velocity is not a constant number

[285639863]

I saw that the velocity is a constant 1 in your code. I want to use a more complex velocity model to replace this constant velocity model for training, such as the marmousi velocity model, but I don't know how to set it up because this velocity requires the use of a tensor related to the x and y coordinates of space.Do you have some ideas about this question? Best wishes!

[R0S3NB3RG]

Hi, Thanks for your email. However, I’m on my way these days. I will give my suggestion as soon as possible. Cheers.

发自我的iPhone

------------------ Original ------------------ From: 285639863 @.> Date: Mon,Dec 18,2023 7:45 AM To: R0S3NB3RG/PINNs-for-wave-equation @.> Cc: Subscribed @.***> Subject: Re: [R0S3NB3RG/PINNs-for-wave-equation] velocity is not a constant number(Issue #1)

I saw that the velocity is a constant 1 in your code. I want to use a more complex velocity model to replace this constant velocity model for training, such as the marmousi velocity model, but I don't know how to set it up because this velocity requires the use of a tensor related to the x and y coordinates of space.Do you have some ideas about this question? Best wishes!

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: @.***>

[R0S3NB3RG]

Hi, I think you can write a function for your velocity like velocity(x,t), which will return the the velocity at the specific point. When you doing the loss function part, use this return value to construct the relationship between your inputs and outputs. And my friend provide another creative way, which is to use the Lagrangian L(q,q_dot) which is a function depends on general position and general velocity and use Euler-Lagrangian Equation and least action to construct your loss function. The main challenge for this method is how to get the Lagrangian based on your PDE. (In your writing style, it seems that you are not a English speaker. If you are a Chinese, we can discuss it on Wechat if you like. If not, just forget about it) Cheers!

Rosen @.***

 

------------------ 原始邮件 ------------------ 发件人: "R0S3NB3RG/PINNs-for-wave-equation" @.>; 发送时间: 2023年12月18日(星期一) 下午3:45 @.>; @.***>; 主题: [R0S3NB3RG/PINNs-for-wave-equation] velocity is not a constant number (Issue #1)

I saw that the velocity is a constant 1 in your code. I want to use a more complex velocity model to replace this constant velocity model for training, such as the marmousi velocity model, but I don't know how to set it up because this velocity requires the use of a tensor related to the x and y coordinates of space.Do you have some ideas about this question? Best wishes!

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: @.***>

[285639863]

Thanks a lot！I have solved it using your method with a velocity function!

dreamland @.***

 

------------------ 原始邮件 ------------------ 发件人: "R0S3NB3RG/PINNs-for-wave-equation" @.>; 发送时间: 2023年12月20日(星期三) 中午12:00 @.>; @.**@.>; 主题: Re: [R0S3NB3RG/PINNs-for-wave-equation] velocity is not a constant number (Issue #1)

Hi, I think you can write a function for your velocity like velocity(x,t), which will return the the velocity at the specific point. When you doing the loss function part, use this return value to construct the relationship between your inputs and outputs. And my friend provide another creative way, which is to use the Lagrangian L(q,q_dot) which is a function depends on general position and general velocity and use Euler-Lagrangian Equation and least action to construct your loss function. The main challenge for this method is how to get the Lagrangian based on your PDE. (In your writing style, it seems that you are not a English speaker. If you are a Chinese, we can discuss it on Wechat if you like. If not, just forget about it) Cheers!

Rosen @.***

 

------------------ 原始邮件 ------------------ 发件人: "R0S3NB3RG/PINNs-for-wave-equation" @.>; 发送时间: 2023年12月18日(星期一) 下午3:45 @.>; @.***>; 主题: [R0S3NB3RG/PINNs-for-wave-equation] velocity is not a constant number (Issue #1)

I saw that the velocity is a constant 1 in your code. I want to use a more complex velocity model to replace this constant velocity model for training, such as the marmousi velocity model, but I don't know how to set it up because this velocity requires the use of a tensor related to the x and y coordinates of space.Do you have some ideas about this question? Best wishes!

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: @.> — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.>