-
Notifications
You must be signed in to change notification settings - Fork 152
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Does spotpy's Dream not exactly Vrugt, J. A. (2016) proposed Dream? #284
Comments
Hi @BeingHapppy, |
Hi @thouska, thank you for your reply! Implementing the missing features is a challenge for me, but I'll try. I appreciate that you said you can provide help. |
Thanks @BeingHapppy for reporting. Unfortunately as @thouska also mentioned, for concrete implementation I also currently don't have time. Let's see what's going on in autumn. Cool that you use this tool. |
Thank you for your reply. MCMC method is the foundation of our team's research, and spotpy dream is the easiest tool we've ever used. Before meeting spotpy, we spent almost a whole year learning MCMC and still couldn't program it. I just did very little work and thank you for opening the doors of MCMC to us. Looking forward to the future of MCMC, your work is amazing.
Best wishes,
Yantong
…------------------ Original message ------------------
From: "Ben";
Sendtime: Wednesday, Jun 22, 2022 4:31 AM
To: "thouska/spotpy";
Cc: "Yantong ***@***.***>; "Mention";
Subject: Re: [thouska/spotpy] Does spotpy's Dream not exactly Vrugt, J. A. (2016) proposed Dream? (Issue #284)
Thanks @BeingHapppy for reporting. Unfortunately as @thouska also mentioned, for concrete implementation I also currently don't have time. Let's see what's going on in autumn. Cool that you use this tool.
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you were mentioned.Message ID: ***@***.***>
|
Hi @thouska and @bees4ever
Thank you very much for your work. I've been using spotpy for almost three years. Today, I find some differences between spotpy's Dream method and the Dream method proposed by Vrugt, J. A. I'm curious why they are different.
(1)In spotpy, it seems we use only one couple chains to get_new proposal vector, but Vrugt seems use random number of chains.
here is the code from spotpy:
` def _get_gamma(self,N):
#N = Number of parameters
p = np.random.uniform(low=0,high=1)
if p >=0.2:
gamma = 2.38/np.sqrt(2*int(N))#/self.gammalevel
else:
gamma = 1
return gamma
below is from Vrugt:
The highlighted part in the figure is the evidence that Vrugt uses multiple chains.
And Vrugt also point out the total number of chains should equal 2*sita+1, sita is the max number of couple chains when proposal vetor.
(2) outlier chain detection
In the paper of Vrugt2016, he says that he use outlier chain detection during the sampling, but he did not show how to do. It seems spotpy Dream does not have outlier chain detection.
The text was updated successfully, but these errors were encountered: