One of the most insightful blogs I’ve ever read was Tim Vieira’s post on “Counterfactual Reasoning and Learning from Logged Data”; I highly recommend it to anyone interested in reinforcement learning and policy optimization techniques. That said, with this post here, I hope to distill some wisdom that analyst and app developers can use to improve their platform’s engagement.
For those unfamiliar with the concept of counterfactuals, you can think of them as data that would have arisen under an alternative scenario i.
“Short Short Courses” are questions meant to prompt the reader to think and reflect. In my time as a tech lead I’ve found that asking questions is far more important than laying out answers. Imagine you’re in a meeting with your team and someone asks these questions of you, would you have a good answer? And if you don’t have good answers, now is your chance to start asking these questions.
In the interests of learning and developing a better sense on how to to use Bayes Rule we’re going to do two things in this post.
We’re going use Bayes Rules to come up with an analytical solution to a probability problem (like you’d see in an interview setting). And
We’ll simulate some data to see if we can approximate an analytical solution (to help develop our programming chops). The Scenario Imagine you’re a patient who has just received a positive test result for HIV.
I’ve never found a blog post that truly made me understand how to derive the normal equation. I think a reason for this is a lot of authors have their own perspective on the problem, leading to a mental image I can’t quite grasp onto. But having grappled with linear algebra for a bit, I think I’m ready to try my hand at writing a series of posts that helps someone like my former self understand such an equation.
Today we’re going to configure Jupyter notebooks so we can connect to our Nano over https. Note, this same logic could easily be applied to a server in the cloud.
Prereqs/Assumptions You will be doing this for Python3
You already have your default python3 configured for virtualenvs
You have created a virtualenv that we’ll install Jupyter & the Ipython kernel on.
We’re doing this setup while we are sshed into the Nano (ssh <jetson name>@<jetson_ip>)
Background If you’ve never heard of Nvidia’s Jetson Nano you can think of it as a Raspberry Pi with a built-in GPU; their developer community page has some pretty interesting projects if you’d like to learn more.
Now, you’re only going to get some milage out of the rest of this post if you’ve already setup your Jetson Nano hardware. I set mine up to use the barrel jack for a power source, added a wifi card to the board, and connected it to the Raspberry Pi camera (to collect training data and use for real-time inference).
Hello,
My name is Andrew Bibian and welcome to my site! Also, thanks for coming. 😃
I created this site to help myself learn how to write better and explore concepts in data science and deep learning. That was around 2019 and after I started writing posts, it petered out fairly quickly. But it’s now 2022 and I’m taking a different approach to how I document my various life learnings.