update_lambda

Update the forgetting factor <code>lambda</code> using the derivative of
a particular cost function

internal

An implementation of the adaptive forgetting factor scheme described in Bodenham and Adams (2016) <doi:10.1007/s11222-016-9684-8> which adaptively estimates the mean and variance of a stream in order to detect multiple changepoints in streaming data. The implementation is in 'C++' and uses 'Rcpp'. Additionally, implementations of the fixed forgetting factor scheme from the same paper, as well as the classic cumulative sum ('CUSUM') and exponentially weighted moving average ('EWMA') methods, are included.

Dean Bodenham

ffstream

Forgetting Factor Methods for Change Detection in Streaming Data

update_lambda function

<dl><dt>lambda_old</dt>
<dd>The old (previous) value of <code>lambda</code>.</dd>
<dt>signchosen</dt>
<dd>Either <code>+1</code> or <code>-1</code>. Usually <code>-1</code>.</dd>
<dt>alpha</dt>
<dd>The value of the step-size in the gradient descent.
In the paper this is named \(\epsilon\).</dd>
<dt>deriv_new</dt>
<dd>The value of the derivative at the current time.</dd></dl>

Arguments

Update lambda — update_lambda

<dl>

<dt>lambda_old</dt>
<dd>The old (previous) value of <code>lambda</code>.</dd>


<dt>signchosen</dt>
<dd>Either <code>+1</code> or <code>-1</code>. Usually <code>-1</code>.</dd>


<dt>alpha</dt>
<dd>The value of the step-size in the gradient descent.
In the paper this is named \(\epsilon\).</dd>


<dt>deriv_new</dt>
<dd>The value of the derivative at the current time.</dd>

</dl>

update_lambda: Update lambda

Description

Usage

Arguments