Video

Thank you kindly, Ajay :)

Calc textbook in the works, which will cover calc 1+2 (differentiation and integration), and a bit of calc 3 (series and diffEQ, though probably not in as much depth). It's too early for me to give a release date, but hopefully early 2025.

@@mikexcohen1do you think calc3 is unnecessary for AI? I didn’t see much calc3 topics in your deep learning course on udemy. I thought stuff like Jacobian has great application, or is it unnecessary?

Thank you, kind internet stranger.

Awesome :)

Good luck on your exams!

Glad I could help!

Nope, but the book that this course is based on has been translated into Python. See my github repo for the ANTS book for links. That said, I do have a signal-processing course (non-neuroscience-related) that is in both MATLAB and Python.

Yeah, sequential arithmetic is a good way to give yourself a headache :P PEMDAS always takes precedence, and then you implement left-to-right in a sequence of same operations. Best practice is to use parentheses for grouping when there is a risk of ambiguity.

@@mikexcohen1 gotcha, yeah it’s definitely a head ache. Thanks for the tips!

This course focuses on foundations of deep learning and AI. It is true that AI is actively developing, but the developments are mostly in scale, architectural tweaks, and applications. In terms of the foundational math and operations, not much has changed in 25 years.

What about any big project's?

Yes, you can do permutation testing on ANOVAs. The mechanism is the same as with t-tests: Randomly re-assign each data point to be in each factor/level, rerun the ANOVA, and store the F values to build up a distribution of empirical H0 F values. I don't have any videos on this method, though.

I appreciate that!

I do cover some math/theory in this course, but it is much more focused on practical/hands-on/coding. You don't need any particular background to take this course except for basic coding (MATLAB or Python). But to be honest, if you're looking for a course on rigorous maths of signal processing, then this might not be the best fit for you.

Oh great! The Korean translation was just recently released. I hope you enjoy it!

Thank you, kind internet stranger :)

Thanks Jason! I hope you find the book enjoyable and useful :)

Thank you kindly, Drew.

Awesome :)

Hi er. I don't have a video on the Stockwell transform per se, but it is really similar to Morlet wavelet convolution, and I have lots of videos on that. If you understand Morlet wavelets, it's just a tiny step from there to Stockwell. Hope that helps!

Hi Marco. I don't make my pdf slides available, but you get lifetime access to the entire course.

@@mikexcohen1 I trust your expertise so I bought it

Thank you kindly, fpejavier!

Glad you like them!

Glad it was helpful!

Glad you enjoyed it!

Hmm, interesting question. My intuition is that because the ERP is already so smoothed by averaging many trials, that TKEO wouldn't be very insightful. The TKEO filter is designed to amplify broadband energy, and most of the broadband energy is filtered out of the ERP.

I politely disagree, but I appreciate you taking the time to express your view! (Also, for some context: this is one optional video from a 60-hour video course on deep learning, and I don't think a claim like "not a good model" should be made without rigorous justification.)

No video! This is an audiobook :D YT keeps the cover image up the whole time.

Hi Victor. Zero-padding is valid as long as you understand the correct interpretation: The values in between the "true" frequencies are sinc-interpolated values. You can trust them as much as you can trust any other interpolated signal. If the resolution is already fairly good, then interpolating a bit higher is fine. In extreme cases, it would be good to exert caution. Anyway, zero-padding is necessary for some applications like convolution, so it's used even if the interpolated points are not interpreted.

Dear Mike, many thanks for your response. It's clearer now. Best, Victor

The description of this video has the direct link to the zip file with all data and code for this module.

@@mikexcohen1 thanks sir

Thank you kindly, oty. It's an ongoing process...

Yes, it's best to be open and honest about how you used AI in all aspects of the research, not only in the final writing stages. People are becoming more comfortable with the idea that AI is a powerful tool when used appropriately, and open declarations of how it was used is part of that.

@@mikexcohen1 Thanks Mike

It's also a great ice-breaker at socially awkward parties :P

Yes, you can perform a time-frequency analysis on resting-state data. The stationarity assumption is based on the FWHM of the wavelet, not the length of the data. The main difficulty is how you would analyze and interpret the data... a lot happens in the brain in 10 minutes! Of course you will see variations in the spectrum, so it would be good to have some hypotheses about what will change and how that might differ between meditation and WM.

Thank you very much for quick response.. Yeah, we have a specific hypothesis. I was thinking from stationarity perspective, if it's right to do..

They overlap a bit. The "theory" book is longer and more focused on concepts, theory, and proofs (though still with an eye on implementations and applications); while the O'Reilly book is more focused on implementations and applications (thought still with an eye on proofs). So the question is whether you are looking for a deeper and more thorough dive into linear algebra, or want to get up to speed quickly with applications.

@@mikexcohen1 Thank you for your response. I am enrolling in a MSc Computer Science. I thought Linear Algebra is very interesting and I am keen to learn more. I also am learning cpp since uni requires a bit of coding skill though they prefer Python. It seems that maybe your other book is more suitable for me since I don't have an image of where or what I will use LA for apart from learning for the sake of learning it. However, if I pick up this book, the implementations part may come in handy if there is a need for it. I also saw your course on Udemy, "Complete linear algebra: theory and implementation in code". Would you recommend going through this course instead of self-studying those books? Thank you for your time.

Both books overlap enough that you'd benefit from either but probably don't need both. Actually, if you're doing CS, then the O'Reilly book might be better b/c of the stronger focus on the coding/implementation. As for the course: That's also similar but not redundant to the books. Some people prefer to learn from books, others from videos, and I try to cater to both audiences.

Thank you, kind internet stranger.