Machine learning theory and practice is increasing in complexity as methods are applied to more challenging problems.  The principal focus of machine learning has been on maximizing decision performance.  For problems involving the control and allocation of resources there is a need for systems to be accurate and robust in estimation of the uncertainty associated with decisions.   Probabilistic inference focuses on proper calibration of probabilities and minimization of fluctuations in the estimates.  Assessment of the accuracy using the logarithmic scoring rule provides grounding in the rigor of information theory.  To satisfy this requirement  methods which manage the accuracy and robustness of low probability phenomena are of particular importance.  Generalized assessments using for instance the Renyi or Tsallis entropies, which can provide additional insight into the robustness of algorithms,  are of interest.  Papers are sought which evaluate the ability of Markov Chain Monte Carlo, probabilistic programming, and other advanced methods to achieve  accurate, robust probability inference.  Advances in this area are important for scientists, engineers, and other professionals seeking to apply the benefits of machine learning to complex problems and systems.
The goal of this session is to bring together professionals, researchers, and practitioners in the area of probabilistic inference to present, discuss, and share the latest findings in the field, and exchange ideas that address  the challenges and implications of accurate, robust machine learning methods.