Skip to content

Inside a neural network

Recently I took part in a little competition to write a simple machine learning model, which should be able to classify about 120 different fruits.

Since the data itself is very beginner-friendly and I have to very little to none data cleaning work, the whole process was tent to exploring data and picking hyperparameters.

After approximately 7 hours of coding and waiting my (partly pretrained) model was finished I get some really interesting results. Following are some of its output of the network for a set of selected images.

 

I will also include some photos of how the data is processed during the process of prediction.

The first layer of the neural network.
The second layer, which is a convolutional layer for detecting edges
A simple activation layer with RELU
A layer of a network also contains a bias beside of normal neuron. This neuron will not take any input but its ‘opinion’ is still worthy.

The network structure itself is not an impressive work since I did not invent it myself. But behind the scene, there is a ton of works I put in in order to understand the topic, which is, to be honest, an immense computer science field even if you just go just a little in the depth.

Published inMachine Learning

Be First to Comment

Leave a Reply

Your email address will not be published. Required fields are marked *