Ravit's Blog

Society values people who: are knowledgeable, and have well-thought-out ideas Top Tech CEOs are all hardworking and hence know a lot about their companies, which are like their babies. People who can defend their ideas and win debates are better negotiators and get their way. are tall and handsome it's true, but don't worry if you aren't; life is unfair. you can probably get quite far without these. are goal-oriented Dr. Larabee asks Akeelah "Do you have any goals?" are witty A friend might react positively if I say something witty. Maybe the idea behind wit is to laugh in the face of suffering. are relatable If you are like them, they want to be like you. They can learn from your actions. are humble, likable, and optimistic Humble people are humble because they think a lot about how much they don't know, not about how much they know. Likable people tend to attract people. Optimistic people give people around them a reason to live. are trustworthy Trustworthy ...

One of my professors is a theoretical computer scientist. Looking at his past publications, which all are very mathematical in nature, I started to doubt whether I should really go into TinyML, a field that is very high-level for a computer science field. I also started to think about different skills that are required to contribute to these fields; I guess theoretical computer science requires a mathematically-oriented mind, while the majority of ML requires an engineering-oriented mind. I also was thinking about the nature of software engineering. It is exciting to come up with a new idea or improvement that can do good in the world--improve the performance or quality of a tool or introduce a new digital product. For example, it is exciting to propose new OS philosophies, research ways to make networks and distributed systems more efficient, and develop new hardware accelerators. But, it seems to me that at the heart of most of today's software is the repetitive implementation of...

There must be some game that two neural networks can play to solve the problem of NAS. GANs solve the problem of image generation through a very simple zero-sum game between two networks, where both compete to capture or evade the other.  Secondly, there must be a continuous way to deform a simple neural network into an immensely complicated one, like deforming a cow into a sphere. https://www.youtube.com/watch?v=1uYkm8mnXss