AI and Machine Learning

A piece of information attached to a data sample that identifies its class or category, providing the ground truth for machine learning algorithms to learn from and use for future predictions.

The process of identifying raw data (such as neural signals or images) and adding one or more relevant labels to provide context, enabling supervised machine learning models to learn from the data and make accurate predictions.

A type of RNN capable of capturing long-term dependencies in neural signals, such as EEG time series.

A subfield of AI focused on enabling machines to learn from data and improve performance on specific tasks without explicit programming.

Learning the ability to rapidly adapt to new users or contexts using minimal training data.

Performance decline resulting from changes in the user's neural patterns over time.

Updating model parameters in real-time as new brain data arrives during active use.

An architecture specialized for time-series analysis, often used for dynamic decoding of time-varying neural signals.

Techniques aimed at preventing overfitting and ensuring model generalization in BCIs.

Reward-based training where a model learns to optimize performance based on feedback and neural responses.

Combining small amounts of labeled data with large amounts of unlabeled data to improve a model's generalization capabilities.

A type of ML that uses labeled datasets to train models for prediction or classification tasks.

A classic supervised learning algorithm used to separate mental states via optimal separating hyperplanes.

Reusing models pre-trained on one user or dataset to enhance decoding performance in another, reducing calibration time.

A machine learning technique where a model trained on one user or session is adapted for use on a new user or session. This helps to significantly reduce calibration time.

Discovering hidden patterns or groupings in unlabeled brain activity data.