceptrons(MLPs)
1. As it name says, is an a perceptron with multiple layers
2. Are the quintessential deep learning models
PERCEPTRON
1. The perceptron is an algorithm for supervised learning of binary classifiers (functions that can decide whether an input, represented by a vector of numbers, belongs to some specific class or not), It is a type of linear classifier,
2. Was invented in 1957 at the Cornell Aeronautical Laboratory by Frank Rosenblatt
3. Was one of the first artificial neural networks to be produced
4. Perceptrons could not be trained to recognise many classes of patterns, due this, the researches invented the DFF NN
5. Can be trained by a simple learning algorithm that is usually called the delta rule
Some features
1. Don't form a cycle, are different to recurrent
2. You can build only by combining many layers of single perceptron
3. was the first and simplest type of artificial neural network devised.
4. It can aproximate almost any function
5. The information moves in only one direction, forward, from the input nodes, through the hidden nodes (if any) and to the output nodes.
6. It can resolve non linear problems
7. the overall number of layers gives the DEPTH of the model, the name deep learning arose from this terminology
  1. Final layer is the output layer
  2. The training examples specify directly what the output layer must do at each point of an input (p)
    1. The behavior of the other layers is not directly speciﬁed by the training data. this layers are called HIDDEN LAYERS
8. Backpropagation is the most used algorithm to learn in this network
Form the basis of many important commercial applications.
1. The convolutional networks used for object recognition from photos are a specialized kind of feedforward network
2. Some examples
  1. Airline Marketing Tactician
  2. Backgammon
  3. Data Compression
  4. Driving – ALVINN
  5. ECG Noise Filtering
  6. Financial Prediction
  7. Speech Recognition
  8. Sonar Target Recognition
Disadvantages
1. Sometimes need a lot of training time
2. it's bad extrapolating
3. The existence of local minimums in the error function makes training difficult

Next up

Deep Feed forward Neural Networks

Description

Resource summary

Similar

	Created by Rodrigo Burciaga about 6 years ago