Multi-Layer Perceptron

• What is the math behind like?
• Class diagram for MLP
• How to use it?

What is the math behind like?

Class diagram for MLP

How to use it?

• Declare a MLP object

  mlp = MLP()
  

• Input and target data

  x = np.linspace(0,1,40).reshape((40,1))
  x = (x-0.5)*2
  y = np.sin(2*np.pi*x) + np.cos(4*np.pi*x) + np.random.randn(40).reshape((40,1))*0.2
  train = x[0::2,:]
  test = x[1::4,:]
  valid = x[3::4,:]
  traintarget = y[0::2,:]
  testtarget = y[1::4,:]
  validtarget = y[3::4,:]
  

• Declare layers

  X = Layer(value = train)
  target = Layer(value = traintarget)
  h1 = Layer(rows = X.rows, cols = 3)        # first hidden layer
  h2 = Layer(rows = X.rows, cols = 2)        # 2nd hidden layer
  h3 = Layer(rows = X.rows, cols = 1)        # 3rd hidden layer
  output = Layer(rows = X.rows, cols = target.cols)
  

• Connect layers

  X.connect(h1)
  h1.connect(h2)
  h2.connect(h3)
  h3.connect(output)
  

• Add layers to MLP and train

  mlp.addLayers([X, h1,
               h2,
               h3,
               output, target])
  mlp.train()
  

• Predict

  test_x = np.ones((1,1))
  output = mlp.predict(x=test_x)
  print(output)