Neural Network Visualiser

Visualise and experiment with a simple neural network right here in your browser

Input Layer

Bias0.68
Input 10.50
Input 20.50

Hidden Layer 1

Bias-1.98
0.07
1.81
0.00
2.32

Output Layer

Output 14.33
Regression (Linear)

How to use?

Each input value can be altered by moving the slide. The network will update in real-time to reflect the new input values. Each node is governed by an activation function. Hidden layers use RELU, output nodes use either linear or Sigmoid. Clicking on the node will create display a graph of the activation function.

Experiment with adding, and removing layers and nodes to and from each layer. Each time the netowrk structure is changed, new weights and biases are randomly generated. The network will update in real-time to reflect the new structure.

The thickness of each nodes border and lines leaving a node represent the strength of its output signal (the larger the value, the thicker the line).

Why This Site Exists?

Neural networks are the foundation of modern AI, and understanding how they work is crucial for anyone interested in the field. Understanding the flow of information through a network can often be difficult to visualise. This simple tool is designed to help you visualise the flow of information through a simple neural network. You can experiment with different network configurations and see how they affect the network's output.

What is a Neural Network?

In the human brain, neurons work together by receiving and summing signals from other neurons. When the combined signal is strong enough, a neuron "fires" and passes information on to the next set of neurons. This summed interaction allows the brain to recognize complex patterns and make decisions.

An artificial neural network mimics this process by creating artificial "neurons" or nodes that are interconnected. Each node in the network receives inputs, sums them, and, if they exceed a certain threshold, activates to send signals forward through the network. By chaining these activations together, the network can process intricate information.

Training these networks with large datasets enables them to learn from experience. Each attempt refines the network's connections, reinforcing those that lead to better results and weakening less effective ones. Over time, this allows neural networks to recognize complex patterns, making them powerful tools in AI.

Where Can I Learn More?

Curious to dig deeper into neural networks? Here are two awesome resources that really bring the concepts to life:

  • 3Blue1Brown Neural Network Series – This YouTube series is hands-down one of the best for visual learners. 3Blue1Brown explains neural networks with stunning animations in an easy-to-follow and intuitive way.
  • Coursera: Machine Learning Specialization – If you want a structured, in-depth course, this specialization on Coursera is a fantastic choice. It covers all the machine learning essentials, with a solid section on neural networks, taught by top experts in the field.