Merge pull request #25 from cjekel/pcm

Don't scale length segments by max x and y with PCM

.github/workflows/ci.yml

@@ -9,7 +9,7 @@ jobs:
  runs-on: ubuntu-latest
  strategy:
  matrix:
- python-version: [3.5, 3.6, 3.7, 3.8, 3.9]
+ python-version: [3.6, 3.7, 3.8, 3.9]
 
  steps:
  - uses: actions/checkout@v2

CHANGELOG.md

@@ -4,6 +4,11 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.6.0] - 2022-10-08
+### Changed
+- `similaritymeasures.pcm` now produces different values! This was done to better follow the original algorithm. To get the same results from previous versions, set `norm_seg_length=True`. What this option does is scale each segment length by the maximum values of the curve (borrowed from the `curve_length_measure`). This scaling should not be needed with the PCM method because both curves are always scaled initially.
+- Fix docstring documentation for returns in `similaritymeasures.dtw` and `similaritymeasures.curve_length_measure`
+
 ## [0.5.0] - 2022-08-06
 ### Added
 - Added mean absolute error as `similaritymeasures.mae` thanks to a contribution by [HarshRaoD](https://github.com/HarshRaoD)

docs/similaritymeasures.html

@@ -332,7 +332,7 @@ <h1 class="title">Module <code>similaritymeasures.similaritymeasures</code></h1>
  return np.sum(area)
 
 
-def get_length(x, y):
+def get_length(x, y, norm_seg_length=True):
  r&#34;&#34;&#34;
  Compute arc lengths of an x y curve.
 
@@ -346,6 +346,8 @@ <h1 class="title">Module <code>similaritymeasures.similaritymeasures</code></h1>
  the x locations of a curve
  y : array_like
  the y locations of a curve
+ norm_seg_length : boolean
+ Whether to divide the segment length of each curve by the maximum.
 
  Returns
  -------
@@ -362,15 +364,20 @@ <h1 class="title">Module <code>similaritymeasures.similaritymeasures</code></h1>
 
  &#34;&#34;&#34;
  n = len(x)
- xmax = np.max(np.abs(x))
- ymax = np.max(np.abs(y))
 
- # if your max x or y value is zero... you&#39;ll get np.inf
- # as your curve length based measure
- if xmax == 0:
- xmax = 1e-15
- if ymax == 0:
- ymax = 1e-15
+ if norm_seg_length:
+ xmax = np.max(np.abs(x))
+ ymax = np.max(np.abs(y))
+
+ # if your max x or y value is zero... you&#39;ll get np.inf
+ # as your curve length based measure
+ if xmax == 0:
+ xmax = 1e-15
+ if ymax == 0:
+ ymax = 1e-15
+ else:
+ ymax = 1.0
+ xmax = 1.0
 
  le = np.zeros(n)
  le[0] = 0.0
@@ -397,7 +404,7 @@ <h1 class="title">Module <code>similaritymeasures.similaritymeasures</code></h1>
  num_data : ndarray (2-D)
  Curve from your numerical data.
 
- Retruns
+ Returns
  -------
  r : float
  curve length based similarity distance
@@ -587,7 +594,7 @@ <h1 class="title">Module <code>similaritymeasures.similaritymeasures</code></h1>
  return xi, eta, xiP, etaP
 
 
-def pcm(exp_data, num_data):
+def pcm(exp_data, num_data, norm_seg_length=False):
  &#34;&#34;&#34;
  Compute the Partial Curve Mapping area.
 
@@ -600,6 +607,10 @@ <h1 class="title">Module <code>similaritymeasures.similaritymeasures</code></h1>
  Curve from your experimental data.
  num_data : ndarray (2-D)
  Curve from your numerical data.
+ norm_seg_length : boolean
+ Whether to divide the segment length of each curve by the maximum. The
+ default value is false, which more closely follows the algorithm
+ proposed in [1]_. Versions prior to 0.6.0 used `norm_seg_length=True`.
 
  Returns
  -------
@@ -611,6 +622,9 @@ <h1 class="title">Module <code>similaritymeasures.similaritymeasures</code></h1>
  Your x locations of data points should be exp_data[:, 0], and the y
  locations of the data points should be exp_data[:, 1]. Same for num_data.
 
+ PCM distance was changed in version 0.6.0. To get the same results from
+ previous versions, set `norm_seg_length=True`.
+
  References
  ----------
  .. [1] Katharina Witowski and Nielen Stander. &#34;Parameter Identification of
@@ -640,8 +654,8 @@ <h1 class="title">Module <code>similaritymeasures.similaritymeasures</code></h1>
  xi1, eta1, xi2, eta2 = normalizeTwoCurves(exp_data[:, 0], exp_data[:, 1],
  num_data[:, 0], num_data[:, 1])
  # compute the arc lengths of each curve
- le, le_nj, le_sum = get_length(xi1, eta1)
- lc, lc_nj, lc_sum = get_length(xi2, eta2)
+ le, le_nj, le_sum = get_length(xi1, eta1, norm_seg_length)
+ lc, lc_nj, lc_sum = get_length(xi2, eta2, norm_seg_length)
  # scale each segment to the total polygon length
  le = le / le_nj
  le_sum = le_sum / le_nj
@@ -651,8 +665,8 @@ <h1 class="title">Module <code>similaritymeasures.similaritymeasures</code></h1>
  # make sure a is shorter than a&#39;, if not swap the defintion
  if lc_nj &gt; le_nj:
  # compute the arc lengths of each curve
- le, le_nj, le_sum = get_length(xi2, eta2)
- lc, lc_nj, lc_sum = get_length(xi1, eta1)
+ le, le_nj, le_sum = get_length(xi2, eta2, norm_seg_length)
+ lc, lc_nj, lc_sum = get_length(xi1, eta1, norm_seg_length)
  # scale each segment to the total polygon length
  le = le / le_nj
  le_sum = le_sum / le_nj
@@ -723,7 +737,6 @@ <h1 class="title">Module <code>similaritymeasures.similaritymeasures</code></h1>
  **kwargs : dict, optional
  Extra arguments to `metric`: refer to each metric documentation in
  scipy.spatial.distance.
-
  Some examples:
 
  p : scalar
@@ -746,7 +759,7 @@ <h1 class="title">Module <code>similaritymeasures.similaritymeasures</code></h1>
  The output array
  If not None, the distance matrix Y is stored in this array.
 
- Retruns
+ Returns
  -------
  r : float
  DTW distance.
@@ -1109,9 +1122,11 @@ <h2 id="parameters">Parameters</h2>
 <dt><strong><code>num_data</code></strong> :&ensp;<code>ndarray (2-D)</code></dt>
 <dd>Curve from your numerical data.</dd>
 </dl>
-<h2 id="retruns">Retruns</h2>
-<p>r : float
-curve length based similarity distance</p>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><strong><code>r</code></strong> :&ensp;<code>float</code></dt>
+<dd>curve length based similarity distance</dd>
+</dl>
 <h2 id="notes">Notes</h2>
 <p>This uses the OF2 method from [1]_.</p>
 <h2 id="references">References</h2>
@@ -1154,7 +1169,7 @@ <h2 id="examples">Examples</h2>
  num_data : ndarray (2-D)
  Curve from your numerical data.
 
- Retruns
+ Returns
  -------
  r : float
  curve length based similarity distance
@@ -1239,8 +1254,8 @@ <h2 id="parameters">Parameters</h2>
 <dt><strong><code>**kwargs</code></strong> :&ensp;<code>dict</code>, optional</dt>
 <dd>
 <p>Extra arguments to <code>metric</code>: refer to each metric documentation in
-scipy.spatial.distance.</p>
-<p>Some examples:</p>
+scipy.spatial.distance.
+Some examples:</p>
 <p>p : scalar
 The p-norm to apply for Minkowski, weighted and unweighted.
 Default: 2.</p>
@@ -1258,11 +1273,13 @@ <h2 id="parameters">Parameters</h2>
 If not None, the distance matrix Y is stored in this array.</p>
 </dd>
 </dl>
-<h2 id="retruns">Retruns</h2>
-<p>r : float
-DTW distance.
-d : ndarray (2-D)
-Cumulative distance matrix</p>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><strong><code>r</code></strong> :&ensp;<code>float</code></dt>
+<dd>DTW distance.</dd>
+<dt><strong><code>d</code></strong> :&ensp;<code>ndarray (2-D)</code></dt>
+<dd>Cumulative distance matrix</dd>
+</dl>
 <h2 id="notes">Notes</h2>
 <p>The DTW distance is d[-1, -1].</p>
 <p>This has O(M, P) computational cost.</p>
@@ -1331,7 +1348,6 @@ <h2 id="examples">Examples</h2>
  **kwargs : dict, optional
  Extra arguments to `metric`: refer to each metric documentation in
  scipy.spatial.distance.
-
  Some examples:
 
  p : scalar
@@ -1354,7 +1370,7 @@ <h2 id="examples">Examples</h2>
  The output array
  If not None, the distance matrix Y is stored in this array.
 
- Retruns
+ Returns
  -------
  r : float
  DTW distance.
@@ -1776,7 +1792,7 @@ <h2 id="notes">Notes</h2>
 </details>
 </dd>
 <dt id="similaritymeasures.similaritymeasures.get_length"><code class="name flex">
-<span>def <span class="ident">get_length</span></span>(<span>x, y)</span>
+<span>def <span class="ident">get_length</span></span>(<span>x, y, norm_seg_length=True)</span>
 </code></dt>
 <dd>
 <div class="desc"><p>Compute arc lengths of an x y curve.</p>
@@ -1789,6 +1805,8 @@ <h2 id="parameters">Parameters</h2>
 <dd>the x locations of a curve</dd>
 <dt><strong><code>y</code></strong> :&ensp;<code>array_like</code></dt>
 <dd>the y locations of a curve</dd>
+<dt><strong><code>norm_seg_length</code></strong> :&ensp;<code>boolean</code></dt>
+<dd>Whether to divide the segment length of each curve by the maximum.</dd>
 </dl>
 <h2 id="returns">Returns</h2>
 <dl>
@@ -1806,7 +1824,7 @@ <h2 id="examples">Examples</h2>
 <summary>
 <span>Expand source code</span>
 </summary>
-<pre><code class="python">def get_length(x, y):
+<pre><code class="python">def get_length(x, y, norm_seg_length=True):
  r&#34;&#34;&#34;
  Compute arc lengths of an x y curve.
 
@@ -1820,6 +1838,8 @@ <h2 id="examples">Examples</h2>
  the x locations of a curve
  y : array_like
  the y locations of a curve
+ norm_seg_length : boolean
+ Whether to divide the segment length of each curve by the maximum.
 
  Returns
  -------
@@ -1836,15 +1856,20 @@ <h2 id="examples">Examples</h2>
 
  &#34;&#34;&#34;
  n = len(x)
- xmax = np.max(np.abs(x))
- ymax = np.max(np.abs(y))
 
- # if your max x or y value is zero... you&#39;ll get np.inf
- # as your curve length based measure
- if xmax == 0:
- xmax = 1e-15
- if ymax == 0:
- ymax = 1e-15
+ if norm_seg_length:
+ xmax = np.max(np.abs(x))
+ ymax = np.max(np.abs(y))
+
+ # if your max x or y value is zero... you&#39;ll get np.inf
+ # as your curve length based measure
+ if xmax == 0:
+ xmax = 1e-15
+ if ymax == 0:
+ ymax = 1e-15
+ else:
+ ymax = 1.0
+ xmax = 1.0
 
  le = np.zeros(n)
  le[0] = 0.0
@@ -2223,7 +2248,7 @@ <h2 id="examples">Examples</h2>
 </details>
 </dd>
 <dt id="similaritymeasures.similaritymeasures.pcm"><code class="name flex">
-<span>def <span class="ident">pcm</span></span>(<span>exp_data, num_data)</span>
+<span>def <span class="ident">pcm</span></span>(<span>exp_data, num_data, norm_seg_length=False)</span>
 </code></dt>
 <dd>
 <div class="desc"><p>Compute the Partial Curve Mapping area.</p>
@@ -2235,6 +2260,10 @@ <h2 id="parameters">Parameters</h2>
 <dd>Curve from your experimental data.</dd>
 <dt><strong><code>num_data</code></strong> :&ensp;<code>ndarray (2-D)</code></dt>
 <dd>Curve from your numerical data.</dd>
+<dt><strong><code>norm_seg_length</code></strong> :&ensp;<code>boolean</code></dt>
+<dd>Whether to divide the segment length of each curve by the maximum. The
+default value is false, which more closely follows the algorithm
+proposed in [1]_. Versions prior to 0.6.0 used <code>norm_seg_length=True</code>.</dd>
 </dl>
 <h2 id="returns">Returns</h2>
 <dl>
@@ -2244,6 +2273,8 @@ <h2 id="returns">Returns</h2>
 <h2 id="notes">Notes</h2>
 <p>Your x locations of data points should be exp_data[:, 0], and the y
 locations of the data points should be exp_data[:, 1]. Same for num_data.</p>
+<p>PCM distance was changed in version 0.6.0. To get the same results from
+previous versions, set <code>norm_seg_length=True</code>.</p>
 <h2 id="references">References</h2>
 <p>.. [1] Katharina Witowski and Nielen Stander. "Parameter Identification of
 Hysteretic Models Using Partial Curve Mapping", 12th AIAA Aviation
@@ -2270,7 +2301,7 @@ <h2 id="examples">Examples</h2>
 <summary>
 <span>Expand source code</span>
 </summary>
-<pre><code class="python">def pcm(exp_data, num_data):
+<pre><code class="python">def pcm(exp_data, num_data, norm_seg_length=False):
  &#34;&#34;&#34;
  Compute the Partial Curve Mapping area.
 
@@ -2283,6 +2314,10 @@ <h2 id="examples">Examples</h2>
  Curve from your experimental data.
  num_data : ndarray (2-D)
  Curve from your numerical data.
+ norm_seg_length : boolean
+ Whether to divide the segment length of each curve by the maximum. The
+ default value is false, which more closely follows the algorithm
+ proposed in [1]_. Versions prior to 0.6.0 used `norm_seg_length=True`.
 
  Returns
  -------
@@ -2294,6 +2329,9 @@ <h2 id="examples">Examples</h2>
  Your x locations of data points should be exp_data[:, 0], and the y
  locations of the data points should be exp_data[:, 1]. Same for num_data.
 
+ PCM distance was changed in version 0.6.0. To get the same results from
+ previous versions, set `norm_seg_length=True`.
+
  References
  ----------
  .. [1] Katharina Witowski and Nielen Stander. &#34;Parameter Identification of
@@ -2323,8 +2361,8 @@ <h2 id="examples">Examples</h2>
  xi1, eta1, xi2, eta2 = normalizeTwoCurves(exp_data[:, 0], exp_data[:, 1],
  num_data[:, 0], num_data[:, 1])
  # compute the arc lengths of each curve
- le, le_nj, le_sum = get_length(xi1, eta1)
- lc, lc_nj, lc_sum = get_length(xi2, eta2)
+ le, le_nj, le_sum = get_length(xi1, eta1, norm_seg_length)
+ lc, lc_nj, lc_sum = get_length(xi2, eta2, norm_seg_length)
  # scale each segment to the total polygon length
  le = le / le_nj
  le_sum = le_sum / le_nj
@@ -2334,8 +2372,8 @@ <h2 id="examples">Examples</h2>
  # make sure a is shorter than a&#39;, if not swap the defintion
  if lc_nj &gt; le_nj:
  # compute the arc lengths of each curve
- le, le_nj, le_sum = get_length(xi2, eta2)
- lc, lc_nj, lc_sum = get_length(xi1, eta1)
+ le, le_nj, le_sum = get_length(xi2, eta2, norm_seg_length)
+ lc, lc_nj, lc_sum = get_length(xi1, eta1, norm_seg_length)
  # scale each segment to the total polygon length
  le = le / le_nj
  le_sum = le_sum / le_nj

docs/version.html

@@ -26,7 +26,7 @@ <h1 class="title">Module <code>similaritymeasures.version</code></h1>
 <summary>
 <span>Expand source code</span>
 </summary>
-<pre><code class="python">__version__ = &#34;0.5.0&#34;</code></pre>
+<pre><code class="python">__version__ = &#34;0.6.0&#34;</code></pre>
 </details>
 </section>
 <section>