Get the latest tech news
Building an EEG with a Children's Toy
How I Trained Neural Networks with Brainwave Data from a Children’s Toy.
After these failures, I decided to investigate how my interpretability engine could improve model inference efficiency in a way that may make it possible to distinguish signal from noise on-head. The analog electrode data is collected, amplified, and converted to a digital format by the implant, before being filtered to detect neuron spiking events, also known as action potentials. The interpretability engine proved effective in simplifying multi-layer perceptrons into singular perceptions for each input, allowing for a clear identification of parameter significance.
Or read this on Hacker News