yannikschaelte
commented
2 years ago Hi @hpesonen , thanks for raising these points. I hope to have addressed all of them in #578.
L34: I’m not sure if if completely correct to describe the forward process model as a “black-box” model? Wouldn’t this make it impossible to model the parameters of the simulator a priori (as we do in ABC)?
The simulator model used in ABC methods has been referred to as "black-box" in various papers (e.g. Palmier et al. 2020, Auzina et al. 2021, Dyer et al. 2022), where "black-box" indicates that the likelihood cannot be assessed. As this can however lead to understanding problems, may not be standard terminology, and is not necessary here, I have removed the term from the manuscript.
Figure 1 : What does the illustration of a graph above the summary statistics in Figure 1 refer to? I’m also not sure what the illustration above “model” is either.
The network is supposed to indicate a neural network, an approach used in one of the contributions. The model illustration is supposed to be bacteria and viruses, exemplary for biological systems. I have updated the caption to include more details.
Plotting in the colab-notebooks is not currently working.
Colab is still running Python 3.7 by default, I would hope that this will be updated soon, as many basic packages including matplotlib, numpy, scipy (and pyabc) now require Python >= 3.8. Some things still work, but apparently not all. I added a comment to the documentation remarking this.
Hi!
Here's a few minor comments about the paper part of the submission
I agree with Issue #566
L17-18 Clarify what type of noise is referred to in the following sentence “In particular, we implement algorithms to account for noise…”
L34-35: Following sentence is incomplete “In a nutshell, ABC circumvents likelihood evaluation by accepting if a distance between simulated and observed data is below a threshold”.
L34: I’m not sure if if completely correct to describe the forward process model as a “black-box” model? Wouldn’t this make it impossible to model the parameters of the simulator a priori (as we do in ABC)?
L40: I feel that the following sentence would benefit for declaring why ABC is computationally expensive “While conceptually simple and widely available, ABC is computationally expensive…”
Figure 1 : What does the illustration of a graph above the summary statistics in Figure 1 refer to? I’m also not sure what the illustration above “model” is either.
References:
Some of the titles are incorrectly all lower case, e.g. "Population growth of human y chromosomes…"
In Lintusaari et al. names Kangasraasio and Jarvenpaa should be Kangasrääsiö and Järvenpää, respectively.
Additional comment:
Plotting in the colab-notebooks is not currently working.