Author Archives: Advaith Veturi

Last year around the same time, I did my first interview ever – for a machine learning research position at a MedTech startup. Naturally, as one usually is, I was super nervous and did a lot of prep on my fundamental maths and programming/software development-related skills. But to my surprise, one of the first questionsContinue reading “AI and Medicine: 7 ways to bridge the divide”

A couple of weeks back, I met a colleague who works as a clinician and was interested to learn more about AI and its applications. He mentioned to me how he read many different articles about AI and how they would say, “AI has learned to do task X,” but he never understood what thatContinue reading “Machine Learning? How do machines Learn?”

When you Google the terms “Deep learning and Medicine” or “Machine learning and medicine”, most of the articles you see will probably be about its uses in radiology-based diagnosis, electronic healthcare records, basically the clinical data. Over the past couple of years, there has been increasing ML research on the more cellular level, focusing onContinue reading “Zooming into the cells – Genomics and Bioinformatics”

Generative modeling and GANs are popular in the media today because of their success in various tasks like generating deep fakes, digitally painting realistic portraits, converting images from one style to another, denoising images, etc. While these are really cool innovations, I think the translation of these ideas into the industry and our daily livesContinue reading “3 Ways GANs can Change Medicine”

Previously, I discussed in-depth the vanishing gradient problem in GANs and how it leads to unstable and ineffective learning. I then introduced the Wasserstein GAN, one of the popular solutions to this problem. While most research efforts into GANs have been in line with this goal of improving training stability, the game doesn’t end here. ThereContinue reading “Improving the GAN Part 2: Conditional and Controllable generation”

In my previous post, I gave an introduction to generative adversarial networks (GANs), discussed the basic training algorithm, and implemented a basic GAN for generating handwritten digits. While this basic model works well and has been shown to produce decent images, there are many difficulties that one can encounter during training, a popular one beingContinue reading “Improving the GAN”

You must be wondering who these random people in the above picture are. What if I told you they don’t exist and were created by a computer? I was pretty amazed when I saw this at an AI talk I attended a couple of years back, and it still amazes me to this day. TheseContinue reading “Introduction to Generative Adversarial Networks”