Studying LSTMs in R

Hello. Could you, please, recommend me any courses (both online and offline) or materials, which could be used for studying LSTM networks creation (immediately in R)? Level: Pre-Intermediate.

Thank you for your help.

Vignette in rnn package.

Hi @Maxim,
A couple of references come to mind:

Chris Olah's blog on Understanding LSTMs

Francois Chollet and J.J. Allaire's Book Deep learning with R in particular section 6.2-6.3.

1 Like

@jrmuirhead Thank you for your help! By the way, do you know anything about the web courses (or simply videos), where it is shown how to create an LSTM model in R step by step? Especially I'm interested in LSTM networks, designed for solving time series forecasting problems.

Hi @Maxim. I don't know of any web courses or videos off-hand as I've pretty much learned it by reading and working out examples. Section 6.3 in Deep Learning In R covers has a worked out example for using LSTM for time series data. What is particularly helpful is the data pre-processing step where they explain how to structure the time series data to feed into the neural network. Although they are in python, I've found other examples such as https://machinelearningmastery.com/time-series-prediction-lstm-recurrent-neural-networks-python-keras/ very useful to get an overall picture of the workflow.

1 Like

But this is only 1 section in a book, which includes only several dozens of pages. I need something more complex and fundamental.

Unfortunately, I don't know Python at the level, which gives an undestanding of complicated neural networks and how they function.

Hi Maxim,
Section 6.3 deals with implementing LSTMs in R ( with Keras) specifically. The first half in the book covers some of the fundamentals and basic theory for all neural networks but at a fairly high level in order to get you up and running quickly, but I suspect only a course would suit your needs.

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.