Visualizing high-dimensional data in the industry
For machines and the associated software, there are countless ways to collect and evaluate data. The resulting convolutes are high-dimensional and therefore difficult to interpret and visualize.
Deep neural networks
Deep neural networks reduce this multitude of dimensions, making data easier to process. In this way, for example, the data from machines with hundreds of sensors can be brought forward into a simple, easily understandable form without losing the inherent structure of the data.
To this end, the AF Institute has developed the Deep Autoencoder (a special kind of deep neural network), which can be used to simplify high-dimensional data and display it as point clouds in a three-dimensional virtual reality.
Via a VR controller, it is also possible to interact with the data in the point cloud, for example, by selecting and viewing individual points. This could identify deviations or (un) advantageous machine configurations and optimize machine work.