Open xgxg1314 opened 1 year ago
Hello, Xgxg1314,
Thank you so much for your kind words. I am glad you found them useful.
With deepONets, you approximate the evaluated operator (i.e., numbers). The output of your operator say (i.e., $G(f(x))=h(x)$ ) is a function. Since $h(x)$ is a function, when you evaluate it at $x=y$, you get a set of numbers, say $h(y)=z$. Those numbers $z$ are the ones you approximate with your DeepOnet. So they are kinda "numerical approximators."
Regarding your specific problem, I am not sure if you could solve it with a Physics-Informed DeepONet (i.e., it would be tricky to include the integral in the loss function related to your PDE). But, you should be able to solve it with a traditional DeepONet. You may need some input-output functions $f(x)-h(x)$ to compute your loss and train your operator. Please review: https://github.com/jdtoscano94/Learning-Python-Physics-Informed-Machine-Learning-PINNs-DeepONets/blob/main/DeepONets/7_Antiderivative.ipynb
I am not sure if that answers your question. If not, please let me know. Also, I would suggest you review this research group: https://www.brown.edu/research/projects/crunch/home
They are the pioneers of DeepONets and have many studies in that field. So maybe, they have faced a similar problem as yours in the past.
Thank you so much for your comment and for watching my videos.
I hope you have a nice day.
Hi, Juan Diego Toscano. You can call me Xiong. Thank you so much for your nice answers and useful advice. I really appreciate it .It helps a lot! I will try to solve my problem with DeepOnet. I think it works now. Looking forword to your next video on Youtube. Have a nice day.
Thank you so much for your kind words @xgxg1314. I am glad you solve your problem. Good luck with your next project!
Sincerely,
Juan Diego Toscano
Hi, Juan Diego Toscano. I turn to you for help agian!I want to know how to get access to videos recording about "Machine Learning + X seminars 2022" in Brown University.
September 1 Recording: A comprehensive study of non-adaptive and residual-based adaptive sampling for physics-informed neural network by Chenxi Wu, Brown University
Recently , i am really interested in this paper titled "A comprehensive study of non-adaptive and residual-based adaptive sampling for physics-informed neural network", but i have no access to video recoding about it.
Could you please help me to find the video? Thank you in advance,anyway! Regards, Fei Xiong
Hi Xiong, I hope you are doing well,
Sure, actually there is a youtube channel where you can find all those recordings (from September 9). https://www.youtube.com/channel/UC2ZZB80udkRvWQ4N3a8DOKQ
Oh, I see that video is not available on youtube yet. I will try to ask them to upload all the previous videos to the youtube channel.
In the meantime, I can share the video you need. Can you give me your email, please?
Sincerely,
Diego Toscano
Hi Toscano,
I am sorry for the late response.Sure, i really need the video, you do me a great favor today! Thank you so so so much for sharing these nice videos, it is very useful for many researches working on scientific machine learning.
My gmail address is xiongxiongxgxg@gmail.com
Have a nice day,
Fei Xiong
Hi Fei Xiong.
I have just sent the video to your email. Please let me know if you received it. Thank you for your nice words, and I am glad you found them useful.
Have a nice day too.
Sincerely,
Diego Toscano
Hi Toscano, I have received the video you shared, but I have no access to the video.It requires access authority.I have tried to apply for access.Just wait for permission. Thank you so much for your sharing!
Sincerely, Fei Xiong
Can you check again, please?
Yes , it works now.I can see the video. Thanks a lot,Diego Toscano!
A great repository!Thank you very much for your open source, i also really like your videos about this on your Youtube! It is so awesome!
I have a question about deeponet.
$$ \frac{d y}{d x}+y(x)=\int_0^x e^{t-x} y(t) d t $$
i want to use DeepOnet to solve this integro-differential equations, but i have no idea how to implement it. The main problem is that i don't want to approximate integral op- erators numerically, i just hope to make it with DeepOnet
Is it possible to solve it using DeepOnet only?
Best Regrads!