Hands-On Machine Learning for Algorithmic Trading

How to define the model architecture

We also need to simplify the AlexNet architecture in response to the lower dimensionality of CIFAR10 images, relative to the ImageNet samples used in the competition. We use the original number of filters, but make them smaller (see notebook for implementation). The summary shows the five convolutional layers, followed by two fully-connected layers with frequent use of batch normalization, for a total of 21.5 million parameters:

_________________________________________________________________ Layer (type) Output Shape Param # ================================================================= CONV_1 (Conv2D) (None, 16, 16, 96) 2688 _________________________________________________________________ POOL_1 ...