This blog explains how we built a low-cost drone to create a local AQI map.

Introduction

This term was mentioned in the Batch Normalization paper. The author’s described it in the following way We define Internal Covariate Shift as the change in the distribution of network activations due to the change in network parameters during training

In a neural network, The output of the first layer feeds into the second layer , output of second feeds into third and so on. When the parameters of layers change, so does the distribution of input to subsequent layers.

Covariate shift refers to change in input distribution to a learning system. Now, a change in input to the first layer of the model, gets amplified to subsequent layers, Hence even a small change in input distribution of the entire model can result in a large change in input distribution of deeper layers.

Explanation

Explanation