CODEWITHSUNDEEP
pythonmachine-learningscikit-learnneural-networkdeep-learning
hma
hma

Reputation: 566

Back propagation and Structure of a Neural Network in scikit-neuralnetwork

I am trying to learn Neural Networks using scikit-neuralnetwork framework and I know basics about Neural Networks and now trying to implement it with scikit-learn. but I am confused on 2 points.

1- what is the structure of this NN given below? Somehow, in some examples felt to me, some people don't put input layer as a layer. Otherwise, I am thinking this as a 2 layer NN has input layer with 100 nodes and 1 node at the ouput layer.

from sknn.mlp import Classifier, Layer

nn = Classifier(
   layers=[
    Layer("Maxout", units=100, pieces=2),
    Layer("Softmax")],
learning_rate=0.001,
n_iter=25)

nn.fit(X_train, y_train)

2- Does scikit-neuralnetwork do back propagation within the code that I put above?

Thank you!

Upvotes: 0

Views: 913

Answers (2)

hma
hma

Reputation: 566

1- In scikit-neuralnetwork framework, the input layer is not shown but it is creared in the background as a layer when we feed the system with training data. 

nn.fit(X_train, y_train)

for example, let's suppose we are using iris flower dataset. It has 3 classes so if that's the case framework understands that from the y_train and it creates 3 input layer with 3 classes. 

y_train = [0,0,0,1,2,]
X_train = [[ 7.1  3.   5.9  2.1]
           [ 5.9  3.   4.2  1.5]
           [ 5.5  2.4  3.7  1. ]
           [ 6.1  2.8  4.7  1.2]
           [ 5.   2.3  3.3  1. ]]

Except for the last layer, all of the other layers are hidden layer. I saw that when I manipulate the last layer, got this error

    nn = Classifier(
    layers=[
            Layer("Maxout", units=100, pieces=2),
            Layer("Softmax",units=1)],   -----> if units is 3 that's ok!
            learning_rate=0.001,
            n_iter=25)
    "Mismatch between dataset size and units in output layer."
    AssertionError: Mismatch between dataset size and units in output layer.

2- More detailed explanation; thanks to @robintibor at datascience.stackexchange. He explained both questions here Explanation

and thank you @user7534232 for your answer too :)

Upvotes: 0

user7534232
user7534232

Reputation: 11

1. Assuming that each training example in X_train has M features, and there are C classes in y_train: The input layer (not explicitly shown in the code) has M nodes. The hidden layer has 100 nodes. The output layer has C nodes (each one encoding the score for each class).

2. .fit() is a method that does that - feeds forward the training examples and uses back propagation to train the NN.

Also: perhaps you have to add units=C for the final layer - I assume this is a classification problem. If you need one value only (a score, not a class label), then use Regressor.

Upvotes: 1

Related Questions