Brain-computer interfacing (BCI) has attracted rapidly increasing research interest in the last decade due to recent advances in neurosciences, wearable/mobile biosensors, and analytics. The ultimate goal is to provide a pathway from the brain to the external world via mapping, assisting, augmenting or repairing human cognitive or sensory-motor functions. Recently, many advanced machine learning technologies have appeared, such as deep learning, transfer learning, and so on. Deep learning method has achieved great success in image and video analysis, natural language processing, speech recognition, etc., and recently has also started to find applications in BCI. Transfer learning, which improves learning in a new task by leveraging data or knowledge from other relevant tasks, can be particularly useful in BCI to cope with variability across individuals or tasks, accelerating learning and improving performance. Advanced deep-transfer-leveraged learning technologies can also be integrated to take advantage of both domains.
Although the studies of BCI using the advanced deep learning and/or transfer learning become more and more popular, there still exist many unsolved fundamental problems so far, such as deep learning representation of some EEG-based BCI data from multiple modalities, mapping data from one modality to another to achieve cross-source BCI data analysis, identifying and utilizing relations between elements from two or more different signal sources for comprehensive BCI data analysis, fusing information from two or more signal sources to perform a more accurate prediction, transferring knowledge between modalities and their representations, and recovering missing modality data given the observed ones.
In the past decade, several EEG-based BCI methods and technologies have been developed and shown promising results in some real-world examples such as neuroscience, medicine, and rehabilitation, which led to a proliferation of papers showing accuracy/performance and comparison, but most do not advance to real-time, translation or application. Then, these papers do not fare well, either because of lack of novelty (known technique) or no bio/med/experiment/clinical translation. For all the reasons mentioned above, it inspires us to exploit and develop effective advanced deep learning and/or transfer learning algorithms for addressing fundamental issues in BCI and rehabilitation fields.
The list of possible topics includes, but not limited to:
• Deep feedforward/belief/residual networks for BCI
• Generative adversarial networks for BCI
• Transfer learning for BCI
• Domain adaptation for BCI
• Deep-transfer learning for BCI
• BCI based healthcare system
• Highly interpretable intelligent system for BCI
Brain-computer interfacing (BCI) has attracted rapidly increasing research interest in the last decade due to recent advances in neurosciences, wearable/mobile biosensors, and analytics. The ultimate goal is to provide a pathway from the brain to the external world via mapping, assisting, augmenting or repairing human cognitive or sensory-motor functions. Recently, many advanced machine learning technologies have appeared, such as deep learning, transfer learning, and so on. Deep learning method has achieved great success in image and video analysis, natural language processing, speech recognition, etc., and recently has also started to find applications in BCI. Transfer learning, which improves learning in a new task by leveraging data or knowledge from other relevant tasks, can be particularly useful in BCI to cope with variability across individuals or tasks, accelerating learning and improving performance. Advanced deep-transfer-leveraged learning technologies can also be integrated to take advantage of both domains.
Although the studies of BCI using the advanced deep learning and/or transfer learning become more and more popular, there still exist many unsolved fundamental problems so far, such as deep learning representation of some EEG-based BCI data from multiple modalities, mapping data from one modality to another to achieve cross-source BCI data analysis, identifying and utilizing relations between elements from two or more different signal sources for comprehensive BCI data analysis, fusing information from two or more signal sources to perform a more accurate prediction, transferring knowledge between modalities and their representations, and recovering missing modality data given the observed ones.
In the past decade, several EEG-based BCI methods and technologies have been developed and shown promising results in some real-world examples such as neuroscience, medicine, and rehabilitation, which led to a proliferation of papers showing accuracy/performance and comparison, but most do not advance to real-time, translation or application. Then, these papers do not fare well, either because of lack of novelty (known technique) or no bio/med/experiment/clinical translation. For all the reasons mentioned above, it inspires us to exploit and develop effective advanced deep learning and/or transfer learning algorithms for addressing fundamental issues in BCI and rehabilitation fields.
The list of possible topics includes, but not limited to:
• Deep feedforward/belief/residual networks for BCI
• Generative adversarial networks for BCI
• Transfer learning for BCI
• Domain adaptation for BCI
• Deep-transfer learning for BCI
• BCI based healthcare system
• Highly interpretable intelligent system for BCI
