Deep Learning with Keras and TensorFlow in R Workflow
1/27/20—1/28/20
9:00 AM-5:00 PM
2 Day Workshop
Brad Boehmke
Data Scientist
84.51°
Overview
This two-day workshop introduces the essential concepts of building deep learning models with TensorFlow and Keras via R. First, we’ll establish a mental model of where deep learning fits in the spectrum of machine learning, highlight its benefits and limitations, and discuss how the TensorFlow - Keras - R toolchain work together. We'll then build an understanding of deep learning through first principles and practical applications covering a variety of tasks such as computer vision, natural language processing, anomaly detection, and more. Throughout the workshop you will gain an intuitive understanding of the architectures and engines that make up deep learning models, apply a variety of deep learning algorithms (i.e. MLPs, CNNs, RNNs, LSTMs, autoencoders), understand when and how to tune the various hyperparameters, and be able to interpret model results. Leaving this workshop, you should have a firm grasp of deep learning and be able to implement a systematic approach for producing high quality modeling results.
Is this workshop for you? If you answer "yes" to these three questions, then this workshop is likely a good fit:
Are you relatively new to the field of deep learning and neural networks but eager to learn? Or maybe you have applied a basic feedforward neural network but aren't familiar with the other deep learning frameworks?
Are you an experienced R user comfortable with the tidyverse, creating functions, and applying control (i.e. if, ifelse) and iteration (i.e. for, while) statements?
Are you familiar with the machine learning process such as data splitting, feature engineering, resampling procedures (i.e. k-fold cross validation), hyperparameter tuning, and model validation? This workshop will provide some review of these topics but coming in with some exposure will help you stay focused on the deep learning details rather than the general modeling procedure details.
Prep work
This is an applied workshop meaning we will be running lots of code so be sure to bring your !
We will be using an RStudio server on AWS. This means you do not need to worry about downloading certain packages or datasets ahead of time; everything will be prepared for you. Bonus - we get to play with GPUs !
All the content is open source. You can find the workshop repo at rstd.io/conf20-dl
Please review the " Prework " section of the README so you are not surprised by the assumptions I make regarding prerequisite knowledge. Also, although you will be using an RStudio server with all the data and packages installed, I provide source code to install all prerequisites so that you can reproduce the analyses post-workshop on your own PC.
Please review the "Setting up RStudio Cloud environment" instructions. I would highly suggest that you work through the instructions and actually log on prior to Monday morning. There are a few kinks still getting worked out so please don't run any notebooks but you should have no problems completing all the steps in the setup instructions. If you do have problems please let me know.
This workshop is notebook focused, meaning we will be working through many R notebooks. If you are not familiar with R notebooks, take some time to review this: 3.2 Notebook | R Markdown: The Definitive Guide .
Feel free to poke around at some of the notebook material but understand that I am still putting some finishing touches on a few areas so you may notice slight changes between now and Monday. Consequently, if you want to clone/fork this repo I would either wait until we kick off the workshop or just plan on doing a git pull to update your content. If you don't know what clone/fork/pull means, don't worry .
Hey @tdbuskirk, you can find the workshop content at http://rstd.io/conf20-dl. You will be receiving an email this evening and over the next couple days regarding how to prepare.
Yeah, I just noticed that as well. This is new. Let me check with the RStudio engineers. I know they have been doing some server swaps recently so maybe there's a temp outage.
I'm running through 01-requirements.Rmd and am able to get all the data downloaded and moved over to the correct folders for all but Dogs vs Cats. The data download fine and are moved into the main folder, but then they don't seem to be moving into the subfolders correctly.
I've attached a couple of screenshots to show what my output ends up looking like. You can see I have all the images in the "full" folder, but they are not moved into the subfolders. The subfolder structure is there, but all the actual folders are empty.
Are you doing this on your own laptop or on the RStudio workshop servers? If you are doing it on the servers, all the data already exists (albeit in a different directory) and you won't have to worry about doing the downloads or package installs.
The 01-Requirements process is mainly so that you can replicate the notebooks outside of the workshop. If this is happening on your own laptop then it might be a permissions issue, which I can help you with at the workshop.
Yeah, I'm pretty sure I know what it is but we can confirm at the workshop. I think it is a permissions issue when trying to unzip the train.zip file within the overall cats-vs-dogs.zip file.
And so you are putting in your name and email and that's the error it provides? Sorry but just want to confirm. If this is in fact the issue can you send me the exact name and email you are using to sign in so engineering can investigate? You can email that info to me for privacy reasons.
Just an FYI that RStudio will need to do a tear down of the servers over the weekend. Consequently, you may get an error when trying to log in that states "Sorry an instance is not available at this time". But don't fret, we had several folks successfully log in using the instructions so we are confident it will go smoothly Monday morning .
!
!
.
,
.