Machine Learning Approaches to Mental Illness and Chronic Disease aims to leverage modern statistical machine learning techniques to investigate the structure of mental illness and its relationship to chronic illness. Particularly in psychiatry, both basic research and individualized treatment are hampered by a lack of valid diagnostic categories and accurate measurement of a patient’s disease state. We approach these questions by seeking patterns in both clinical records and data from laboratory experiments.
- Laboratory Experiments – For laboratory experiments with psychiatric patients and associated machine-learning based data analytics, the core has a service center that is available to the broader NJ ACTS community: The Rutgers-Princeton Center for Computational Cognitive Neuro-Psychiatry (CCNP). For more information, please visit the CCNP website or email CCNP@ubhc.rutgers.edu.
- Collaboration – The Core supports the broader NJ ACTS community by collaborating with other NJ ACTS researchers to apply the same machine learning-based techniques and software to investigate other problems in clinical research beyond mental health. Modern, scalable statistical approaches extend classic data analysis approaches (such as factor analysis and Cox regression) to large, irregularly structured datasets. We are releasing software to the NJ ACTS community for some of these analyses, and are also keen to collaborate on new data analysis questions. Please email Nathaniel Daw for further information.
Machine Learning Software for Online Studies Decision Making
Machine Learning Core software released on three github websites
- Code for server for online data collection: https://nivlab.github.io/docs/software/nivturk
- Code for decision making tasks to run in web browser: https://nivlab.github.io/docs/software/jspsych-demos
- Code for data analysis and model fitting: https://github.com/ndawlab/em