Incremental Machine Learning/Online ML with River Python (ML on Data in Motion)

[@sandromartins3579]

Thank you! very helpful.

[@cedriclebocq1060]

nice tutorial as usual, thx

[@tariqlimouni9642]

Thanks for video , please I have a question can I use LSTM model weights and do the same framework for time series forecasting?

[@barrydrake473]

Is there a similar demonstration for clustering time series?

[@bxx-py6cd]

hi jesse,i have exsting data to train this model ,and if i want to add new data to train.what's the connection with exsting data.how will it begin train

[@islberfinkoparan7839]

Hello, thank you for this informative video, to clarify, at the "#update the model" part we add a learn_one at the end again, does this line improve our model trained on our train data with test data or am I misunderstanding? At this part, I understood as if we wish to add more data to the model, we can give it as test data so it can improve the model with that line of code, and when we use learn_one again at last, it is the test data, am I right on this ?

[@anil099kumar6]

What main main difference river package fit into feature at the but it take long time in scikit learn package fit at time training data please tell me what main difference scikit-learn and river package

[@selimbenfradj9864]

hello Jesse , I'am new in ML specially could you please answers those qustions: 1- Could river stream Metrics from datetime series collector tools such as Influxdb or Prometheus and feed it to a faillure prediction modul? 3-how to build a faillure prediction modul based on online ML ?

[@vaishnavpk22]

How to change this to an multi label text Classification by using river. Prediction like ['software','api','label1']

[@growwitharosh5052]

Hi Jesse, Thank you very much for your nice video on incremental learning. Is it possible to classify the facial emotion by Creme? I would request you please make a video on facial emotion recognition using Creme.

[@JCharisTech]

Glad it helped. Thanks for the suggestion Arman

[@abdelilahabdallah7133]

12:05

[@makoaewilson5869]

Hi Jesse thanks so much for this. I am new to machine learning and I am fascinated by you videos. Question: after the model has made one wrong prediction, what would be the approach to correcting it on that particular data field that it got wrong so that it gets it right next time? Thanks. Would I throw that specific data field back and train it again and make it learn again. So basically run learn one with the correction?

[@JCharisTech]

Thanks Makoae, about your question, in general (batch machine learning) the accuracy and performance of the ML model is usually dependent on the quality and quantity of the data you are using. Hence the more data the better the model ie if the data is of better quality. However for online ML, you can update the dataset like you said with the correct data and label and retrain. Hope this helps

[@jamalfinnley4135]

i know I am pretty off topic but does anybody know a good site to watch new series online?

[@taolu6067]

How can I use it in cv field?

[@JCharisTech]

Hello Tao lu, do yo mean Computer Vision?

[@mayanktripathi4u]

I am getting below error "ImportError: numpy.core.multiarray failed to import" when ```import river```. Also did tried lowering the numpy version to 1.18.1 ; 1.19.3; 1.19.5.. but none worked. Please help.

[@JCharisTech]

Hello Mayank, please can you try using it within a virtual env eg pipenv or virtualenv Hope this helps

[@MichaelWoodJ]

Very nice video! Just slow down and enunciate a bit more. Awesome job.

[@saurabhghosh7800]

Can we use transformer using river? urgent ,please reply

[@JCharisTech]

Have not yet tried online ML with transformers yet

[@anil099kumar6]

Please explan dask package