Você está na página 1de 22

Redes Neurais Artificiais

Rede Neural Multicamadas Feed-forward


Back-propagation

Projetos RNA Multicamadas Back-propagation

Choose your exploratory data set that contains data and labels of the class
elements. Train an MLP neural network with the backpropagation algorithm,
using 75% of the data examples for training and the remaining 25% for testing.
1. Explore the correlation between the learning rate and momentum and the
speed of training when a training error of 0.001 is chosen to stop the
training process. For example, use different values such as:
i. Learning rate 0.1, momentum 0.5
ii. Learning rate 0.5, momentum 0.5
iii. Learning rate 0.5, momentum 0.9
iv. Learning rate 0.5, momentum 0.1
Which values for the learning rate and momentum do you think are the best
for your neural network?
2. Explore the influence of the number of hidden nodes and number of layers
on the training and test error and on the time for training.
3

For example, you can use one of the following problems for the project:

Iris Classification
The Smoker and the Risk of Cancer
Gas Consumption Prediction
Handwritten Characters Recognition
Bank Loan Decision
Inverted Pendulum
Medical Diagnosis
Mortgage approval
Traveling Salesman
Resource Scheduling
Unemployment prediction
Musical signals-to-notes transformation
Playing ticktacktoe
Predicting beer sales
Stock Market Prediction
Water Flow to a Sewage Plant Prediction

Reconhecimento de Caracteres Escritos Mo


variabilidade de formas em que o nmero 3 pode ser escrito

padro pode ser representado por um conjunto de


aspectos: curvas, linhas retas, pontos, cor, ...
5

Classificao da Iris

150 amostras em trs classes: setosa, versicolor e virginica


4 atributos, medidos em centmetros, para cada amostra:
comp. sepal, comp. petal, larg. sepal e larg. petal.
7

SL - comprimento sepal, PL - comprimento petal


SW - largura sepal, PW - largura petal

Classificao Solos

seis tipos de solos, cada solo caracterizado


por diferentes concentraes de ions

Previso Consumo de Gs

X meses jan, abr, out dos anos 88-91


Y1 consumo de gs (gallons per capita)
Y2 temperatura mnima mdia (0C)

10

11

12

13

14

15

Diagnstico Mdico Baseado em Exemplos


Case

Sex

Age

Lumbar pain

Spinal column stiffness

Male

30

yes

yes

Women

23

not

not

Male

30

not

not

Male

70

not

not

Women

63

not

not

Male

59

not

not

Women

61

not

not

.
.
.

...

16

17

Diagnstico Mdico Baseado em Regras

4 regras envolvendo 4 manifestaes e 4 diagnsticos


18

Controle do Pndulo Invertido

Y fora aplicada no carro para frente e para trs

ngulo do pndulo com a vertical


velocidade angular

19

20

21

22

Você também pode gostar