Machine Learning Moves the Needle on Neural Science

Find out which data scientists won the first Decoding Brain Signals competition

  • Data Scientists Decode Human Perceptions from Brain Signals
  • When we can read and decode brain signals, we can help treat, heal, and retrain a brain after injury.”
  • Being able to decode human perceptions from brain signals can benefit this population greatly.
  • To learn more about upcoming competitions, click on this link or the image below: Cortana Intelligence Competitions .
  • Machine Learning Moves the Needle on Neural Science

Millions of people suffer from brain-related injuries and disorders every year. Being able to decode human perceptions from brain signals can benefit this population greatly. That’s what inspired Stanford University neurosurgeon Dr. Kai Miller to team up with Microsoft to offer the inaugural Cortana Intelligence Competition: Decoding Brain Signals. 


@MSFTnews: Find out which data scientists won the first Decoding Brain Signals competition

This post is by Chirag Dhull, Product Marketing and Hang Zhang, Senior Data Science Manager, at Microsoft.

Millions of people suffer from brain-related injuries and disorders every year. Being able to decode human perceptions from brain signals can benefit this population greatly. That’s what inspired Stanford University neurosurgeon Dr. Kai Miller to team up with Microsoft to offer the inaugural Cortana Intelligence Competition: Decoding Brain Signals. 

“The brain is an electrical organ with over 100 trillion synapses, connecting more than 87 billion neurons,” explains Dr. Miller. “Neuroscientists and computer scientists have been working to understand the secrets hidden in brain signals streaming directly from brains. When we can read and decode brain signals, we can help treat, heal, and retrain a brain after injury.”

The Decoding Brain Signals competition allowed machine learning experts and data scientists from around the world to test their skills while helping further the cause of neuroscience research. The contest asked participants to build intelligent models to decode electrical brain signals that were gathered from Dr. Miller’s research with epilepsy patients.

“I’ve worked with a lot of patients who have electrodes implanted in their brain for seizure detection,” explains Dr. Miller. “We did a series of experiments where we showed the patients pictures of faces and houses. Those two types of pictures produce electric activity in different brain areas. The purpose of the competition was to see if people could come up with inventive new algorithms that would allow us to decode what the patient had seen and give us new intuition into the underlying nature of these signals.”

More than 600 data scientists in over 170 countries responded to the challenge. They submitted over 1,800 solutions to tackle this problem. In the end, the three top winners managed to find solutions that were 10% more accurate than Dr. Miller’s solution.

“With the work that I do, there are different kinds of expertise that can contribute uniquely,” says Dr. Miller. “This competition was a creative way of harnessing the collective expertise of the Data Science community in a way that was fun.”

“It allowed us to ask ‘How good can it get if we apply power tools from data scientists?’ Contestants were able to optimize the mathematical approach used in my proof-of-principle experiment and the crowdsourced results will be useful for me and for my colleagues.”

To learn more about upcoming competitions, click on this link or the image below: Cortana Intelligence Competitions.

Microsoft provides its cloud-based ML technology platform, Azure Machine Learning (which is part of the Cortana Intelligence Suite) at no cost to contest participants. This platform is accessible from any browser, with rich drag-and-drop capabilities that help even new data scientists successfully build and deploy models in very little time.

The winning contestant to Cortana Intelligence’s inaugural competition was Alexandre Barachant from France. Alexandre holds a degree in electrical engineering and a PhD in signal processing. He says, “My background is more about signal processing than Machine Learning, but I caught the virus and now I can’t live without my daily dose of data science.”

In his entry, Alexandre exploited two types of brain activities: event-related potential and induced responses. His final winning solution is an ensemble of 5 different machine learning models.

All three winning solutions have been published in Cortana Intelligence Gallery. You can link to these them on the Decoding Brain Signals competition page. See how well you’re able to compete with contest participants, and learn from these winning solutions. You can also discover more by visiting this link: Tutorial Competition: Decoding Brain Signals , which provides an intuitive and step-by-step tutorial on how to build your first solution within minutes!

Chirag & Hang

Machine Learning Moves the Needle on Neural Science