Browse Source

Remove old paragraphs

icml
Markus Kaiser 3 years ago
parent
commit
f548b71987
  1. BIN
      dynamic_dirichlet_deep_gp.pdf
  2. 15
      dynamic_dirichlet_deep_gp.tex

BIN
dynamic_dirichlet_deep_gp.pdf

Binary file not shown.

15
dynamic_dirichlet_deep_gp.tex

@ -55,20 +55,7 @@ Building a truthful model of such data requires learning two separate models and
A similar example would be if our sensors that measures the lift are faulty in a manner such that we either get a correct reading or a noisy one.
Estimating a model in this scenario is often referred to as a \emph{data association problem}~\parencite{Bar-Shalom:1987, Cox93areview}, where we consider the data to have been generated by a mixture of processes and we are interested in factorising the data into these components.
\todo[inline]{I think this way of actually saying that the two problems are just the same will cut some introduction, what do you think?}
Many real-world dynamical systems can be in multiple different states of operation.
The available sensors often are not sufficient to identify the current state and datasets collected from such systems consequently contain a mixture of information from different sources.
Ideally, a machine learning method should identify these different sources and derive informative models for all situations.
Take faulty sensors as an example, which, at any time, might emit a correct reading or uninformative noise, as shown in \cref{fig:choicenet_data}.
Applying machine learning methods on such data often leads to model pollution, where the desired signal is hard to recover as the model simultaneously has to explain unrelated observations.
Estimating a model in this scenario is often referred to as a \emph{data association problem}~\parencite{Bar-Shalom:1987, Cox93areview}, where we consider the data to have been generated by a mixture of processes and we are interested in factorising the data into these components.
In a slightly different scenario, the goal is not to find signal in noise but rather to separate different signals and learn about each of them.
This problem arises in computer vision~\parencite{Cox93areview}, multi target tracking~\parencite{lazaro-gredilla_overlapping_2012} or industrial systems with latent effects~\parencite{hein_benchmark_2017} such as wear or defective parts which influence the underlying dynamics.
Here, both the different functions and the associations of the observations to a function need to be estimated.
In this work we will investigate a data set derived from the cart-pole benchmark, which contains trajectories of two instances of the benchmark with different pole lengths.
Our goal is to model the resulting multimodalities to yield both informative joint predictions but also extract informative models for these separate states.
\todo[inline]{Introduce model pollution? Cite Daniel? Mention Figure 1?}
\begin{figure}[t]
\centering

Loading…
Cancel
Save