Fast way to cluster time series data in R -

-

i'm trying cluster time-series data: have 16000 time-series vectors, each vector ~1500 samples long.

i tried using dtw package:

d = dist(x = time_series, method = "dtw") hclust(d)

however distance matrix calculation didn't finish running throughout whole weekend.

i'm looking faster way since data set larger.

your data on length 1500. suppose oversampled..

if downsample 1 in 2, dtw 4 times faster. if downsample 1 in 4, dtw 16 times faster. if downsample 1 in 10, dtw 100 times faster.

this might starting point.

are using cdtw or dtw? former significant faster, , can more accurate.

a paper in sigkdd week has faster way cluster dtw using upper , lower bounds [a].

however, matrix of size (16000 * 15999)/2.

so if have 2 days: 2 days / (16000 * 15999)/2 = 337 microseconds

so need each comparison in 337 microseconds, not lot of time. difficult..., doable effort. if stuck, email me (i last author of [a])

[a] nurjahan begum, liudmila ulanova, jun wang, eamonn keogh (2015). accelerating dynamic time warping clustering novel admissible pruning strategy sigkdd 2015


Comments

Post a Comment

Popular posts from this blog

python - pip install -U PySide error -

-
does know how avoid following error when running pip install -u pyside per official website instructions here: https://pypi.python.org/pypi/pyside/#installing-pyside-on-a-mac-os-x-system note have done brew install qt successfully. you using pip version 7.0.3, version 7.1.2 available. should consider upgrading via 'pip install --upgrade pip' command. collecting pyside using cached pyside-1.2.4.tar.gz installing collected packages: pyside running setup.py install pyside complete output command /applications/anaconda/bin/python -c "import setuptools, tokenize;__file__='/private/var/folders/dy/ttgqhqqx3g9bsbsnqnxsk2z80000gq/t/pip-build-cpxemt/pyside/setup.py';exec(compile(getattr(tokenize, 'open', open)(__file__).read().replace('\r\n', '\n'), __file__, 'exec'))" install --record /var/folders/dy/ttgqhqqx3g9bsbsnqnxsk2z80000gq/t/pip-0r7hri-record/install-record.txt --single-version-externally-managed --compile:
Read more

arrays - C++ error: a brace-enclosed initializer is not allowed here before ‘{’ token -

-
simple question: i'm trying initialize array within c++ class declaration: using namespace std; #include <string> class myclass{ public: string myarray[] = {"a","b","c"}; }; and i'm getting error: error: brace-enclosed initializer not allowed here before ‘{’ token no, without complied c++11 compiler, cannot initialize member array in declaration. have initialize array member in constructor. , don't use open array if know number of elements initialize array.
Read more

cytoscape.js - How to add nodes to Dagre layout with Cytoscape -

-
i trying add nodes dagre graph through cityscape.js. problem nodes not painted. code have: cy.add({ group: "nodes", data: { id:"n0",name:"n0",weight: 75 }, }); $.ajax(url).done(function(jsontree){ var newnodes=[]; for(var i=1;i<6;i++){ //var child = jsontree.result[i]; newnodes.push( { group: "nodes", data: { id: "n"+i, name:"n"+i}}, { group: "edges", data: { id: "e"+i, source: "n0", target: "n"+i } } ); } cy.add(newnodes); any ideas? have used loaded function reload graph not work. not know if have use specific function reload graph. thank in advance. you need add positions new nodes. cytoscape.js docs : it important note positions of newly added nodes must defined when calling cy.add(). nodes can not placed in graph without valid position — otherwise not displayed. yo
Read more