Convolutional Neural Networks

A convolutional neural network (CNN, or ConvNet) is a type of feed-forward artificial neural network in which the connectivity pattern between its neurons is inspired by the organization of the animal visual cortex. Individual cortical neurons respond to stimuli in a restricted region of space known as the receptive field. The receptive fields of different neurons partially overlap such that they tile the visual field.

Statistical Invariance — Things that don’t change over time or space.

‘Same’ padding equation:
 out_height = ceil(float(in_height)/ float(strides[1]))
 out_width = ceil(float(in_width)/ float(strides[2]))

‘Valid’ padding equation:
 out_height = ceil(float(in_height — filter — height + 1)/ float(strides[1]))
 out_weight = ceil(float(in_width — filter — width + 1)/ float(strides[2]))


Advanced ConvNet-ology

Pooling:

  1. Max Pooling:
     — Parameter — free!
     — Often more accurate
     — More expensive
     — More hyperparamters — pooling size and pooling strides
  2. Average Pooing:
     — Y = mean(Xi)

1 x 1 convolutions: 1 x 1 convolutions ⇔ Matrix multipliers

Inception modules:

At each layer of your convNet yu make a choice. Have a pooling operation or convolution or 1 x 1, and on on top concatenate the output of each other. We choose in such a way so as to decrease the parameters so that the model performs better than a simple convolution.

Alright that’s it for now! Thank you for spending your time. Cheers!

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.