Installing RStudio/Keras on Windows

There are various issues installing RStudio/keras on Windows. Below is a solution provided by Edward Yu (@edwardmjyu):

The issue in more detail and its solution is found here: https://github.com/rstudio/keras/issues/626

Below are the steps for the solution: Only thing that did work (executed in the following order):

1. installation of R 3.5.2 (independent from directory (standard or any other))
2. installation of RStudio (independent from directory)
3. RStudio -> Tools -> Global Options -> Packages -> Disable both “Use secure download method for HTTP” and “Use Internet Explorer library/proxy for HTTP”
4. installation of Miniconda3 (-> has to be the standard directory!) Version 4.5.11 (it did not work with the newest version)
5. Use RStudio ->

   install.packages("tensorflow")
   install.packages("keras")
   library("keras")
   install_keras() 
   

   One caveat in the installation process is the version of Miniconda, there may be potential issues if the newest version of miniconda is installed.

TensorFlow Lite on iOS

Following tips are from Bryan Kevan @bryanmkevan:

There are a lot of steps missing from the tensorflow lite readme that we talked about in class on Tuesday. I’m giving a heads up just in case others try to replicate those steps and come across these issues. Missing steps are:

Xcode package (no link to this on the page): https://github.com/skafos/ImageClassifier, then run pod install on the directory of the podfile

To run the app, we have to set up a Skafos account and follow the startup guide here: https://docs.metismachine.io/docs but note that the quickstart guide link specified in the xcode project readme is wrong.

FAQs

• Kristen’s question. Intermediate results during TensorFlow training. Checkpoints in TensorFlow: https://www.tensorflow.org/guide/checkpoints, callbacks in Keras: https://keras.io/callbacks/, or callback_model_checkpoint function in RStudio/Keras.

• Stephen’s question. Metrics used in word2vec: PMI (pointwise mutual information), shifted PMI, … See http://blog.aylien.com/overview-word-embeddings-history-word2vec-cbow-glove/

Announcement

• Today’s office hourse: 2:30p-3p.

Checklist on your resume/cv

A data scientist is someone who is better at statistics than any software engineer and better at software engineering than any statistician.

• Git/GitHub (give your GitHub handle)
• Tidyverse
• Data visualization (ggplot2, shiny)
• SQL databases
• Rcpp, parallel computing
• HPC (if you use Hoffman2)
• Cloud computing (GCP, AWS?, Azure?)
• Docker
• Deep learing with Keras+TensorFlow+GPU (PyTorch and Caffe are friendlier for research)
• Frontend development (shiny, smart phone app)

• Apache Hadoop + Spark

• Make your GitHub repo biostat-m280-2019-winter public (after final week) and show your work to back your resume.

• Use these tools in your daily work: use Git/GitHub for all your homework and research projects, write weekly research report using RMarkdown, give presentation using ggplot2 and Shiny, write blog/tutorial, …

• Stop your own GCP instances and release un-used static IPs to avoid charges. Ask me for $50 credit if you use up your $300 free credit.

What’s not covered

• Machine/statistical learning methods. Familiar with methods in Elements of Statistical Learning and software, e.g., scikit-learn.

• Algorithms. Spring quarter’s M280 will cover numerical linear algebra and numerical optimization algorithms.

• Public health applications.

• Be open to languages. Python is a more generic programming language and widely adopted in data science. JavaScript is dominant in web applications. Scala is popular for implementing distributed programs. Julia is attractive for high performance scientific computing.

Today

• Course evaluation: http://my.ucla.edu. Do it now please!

• GAN example.

• Apache Hadoop, Spark, Sparklyr.