Short Bytes: Deep Playground is an online visualization of how a neural network learns over the time and generates an output. Written in TypeScript using d3.js, you can use this app right now in your browser and tinker with a neural network.
Neural network is a technique of making a computer program that learns from data. Just like our human brain, with time, it learns new things and applies its knowledge to solve a problem.
If you wish to make a neural network, you’ll need a collection of software “neurons” that are connected and used to send and receive messages. Over the time, a neural network solves a problem over and over again, thus making the outcome closer to the most accurate result and eliminating the failure chances.
To visualize this learning process, some genius minds have made a Deep Playground app that shows an interactive visualization of neural networks.
This app is written in TypeScript using d3.js and you can go ahead and run it right now in your web browser. Note that this isn’t some official app from Google
As you press the start button shown at the top, a counter shows the number of iterations being performed over the time. You can also adjust the learning rate, problem time and choose your own data set.
From the main area in the middle, you can choose which properties you wish to feed in. The outputs from different neurons are mixed with varying weights and used to calculate the final value.
If you are wondering about the orange and blue colors, they show data, neuron, and weight values — blue is a positive weight, orange is a negative weight.
The description on the page app explains further:
In the hidden layers, the lines are colored by the weights of the connections between neurons. In the output layer, the dots are colored orange or blue depending on their original values. The background color shows what the network is predicting for a particular area. The intensity of the color shows how confident that prediction is.
Go ahead, try it here: Deep Playground