Machine-Brain Intelligence in Brain-Computer Interfaces
summary
Claims that closed‑loop EEG BCIs require mutual adaptation: humans learn to produce signals while the decoder recalibrates online, typically after a 10 minutes calibration using simple classifiers like LDA or SVM. System uses EEG spatial and temporal filtering (roughly 8–30 Hz mu/beta bands), adaptive hyperplane shifts and feedback pacing; a prototype that subtly boosts perceived success to keep users in flow increases real performance, but continual retraining remains necessary and general large neural nets underperform on this sparse, nonstationary data.
AI-written summary · about
chapters (28)
- 0:00Intro
- 0:32About Jelena
- 1:01Projects on the Brain
- 2:26Outline
- 3:20Clinical definition
- 4:20Human computer interaction
- 4:46Principles
- 5:08Filtering
- 5:44Machine Learning
- 7:58Success
- 8:54Second Topic
- 9:10The Past
- 12:49Artificial Neural Networks
- 15:21Computational Model
- 15:40Cow
- 16:19Bayesian brain
- 17:04Bayesian inference
- 17:32Cat experiment
- 18:50Perceptual affordance
- 19:55Active inference
- 20:42Inspired from the brain
- 21:00BrainComputer Interface
- 22:01Optimal AI
- 22:53Emotional Intelligence
- 24:09Questions
- 25:31Why did you not implement Perceptron
- 27:17Training
- 28:32Recalibration
related talks
42mUnveiling the mysteries of Brain Computer Interfaces
18mAdding Human Learning in Brain--Computer Interfaces (BCIs): Towards a Practical Control Modality
09mNeuroAssist: Enhancing Cognition Computer Synergy with Adaptive AI & Enhance Neural Decoding
31mLearn to control your brain: Brain Computer Interfacing with Python
53mBrain: The ultimate hack
05mBrain Computer Interface: Reading Human Brains Using Deep Learning
11mConnected brains
26mBrainDrain Using Machine Learning and Brain Waves to Detect Errors in Human