These experiments show you how a neural network adapts itself during time to produce an almost correct output response in presence of a specific input request. In each case you can "Start", "Stop" and "Restart" the process (the latter just Starting again without Stopping). An "Iteration Number" constantly tells you how many times the neural network has been already tested with the provided examples. During each step a couple of number combinations is passed to the neural network as input and desired output and, immediately after that, the same neural network is asked to produce an output in relation to the same input just passed before. While you play with the different experiments, in an escalation of data complexity, you can appreciate the awesomeness of the approximation concept and contemplate why imperfection is at the very base of the beautiful perfection of the reality.
In this experiment the same "numbers combinations" are repeatedly used for input as well for output.
Input and desired output are visible on the left, within the white boxes, while the produced output is visible within the last gray box on the right.
In this experiment you can see how the neural network learns to produce an almost correct output response in presence of a known input request.
The neural network is trained constantly with the same input and output patterns.
Each iteration it is tested asking to produce the output response for the input request so that you can monitor the learning progress.
You can see how the neural network adapts itself a bit more for each iteration trying to match the expected response.
All the patterns used as input and output in the experiment are those reported in white background.
Below of them, on the left, there are the current training input/output patterns and, on the right, the current querying input/output patterns.
In the center instead there is the number of the current iteration.
To start or restart the experiment use the "Start" button.
To stop the experiment before its normal end, when the error is always equal or smaller than the configured amount, use the "Stop" button.