Deep Multiphysics and Particle–Neuron Duality: A Computational Framework Coupling (Discrete) Multiphysics and Deep Learning

Round 1

Reviewer 1 Report

This work focuses on Deep Multiphysics as an approach to couple first-principle modeling and Machine Learning method. The author applies the method to design of a microfluidic device, which is used for separating cell populations with different levels of stiffness.

The manuscript presents an interesting topic. Nevertheless, the manuscript can be improved in my opinion, as detailed bellow. With appropriate consideration to the suggested comments, as well as other reviewers' comments, I find this paper appropriate for publication in Applied Sciences journal.

 

The authors mentioned “Reinforcement Learning” in the manuscript. I would recommend explaining this term. According to the manuscript, the ANN is capable of predicting the cell stiffness. Is it also capable of predicting the velocity and location of particles? How does the accuracy of the ANN model depend on the number of simulations? I would recommend comparing the results with similar studied in the literature.

 

 

 

 

 

Comments for author File: Comments.pdf

Author Response

Please see attached document

Author Response File: Author Response.docx

Reviewer 2 Report

See attachement.

Comments for author File: Comments.pdf

Author Response

Please read attached document

Author Response File: Author Response.docx

Reviewer 3 Report

This work tried to introduce a computational framework that combines multiphysics and artificial neural network.

However, the title seems very general while the text is very specific to the problem of "stiffness of cells".

I have a general and major question: is the underlying rationale firstly proposed by the author? (According to my literature research, the answer seems to be 'no'.) If not, please give proper context of your work, e.g. citing a review article. Otherwise, the whole idea of the so-called "in-parallel coupling" sounds misleading as it's unclear from the Introduction what the real difference is between "in-series coupling" and "in-parallel coupling".

It would be ideal if the author could give more details on how this method was implemented. Current text has quite vague description of this methodology with those ideas/equations which are well-known already. Besides, the author can choose to attach some video clips of simulation or source code.

Minor issues:

The examples of "in-series coupling" in the Introduction seems off-topic. The author mentioned "ML was used to learn Density Functionals", this is not related to the application system introduced in this work. More related examples are preferred.

 

Author Response

Please see attached document

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

This reviewer would like to thank the author for addressing the raised issues. This reviewer is now convinced about the usefulness of this novel approach, especially by the intriguing simulation videos attached in the SI. 

I am pleased to recommend this work to be published on Applied Sciences.