PoissonLoss

Compute the weighted Poisson loss function, which is <code>seg.mean</code> -
<code>count</code> * log(<code>seg.mean</code>). The edge case is when the mean is zero, in
which case the probability mass function takes a value of 1 when
the data is 0 (and 0 otherwise). Thus the log-likelihood of a
maximum likelihood segment with mean zero must be zero.

Computes optimal changepoint models using the
Poisson likelihood for non-negative count data,
subject to the PeakSeg constraint:
the first change must be up, second change down, third change up, etc.
For more info about the models and algorithms,
read "Constrained Dynamic Programming and Supervised Penalty Learning
Algorithms for Peak Detection"
<https://jmlr.org/papers/v21/18-843.html> by TD Hocking et al.

Toby Hocking

PeakSegOptimal

Optimal Segmentation Subject to Up-Down Constraints

PoissonLoss function

<dl> <dt>count</dt>
<dd>count</dd> <dt>seg.mean</dt>
<dd>seg.mean</dd> <dt>weight</dt>
<dd>weight</dd></dl>

Arguments

Toby Dylan Hocking &lt;toby.hocking@r-project.org&gt; [aut, cre]

Author

PoissonLoss — PoissonLoss

<dl>

 <dt>count</dt>
<dd>count</dd>

 <dt>seg.mean</dt>
<dd>seg.mean</dd>

 <dt>weight</dt>
<dd>weight</dd>

</dl>

PoissonLoss: PoissonLoss

Description

Usage

Arguments

Author

Examples