How to make RStudio work with anaconda/miniconda in macOS Sierra

I tried installing anaconda and r-essentials on my macbook.  The install went fine, but when I downloaded RStudio and tried to start it I got an error saying that it was unable to find the R binary.  So how do we fix this?

  1. Make sure you have R installed via anaconda or miniconda. Do this by typing “which R” without quotes into the terminal.  It will give you an output like “/Users/yourAccountName/miniconda3/bin/R”
  2. Use this to tell R where to start up from.  Edit your .bash_profile with your favorite text editor to set the location of R within the anaconda install directory. I used vi .bash_profile and edited my host file to have
  3. Next type open -a RStudio into the terminal.  Open and close terminal.
  4. Rstudio should then open from the dock.

Building a ML box with Ubuntu for fun and profit & Hello to my blog

Welcome to the inaugural post   

At I’m going to talk about biology, CS, math, life.  Pretty much anything, but the blog will largely focus on biology and CS.  

I’ve been working on the EdX course series PH525x on Biomedical Data Science, and there will be a few blog posts on it forthcoming.  I have also just built a new Ubuntu ML box, and that is where we will start today.

Building a Machine Learning box with Ubuntu 16.04 for fun and profit

Ubuntu Linux is great.  It is used worldwide, it’s free, and the unity/gnome desktop environment has matured to the point at which it is actually usable for daily desktop use.  Gone are the days of manually recompiling a kernel module to load nvidia drivers, all you have to do now is check “Nvidia Drivers” in the control panel, and it works 99% of the time.  It’s also fast, having decent GPU performance as Phoronix tested…


“These tests largely just continue to prove that the performance and features/functionality of the NVIDIA Linux driver remains very close to that of the NVIDIA Windows driver; it’s why many game studios/porters working on Linux games continue recommend using NVIDIA proprietary graphics.”

I wanted to build a decent Ubuntu computer with a discreet GPU for less than $500.  I decided to go with an off-lease dell t3500 workstation.  This thing was cheap, costing only $200 including shipping on ebay.  It has a xeon X5650, which is adequate for what I’m doing which is mostly gpu-based.  It also includes 12gb ram, and although 16 or 32gb would be better, 12 is enough.    

After selecting the system, I decided to go with a Gigabyte GTX 1060 gpu.  This card has 6gb of ram, and a decent GPU processor with fans that will idle at zero rpm, so things will be quiet..which is important.  It will also work with my Monoprice 4k monitor at 60hz, as it only takes 60hz input from displayport.  After the rebate the GTX 1060 GPU came to $230.43.  

For the CPU I chose a Crucial MX300 275 GB SSD for $69.  It is a sata III connection on the drive but only a sata II 3.0GBPS aka 300 megabyte/sec max on the motherboard, but I don’t think that will make much difference on this system in practice.  I also needed a 6-pin to 4-pin adapter to get things working.  Total cost was 502.93, not too bad.  

After opening the case I then had to take the old GPU and harddrive out, which was not too difficult.  I then added the GPU and SSD.  Then I installed ubuntu 16.04.1LTS on the SSD.  After that just a quick apt-get update and apt-get upgrade, install gpu drivers from settings, and I was set.  


100% functional Ubuntu Linux box pushing a 4k display at 60hz for just over $500.  It works just great, this is about the best value you can find.  Off-lease dell/lenovo/hp workstations provide a great starting point for gaming or machine learning/GPU-compute builds. 

In the next post I’lll talk about installing CUDA and tensor flow.