BlackBox 2015 - 2015 Workshop on Black Box Learning and Inference

Probabilistic models have traditionally co-evolved with tailored algorithms for efficient learning and inference. One of the exciting developments of recent years has been the resurgence of black box methods, which make relatively few assumptions about the model structure, allowing application to broader model families.
In probabilistic programming systems, black box methods have greatly improved the capabilities of inference back ends. Similarly, the design of connectionist models has been simplified by the development of black box frameworks for training arbitrary architectures. These innovations open up opportunities to design new classes of models that smoothly negotiate the transition from low-level features of the data to high-level structured representations that are interpretable and generalize well across examples.
This workshop brings together developers of black box inference technologies, probabilistic programming systems, and connectionist computing frameworks. The goal is to formulate a shared understanding of how black box methods can enable advances in the design of intelligent learning systems.
# Call for Abstracts
We invite contributions of two types:
1. Systems Spotlights and Panel
Systems developers are invited to participate in a spotlight session and panel discussion on
? Probabilistic programming systems
? Neural net frameworks
? MCMC libraries
Contributors are invited to submit an abstract outlining the design goals of the system, its intended use cases and modeling syntax. We recommend that these abstracts include at least one code example that is explained in detail.
2. Research Contributions
We invite research abstracts on the following topics, to be presented as posters, contributed talks, or as short spotlight talks:
? Inference in probabilistic programming systems and broad model families:
? Gradient-based methods for parameter estimation, variational inference,
? Metropolis-Hastings variants with efficient rescoring,
? Message passing variants,
? Sequential Monte Carlo variants,
? Learning to infer: Using discriminative methods to amortize probabilistic inference (Variational Autoencoders, Deep Latent Gaussian Models, Restricted Boltzmann Machines, Neural Network based proposals),
? Model specification languages that use black box techniques (probabilistic programming languages, neural network libraries such as Torch, Theano, Caffe),
? Applications to vision, speech, reinforcement learning, motor control, language learning.
Submitted abstracts should be 2?4 pages in NIPS format, sent to nips2015blackbox-AT-gmail.com before October 9 2015 (midnight PDT). Submissions need not be anonymized. Additional space beyond the fourth page is permitted for references, if required.
Authors will be notified of acceptance and presentation type (poster, spotlight, talk) by November 2 2015. Final versions of submitted abstracts are due 2 December 2015 and will be distributed on the workshop website.
# Key Dates
? Abstract submission: 12 October 2015 (note: extended deadline)
? Acceptance notification: 2 November 2015
? Final abstract submission: 2 December 2015
? Workshop: 12 December 2015