diff --git a/1-STUMPY-basics/README.md b/1-STUMPY-basics/README.md index f91c605..9d5a832 100644 --- a/1-STUMPY-basics/README.md +++ b/1-STUMPY-basics/README.md @@ -1,9 +1,12 @@ # 1 - STUMPY Basics -Above what was seen in the tutorial I started on a general purpose motif -function which takes the dataset and a computed matrix profile and returns the -top motif it discovered and all locations where that motif appears. It does not -return overlapping motifs. +Above what was seen in the tutorial I started on a generalized motif and discord +extraction procedure. Since these are both really the same process over different +sort orders of the matrix profile we end up with a very clean functional setup +with a single function doing all the work and two wrappers 'deciding' which +calculation should be done. This allowed me to generalize marking discovered +discord and motif windows on my output plots, a process which should remain more +or less evergreen. ## TODO I remember reading that the distances returned by a matrix profile have an upper diff --git a/1-STUMPY-basics/Steamgen dataset.html b/1-STUMPY-basics/Steamgen dataset.html new file mode 100644 index 0000000..9334cf0 --- /dev/null +++ b/1-STUMPY-basics/Steamgen dataset.html @@ -0,0 +1,15096 @@ + + + + + +Steamgen dataset + + + + + + + + + + + + + + + + + + + + + +
+
+ + +
+
+
+ + +
+ +
+
+
+ + +
+
+
+ + +
+ +
+
+
+ + +
+
+
+ + +
+ +
+
+ + +
+ +
+
+
+ + +
+
+
+ + +
+ +
+
+ + +
+ +
+ + + + + +
+ +
+
+
+ + +
+
+
+ + +
+ +
+
+
+ + +
+
+
+ + +
+ +
+ + + + + +
+ +
+
+
+ + +
+
+
+ + +
+ +
+ + + + + +
+ +
+
+ + +
+ +
+ + + + + +
+ +
+
+
+ + +
+
+ + + + + + + + + diff --git a/1-STUMPY-basics/Steamgen dataset.ipynb b/1-STUMPY-basics/Steamgen dataset.ipynb index add9370..3892c56 100644 --- a/1-STUMPY-basics/Steamgen dataset.ipynb +++ b/1-STUMPY-basics/Steamgen dataset.ipynb @@ -1,5 +1,13 @@ { "cells": [ + { + "cell_type": "markdown", + "id": "cc72af0d", + "metadata": {}, + "source": [ + "This is the basic environment. stumpy is the main library we're working with, numpy and pandas are for all intents and purposes required to work with datasets and matplotlib allows us nice output." + ] + }, { "cell_type": "code", "execution_count": 1, @@ -16,6 +24,14 @@ "import datetime as dt" ] }, + { + "cell_type": "markdown", + "id": "a42ce3b2", + "metadata": {}, + "source": [ + "Basic utilities used to output the time it takes to compute a matrix profile over a given number of windows." + ] + }, { "cell_type": "code", "execution_count": 2, @@ -47,6 +63,14 @@ " return result" ] }, + { + "cell_type": "markdown", + "id": "e9e52f62", + "metadata": {}, + "source": [ + "More utilities to encapsulate the process of plotting our data." + ] + }, { "cell_type": "code", "execution_count": 3, @@ -74,6 +98,14 @@ " plot_.plot(profile[:, 0])" ] }, + { + "cell_type": "markdown", + "id": "07c68fbe", + "metadata": {}, + "source": [ + "Load the data!" + ] + }, { "cell_type": "code", "execution_count": 5, @@ -167,6 +199,14 @@ "steam_data[:4]" ] }, + { + "cell_type": "markdown", + "id": "347f81d3", + "metadata": {}, + "source": [ + "Automates the process of marking a motif or discord. Note that extracting a discord (or radically different sample) from a matrix profile is the same process as extracting a motif but with the matrix profile in reverse order allowing us to generalize into the threshold_extraction function. This function is also capable of marking every repeat in a motif group, rather than simply taking the first two values." + ] + }, { "cell_type": "code", "execution_count": 7, @@ -216,6 +256,14 @@ "\n" ] }, + { + "cell_type": "markdown", + "id": "07e3ce74", + "metadata": {}, + "source": [ + "Now we can actually do our calculations..." + ] + }, { "cell_type": "code", "execution_count": 8, @@ -240,6 +288,14 @@ "discord_list = get_discords(matrix_profile, motif_order, window, 0.005)" ] }, + { + "cell_type": "markdown", + "id": "9e4612eb", + "metadata": {}, + "source": [ + "... and report what we've discovered." + ] + }, { "cell_type": "code", "execution_count": 9, @@ -289,6 +345,14 @@ "plot_matrix_profile(axis[1], matrix_profile)\n", "mark_discovered(discord_list, axis[0], axis[1], window, max(steam_data['steam flow']))" ] + }, + { + "cell_type": "markdown", + "id": "432d3659", + "metadata": {}, + "source": [ + "The graphs show our dataset on top and the distance profile on bottom. The distance profile plots a sliding window's closeness to its nearest neighbor. The low points are places where the data is most similar to some other sample and the high points represent radically different patterns which appear nowhere else in the data. The matrix profile is simply the distance profile sorted from lowest distance value to greatest." + ] } ], "metadata": { @@ -307,7 +371,7 @@ "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", - "version": "3.10.9" + "version": "3.10.10" } }, "nbformat": 4, diff --git a/README.md b/README.md index 7d48b3e..42dcac8 100644 --- a/README.md +++ b/README.md @@ -11,17 +11,12 @@ copied in as I follow along, I have tried to mark as clearly as possible where I extrapolations. All data related to the tutorial is also mirrored in a data directory for each entry. ## 1 - STUMPY Basics -Above what was seen in the tutorial I started on a general purpose motif function which takes the dataset -and a computed matrix profile and returns possibly multiple motif groups. It takes 2 threshold parameters, -one for the absolute value of the matrix profile at the given point and the other for a percentage of the -maximum data magnitude. +[Human readable walkthrough.][3] +[Python code][4] -### TODO -I remember reading somewhere about an upper bound on matrix profile values. I should find that again -and calculate a percentage of the upper bound rather than having mp_thresh be an absolute value. -The Motifs class and function should be broken off into their own module for re-use elsewhere, including -Jupyter. [1]: https://www.cs.ucr.edu/%7Eeamonn/MatrixProfile.html "Resources and papers on the Matrix Profile" -[2]: https://stumpy.readthedocs.io/en/latest/tutorials.html "stumpy tutorial" \ No newline at end of file +[2]: https://stumpy.readthedocs.io/en/latest/tutorials.html "stumpy tutorial" +[3]: ./1-STUMPY-basics/Steamgen\ dataset.html +[4]: ./1-STUMPY-basics/portable.py \ No newline at end of file