Open seanlaw opened 4 years ago
@seanlaw Hi there ! I am a new contributor to this repository and would love to contribute by solving this issue. Could you please assign this issue to me ?
@wiredinhp Thank you for your interest. Please take a look at feel free to ask any clarifying questions.
@wiredinhp Note that you will need an NVIDIA GPU to test on. Please see this stump file as a cross-reference for Pearson Correlation but note that GPU code uses a different paradigm than CPU code.
@seanlaw Actually I am currently using a mac and I do not have NVIDIA GPU to test on but can I send a draft PR or specific commits in my forked repo to you so that you can verify and test on ?
@seanlaw Actually I am currently using a mac and I do not have NVIDIA GPU to test on but can I send a draft PR or specific commits in my forked repo to you so that you can verify and test on ?
In that case, you can try creating your PR and then using Google Colab, which has NVIDIA GPUs, to run our unit test suite on. Please see/copy this example notebook and ask follow up questions. We are here to help!
@seanlaw Sure !! I am aware of Google Colab though I was not sure whether it can handle the unit test suite but I will try and get back to you 👍
Currently,
gpu_stump
uses the sliding dot product approach to computing matrix profiles rather than Pearson Correlation. In theory, we should be process each row at a time (rather than traversing diagonals) and instead of storingQT
from the last row, we store the covariance from the previous row.Doing this should result in a nice speed improvement and makes the codebase more consistent with the
stump
functions since they also use Pearson Correlation