Multivariate Time Series Analysis with an RNN - Training

This is a simple example workflow for multivariant time series analysis using an LSTM based recurrent neural network and implemented via the KNIME Deep Learning - Keras Integration. It is based on the bike demand predition dataset from Kaggle and trains a model to predict the demand in the next hour based on the demand and the other features in the last 10 hours.


This is a companion discussion topic for the original entry at https://kni.me/w/B45XEOAuWeQBzO9b

Hi,

I downloaded this workflow to give it a try since am working on a similar project.However,to my dismay I ran some node successfully but I can’t progress past keras network learner.Nothing seems to be happening here for an entire night.

Below is a screen shot of it and see attached is the console output.

I am working on windows 10.I have also installed all the extensions and I tested another Keras workflow which worked fine(see below)



image

See also attached ananconda dependecies.Anaconda dependencies.txt (6.8 KB) Console output.txt (1.2 KB)

Please let me know what the problem could be the problem

I haven’t finished running this workflow yet myself, but I wanted to note that you can check the progress of the Keras Network Learner by right-clicking on it while it executes and selecting “View: Learning Monitor”. On my machine (no GPU) it’s taking about 40 seconds per epoch, so this should still finish in under an hour.

Can you check this post to see if the suggestions help you? KNIME 4.1: Keras error ( Selected Keras backend 'Keras(Tensorflow)' is not available anymore - #14 by MarcelW

Hello,

I tried recreating the workflow with the knime versions and extensions originally used to develop the workflow but there is no progress.

See below and attached log files and anaconda environment created and all packages associated with it.

There is a big problem somewhere. Does this mean the problem only works on certain OS?
Please let me know what the problem could be.Conda trouble shooting.txt (12.4 KB) Console output.txt (1.2 KB)

at the time of execution it shows they follow it in the console and it takes a long time in 12%

WARN Keras Network Learner 0:148 The number of rows of the input training data table (13921) is not a multiple of the selected training batch size (32). Thus, the last batch of each epoch will continue at the beginning of the training data table after reaching its end. You can avoid that by adjusting the number of rows of the table or the batch size if desired.

WARN Keras Network Learner 0:148 The number of rows of the input validation data table (1736) is not a multiple of the selected validation batch size (32). Thus, the last batch of each validation phase will continue at the beginning of the validation data table after reaching its end. You can avoid that by adjusting the number of rows of the table or the batch size if desired.

WARN Keras Network Learner 0:148 C:\Users\PC\anaconda3\envs\py3_knime_dl_1\lib\site-packages\keras\engine\saving.py:292: UserWarning: No training configuration found in save file: the model was not compiled. Compile it manually.

keras

HELLO @Kathrin
excellent exercise, if I want to use this workflow for academic purposes, what is the correct way to cite it or refer to the author.

I quote @Kathrin?

Hi @jhonguzman88,

thank you for you feedback and for quoting me in your exercise. What about just adding my full name and LinkedIn account?

Regarding your question from above. The first two warnings don’t cause any problems. I’ve never seen the third warning though. How long does it roughly take at 12%? Are you in general able to execute the workflow?

Cheers
Kathrin

Hi @Kathrin

If you please full name and link of linkedín

Regarding the execution of the route, I have already finished and it took 1 hour 40 minutes

Hi @jhonguzman88,

my full name is Kathrin Melcher and this is my LinkedIn Account: https://www.linkedin.com/in/kathrin-melcher-b44542155/

Cheers
Kathrin