mymoza / beatpd-clsp-jhu Goto Github PK

BeatPD-CLSP-JHU

Jupyter Notebook 74.72% Python 17.12% Shell 8.16%

beatpd-clsp-jhu's Introduction

Hi there 👋

Hello! I’m Marie-Philippe Gill.

I am a content creator on Instagram, YouTube, and my blog GirlKnowsTech.com!

Professionally, I work full-time as a machine learning engineer. I graduated from software engineering and a Master (M. Sc.) at École de Technologie Supérieure (ÉTS) in Montreal. I also have a college degree (DEC) in Computer Science.

I completed an internship at Johns Hopkins University. My research was on using smartwatches to predict the severity of Parkinson’s Disease symptoms, the repo is public.

Pronouns: She & Her

beatpd-clsp-jhu's People

Contributors

Stargazers

Watchers

beatpd-clsp-jhu's Issues

Create softlink to /export/ instead of Sonal home folder to store the data

Test if generating kfold_predictions_*.csv files works

Just test & make sure the generation of these files works.

530d677

Verify that "Get First Derivative" still works

In prepare_data.ipynb

Clean & Good Example to extract + preds on tsfresh

Launch i-vectors training for REAL-PD

Check the highpass filter

Check if the highpass filter is working good enough

Casting to int twice

In pcq_knn_bpd, I'm casting VLTest and vTestSubjectId to int twice.

Could be improved :)))

Speed up getting derivative with convolution?

It takes a while right now to get the velocity vectors

Error message with REAL-PD

When running this:

do_work = partial(
    get_first_derivative, 
    derivative_path="real-pd.training_data.derivative_original_data/"+data_real_subtype,
    padding=True, 
    mask_path=data_dir+'/real-pd.training_data.high_pass_mask/'+data_real_subtype
)

num_jobs = 8
with ProcessPoolExecutor(num_jobs) as ex:
    results = list(ex.map(do_work, df_train_label['measurement_id']))

I get weird errors, to investigate:

measurement_id  7d3f4b7a-167f-4a26-9062-94ce9d8794c1
measurement_id  99af8d14-cd09-4107-9502-355378ba4e08

measurement_id  c0445431-35c3-402a-9b71-1c413604327b
measurement_id  e07708ff-7b8d-4070-af70-3aa81423ab5b

Docstring format

Make the docstring format uniform

Get masks for REAL-PD database

pcc_knn_bpd and pca_everyone_svr_bpd.py have duplicated code

Both files have duplicated code

Verification: Count the number of files

In get_excel_results, the verification of the number of files is not completed.

No need to dump nb of files for training, it's in the len of labels

len(vLTrai_subjectid), len(vLTest_subjectid)

Could use directly len(predictionsTrai)

[predictionsTrai,vLTrai_subjectid,predictions,vLTest_subjectid, vTraiMeasurementId, \
 mse_trai_subjectid, \
 mse_test_subjectid, \
 nb_files_trai_subjectid, nb_files_test_subjectid, \
 lTestMeasId_subjectid] = pickle.load(open(fold_folder+sFileName, "rb" ) )