Elvin Aghammadzada

An AI Researcher, 3x Kaggler | Creative Technologist | AI Product

[Elvin Aghammadzada @aghammadzada]

A podcast about now and the future of AI.

https://t.co/HGWAAKlH3x

[Elvin Aghammadzada @aghammadzada]

@calcsam book

[Elvin Aghammadzada @aghammadzada]

got featured in this article

https://t.co/uQ3rDwQjsk via @featureddotcom

[Elvin Aghammadzada @aghammadzada]

American Truly Open Models (ATOM) Project https://t.co/BH37gpLK7r

[Elvin Aghammadzada @aghammadzada]

@mikemajzoub @Adobe reach-out, please let me know. :)
Sincerely, Elvin

[Elvin Aghammadzada @aghammadzada]

@mikemajzoub @Adobe doing things like building multi-modal agents and RAG pipelines for design and working closely with designers and creative directors.

I'm really inspired by the work Adobe (and maybe your team specifically) is doing in this space. If you think this sounds like a sincere...:)

[Elvin Aghammadzada @aghammadzada]

@ugurkaner Hi Ugur. Please let me know if we could chat more about this? Can't dm you. Thanks :)

[Elvin Aghammadzada @aghammadzada]

2,000,000,000,000
is approximate number of galaxies in the observable universe. Even if you travel at light speed, 95% of the universe is already unreachable. Gone forever...

[Elvin Aghammadzada @aghammadzada]

@bindureddy What's the idea?

[Elvin Aghammadzada @aghammadzada]

@EugeneVinitsky 😄

[Elvin Aghammadzada @aghammadzada]

@seaamiacsr Is it just me, or are you and depression on a first-name basis these days?). thnks

[Elvin Aghammadzada @aghammadzada]

@seaamiacsr to build a billion-dollar AI-based Chrome extension

[Elvin Aghammadzada @aghammadzada]

Explain like I am 5:
While log regression can be affected by data distribution difference among features, non-param models like decision trees are not.
This means that the values of different features may vary in terms of their range, spread, central tendency, etc

[Elvin Aghammadzada @aghammadzada]

I am seeing the word non-parametric everywhere in ML papers. Assuming you do as well, here is what it is in simpler terms.

Non-parametric models are models which do not assume a specific functional form or distribution for the data, providing greater flexibility and adaptability

[Elvin Aghammadzada @aghammadzada]

Non-parametric methods often rely on techniques such as nearest neighbors, decision trees, support vector machines with flexible kernels. These methods can handle a wider range of data distributions and are capable of capturing more complex patterns and relationships in the data.

[Elvin Aghammadzada @aghammadzada]

They offer more flexibility and adaptability by not assuming a specific functional form or distribution. Instead, they aim to learn directly from the data without imposing strong assumptions. They can be more expressive and better suited for complex or unknown data distributions.

[Elvin Aghammadzada @aghammadzada]

Couldn't find any big tech/chip companies building something towards carbon tubes instead of silicon transistors. Is carbon computing a thing in academia only?! (yet)

[Elvin Aghammadzada @aghammadzada]

@adilaliyev this was a comparison between today's most advanced chips vs. the human brain to confirm that we still have a long way to go to catch up with the human brain level (maybe a few more years 😅). I feel like there will be a revolution in the silicon chip industry soon...

[Elvin Aghammadzada @aghammadzada]

Every transistor roughly talks to about 10 transistors
Every neuron, on the other hand, roughly talks to about 10,000 other neurons