Multi-task Learning for Predicting Health , Stress , and Happiness

Multi-task Learning (MTL) is applied to the problem of predicting next-day health, stress, and happiness using data from wearable sensors and smartphone logs. Three formulations of MTL are compared: i) Multi-task Multi-Kernel learning, which feeds information across tasks through kernel weights on feature types, ii) a Hierarchical Bayes model in which tasks share a common Dirichlet prior, and iii) Deep Neural Networks, which share several hidden layers but have final layers unique to each task. We show that by using MTL to leverage data from across the population while still customizing a model for each person, we can account for individual differences, and obtain state-of-the-art performance on this dataset.

[1]  P. Rousseeuw.  Silhouettes: a graphical aid to the interpretation and validation of cluster analysis , 1987 .

[2]  Andrew P. Smith,et al.  Psychological stress and susceptibility to the common cold. , 1991, The New England journal of medicine.

[3]  Rich Caruana,et al.  Multitask Learning , 1998, Encyclopedia of Machine Learning and Data Mining.

[4]  Zoubin Ghahramani,et al.  Propagation Algorithms for Variational Bayesian Learning , 2000, NIPS.

[5]  Adrian Furnham,et al.  Personality, self-esteem, and demographic predictions of happiness and depression , 2003 .

[6]  Ashish Kapoor,et al.  Multimodal affect recognition in learning environments , 2005, ACM Multimedia.

[7]  Thomas G. Dietterich,et al.  To transfer or not to transfer , 2005, NIPS 2005.

[8]  Lawrence Carin,et al.  Multi-Task Learning for Classification with Dirichlet Process Priors , 2007, J. Mach. Learn. Res..

[9]  Alan Fern,et al.  Multi-task reinforcement learning: a hierarchical Bayesian approach , 2007, ICML '07.

[10]  Shiliang Sun,et al.  Neural network multitask learning for traffic flow forecasting , 2008, 2008 IEEE International Joint Conference on Neural Networks (IEEE World Congress on Computational Intelligence).

[11]  R. Veenhoven,et al.  Healthy happiness: effects of happiness on physical health and the consequences for preventive health care , 2008 .

[12]  Arthur C. Graesser,et al.  Multimodal semi-automated affect detection from conversational cues, gross body language, and facial features , 2010, User Modeling and User-Adapted Interaction.

[13]  L. Wisk,et al.  Does the perception that stress affects health matter? The association with health and mortality. , 2012, Health psychology : official journal of the Division of Health Psychology, American Psychological Association.

[14]  N. Lane,et al.  MoodScope: building a mood sensor from smartphone usage patterns , 2013, MobiSys '13.

[15]  Nitish Srivastava,et al.  Dropout: a simple way to prevent neural networks from overfitting , 2014, J. Mach. Learn. Res..

[16]  Samuel Kaski,et al.  Multi-task and multi-view learning of user state , 2014, Neurocomputing.

[17]  Alex Pentland,et al.  Daily Stress Recognition from Mobile Phone Data, Weather Conditions and Individual Traits , 2014, ACM Multimedia.

[18]  Akane Sano,et al.  Predicting students' happiness from physiology, phone, mobility, and behavioral data , 2015, 2015 International Conference on Affective Computing and Intelligent Interaction (ACII).

[19]  Akane Sano,et al.  Prediction of Happy-Sad mood from daily behaviors and previous sleep history , 2015, 2015 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC).

[20]  Akane Sano,et al.  Multi-task , Multi-Kernel Learning for Estimating Individual Wellbeing , 2015 .