In this workshop we give a spin on time series analytics. In particular, we avoid the traditional computation of statistics to focus on patterns, motifs, discords, shapelets and so on. Many of the algorithms and use cases used in this workshop has been first published by the research team of Professor Eamonn Keogh.
To set-up the environment, yo need to install Docker on your machine. After that, you need to perform the following steps:
Go to the directory where you have clone this repository. If you have an NVidia graphics card and want to exploit its capabilities, follow GPU section, if you don't, just go throught the CPU section:
* CPU:
docker build --rm -t shapelets/timeseries-workshop .
docker run --rm -p 8888:8888 -v <absolute path where you cloned the repo>:/home/khiva-binder -ti shapelets/timeseries-workshop
open your browser and copy the url that is printed in the docker log, (similar to this one: http://localhost:8888/?token=<jupyter token>)
* GPU:
docker build --rm -t shapelets/timeseries-workshop . –f Dockerfile-cuda
docker run –rm --runtime=nvidia -p 8888:8888 -v <absolute path where you cloned the repo>:/home/khiva-binder -ti shapelets/timeseries-workshop
open your browser and copy the url that is printed in the docker log, (similar to this one: http://localhost:8888/?token=<jupyter token>)