Data Science - Part XVI - Fourier Analysis

[@LucienHughes]

Hi there, great video. This may be a stupid question but I'm just wondering what the advantage of using neural networks to predict the subseries rather than just extrapolating forward the results of the FFT are? Surely a neural network with no drivers other than a time series can never be more accurate than the extrapolated recomposed waveform?

[@o0prince]

This is a very well put together lecture. I like to explain to people that fourier transform is just looking at waves in a different dimension kind of like drawings on butter paper put together

[@dimar4150]

great presentation, one suggestion. at 41:00 you have a graph with unlabeled x axis unlabeled y axis and no legend to indicate differentiation of colors. I was puzzled looking at this graph that has only a title.

[@zhubarb]

first of all, thank you for taking the time to prepare the slides. in my humble opinion, the first 25 mins can be digested in 5 minutes from the wikipedia article. the added value is in the practical example. I wish there was more emphasis on this aspect.

[@zhubarb]

+zhubarb And I just realised the practical application (word by word) is indeed from this 2011-dated R blog entry: www.inside-r.org/howto/time-series-analysis-and-order-prediction-r

[@DerekKaneDataScience]

+zhubarb Thank you for some good feedback. Much appreciated. I completely understand your perspective and I think you have some valid points. What would you like to see differently? More applications of the technique from different R packages? It is surprising very difficult to find information related to practical applications with the FFT in R and Prof. Andras Pataricza has the best example on the subject which I used for the lecture. I am still awestruck by his brilliance and I have leveraged this approach to FFT in some time series applications in R in my professional work (which I might add to the revised lecture). Some people aren't even aware of the basics for the application and this was the target audience for this lecture. Hopefully, this provided some value for you with your work, even if in a minor way. Good luck and thanks again for watching.

[@xinyuliu2920]

Hi Derek, it is a very nice presentation! I am wondering if you could please send me the R code that I can play with a little? Thank you for your help!

[@MrInexistent]

I wonder what was the test data since we do not know the next quarter Fourier series we cannot give input at test time. If you used Lstm Instead of Mlp, I would understand but I don't get it.

[@flame-on-3]

You Rock Dude, I bet you play a mean bassoon.

[@newmohak]

Excellent work and very helpful.

[@fun_vandanrevanur]

Hey Derek, a comprehensive video on fft for time series. Just wanted to know if you have a github repo for this project? If not could you share the source code, privately?

[@anitachang4633]

Just wondering if anyone has managed to find the R code for this lecture?

[@anitachang4633]

If anyone is still looking for the script: github.com/DerekKane/Data_Science/blob/master/Fourier%20Analysis/Fourier%20Analysis%20Tutorial.R

[@ahmedaghmadi4626]

Hello Derek, it is a very good presentation, I really appreciate your work and I would like to ask you if you have some Matlab code that can I work with it please? thank you in advance

[@WahranRai]

Open link from berkmeister or google time series analysis...you will find the original paper of the professor and his team and inside the code in R (similar to matlab easy to understand)

[@MrSteniorod]

Derek. The video is very nice but you need to post the link from the code and from the dataset.

[@DerekKaneDataScience]

Jose Rodrigues Thanks for the feedback Jose. I am planning on either setting up a GitHub account for those who want to play with the data/code or something within SlideShare. In the meantime, please send me a private message and I can get you the materials. Thanks for watching and enjoy.

[@MrSteniorod]

That would be good. Just do not forget to put a video with the instructions on how to download it.

[@zhubarb]

+Jose Rodrigues www.inside-r.org/howto/time-series-analysis-and-order-prediction-r

[@陳中彥]

8:30 Heisenberg from Breaking Bad lol