Inception Module Classification

Inception Network Andrew Ng - Inception Network, Deeplearning AI

Inception Network Motivation Andrew Ng - Inception Network Motivation, Deep Learning AI

import numpy_neural_network as npnn
import npnn_datasets

model = npnn.Sequential()
model.layers = [
  npnn.Inception((28, 28, 1),
        2,
    2,  4,
    2,  4,
        2
  ),
  npnn.MaxPool(shape_in=(28, 28, 12), shape_out=(14, 14, 12), kernel_size=2),
  npnn.Inception((14, 14, 12),
        2,
    4,  6,
    4,  6,
        2
  ),
  npnn.MaxPool(shape_in=(14, 14, 16), shape_out=(7, 7, 16), kernel_size=2),
  npnn.Inception((7, 7, 16),
        2,
    6,  6,
    6,  6,
        2
  ),
  npnn.Dense((7, 7, 16), 140),
  npnn.LeakyReLU(140),
  npnn.Dense(140, 10),
  npnn.Softmax(10)
]

loss_layer = npnn.loss_layer.CrossEntropyLoss(10)
optimizer  = npnn.optimizer.Adam(alpha=1e-3)
dataset    = npnn_datasets.MNIST_28x28_2560()

optimizer.norm  = dataset.norm
optimizer.model = model
optimizer.model.chain = loss_layer

MNIST Handwritten Digits Classification using an Inception Module network

MNIST Handwritten Digits Classification using an Inception Module network
plot of network validation batch data target values (green) and predicted network output values (orange)