README.md

One dimensional CNN library

Usage

• MakeWeight(x, y)

  • Make weight array. (w[x][y])

• Open_data(filename)

  • Open [filename] file.

FFNN Function

Function for calculate node.

• Logistic_Func(x)

  • Calc logistic function. It's used as activation function.

• FullConect_Func(x, w)

  • Full connect layer's Calc function.

• Conv_Func(x, w)

  • Convolution layer's Calc function.

• Max_Pool_Func(x, kernel_size)

  • Max pooling layer's Calc function.

Pass NN function

• Pass_FC(old_layer, new_layer, w)

  • Pass full connect layer automatically.

• Pass_Conv(old_layer, new_layer, w)

  • Pass convolution layer automatically.

• Pass_Max_Pool(old_layer, new_layer, kernel_size)

  • Pass max pooling layer automatically.

• Pass_FC_Out(old_layer, new_layer, w)

  • Pass Output(full connect) layer automatically.

BPNN Function

• Dif_Logistic_Func(x)

  • Differential logistic function.

• Cross_Entropy(x, d)

  • Calculate cross entropy from output node.

Calculate delta

• FC_Delta(node, delta, w)

  • Calculate full connect layer's delta. (For next bp node)

• Conv_Delta(node, delta, w)

  • Calculate convolution layer's delta. (For next bp node)

• Max_Pool_Delta(layer, delta)

  • Calculate max pooling layer's delta. (For next bp node)

Update node

• FC_Update(x, delta, w)

  • Update full connect layer's weight.

• Conv_Update(node, delta, w)

  • Update convolution layer's wieght.

example

$python main.py learn_data train_data