It’s called Cebra and it’s a new machine learning system capable of reconstructing the hidden structure of the neural code. The discovery published in the journal Nature was made by a group of scientists from the Federal Polytechnic School of Lausanne (Epfl). The team led by Mackenzie Mathis subjected mice to black and white footage of a man running towards a car. Using the signals produced by the visual cortex of mice, the Cebra algorithm has made it possible to train a deep learning model called “deep learning” which decodes what the animal is looking at.
Understand how the brain processes information
The algorithm was able to read the stimuli received by the brain and reconstruct a sequence of images similar to the one projected for animals on another screen. “This work – argues Mathis – represents a substantial step forward towards the algorithms applied to neurotechnology and mind-computer interfaces (BMI).” The brain signals are obtained directly by measuring brain activity through electrodes inserted in the area of the visual cortex of the mouse brain. Cebra’s ultimate goal will be to identify a structure in complex systems and help us understand how the brain can process information. “The potential clinical applications are exciting and are not limited to neuroscience research,” concludes the researcher .