~~NOCACHE~~
/* DO NOT EDIT THIS FILE             */

[[en:seminaires:acs:index|Analysis and conception of systems]]\\
Monday January 7, 2019, 10:30AM, Salle 3052\\
**Louis Mandel** (IBM Watson) //Reactive Probabilistic Programming//
<button type="link" size="xs" collapse="b18-3485">{{icon>info-circle?lg&color=#777}}</button>\\
<collapse id="b18-3485" collapsed="true">
<well size="sm">
The idea of Probabilistic Programming is to use the expressiveness of
programming languages to build probabilistic models. To implement this
idea, a common approach is to take a general purpose language and
extend it with (1) a function that allows to sample a value from a
distribution, (2) a function that allows to condition values of
variables in a program via observations, and (3) an inference
procedure that build a distribution for a program using the two
previous constructs.

Following this approach, we propose to extends a reactive programming
language with probabilistic constructs. This new language enables the
modeling of probabilistic reactive systems, that is, probabilistic
models in constant interaction with their environment. Examples of
such systems include time-series prediction, agent-based systems, or
infrastructure self-tuning.

To demonstrate this approach, we started from ReactiveML, a reactive
extension of OCaml that supports parallel composition of processes,
communication through signals, preemption, and suspension.  We extend
the language with classic probabilistic constructs (sample, factor à
la WebPPL) and propose an inference scheme for reactive processes,
that are, non-terminating functions.
</well>
</collapse>