Explainable Hopfield Neural Networks Using an Automatic Video-Generation System
Round 1
Reviewer 1 Report
-The paper should be interesting ;;;
-add some photos about sensors, measurement - add arrows what is what;;;
-please add a block diagram of the proposed method;;;;
-please add a block diagram of the proposed research;;;; what is the result of the paper?
-please add photos of the application of the proposed research, 2-3 photos ;;; for example fault diagnosis
-please add sentences about future analysis;;;
-fonts of figures should be the same/bigger, ;;;
-the paper should be more scientific, there are too many algorithms (code)
-add arrows, axes, labels what is what;;;;;
-formulas and fonts should be formatted;;;;
-references should be 2019-2021 Web of Science about 50% or more;; 30-40 at least.;;;; show knowledge
-is there a possibility to use the proposed methods for other problems? ;;;;
-please compare different methods advantages/disadvantages,
for example for analysis of thermal images:
1)
Fault diagnosis of electric impact drills using thermal imaging, Measurement, Volume 171, 2021,
https://doi.org/10.1016/j.measurement.2020.108815
-Conclusion: point out what are you done;;;;
Author Response
Response to Reviewer 1 Comments
Point 0. About English.
Response about English: Erratas and typos have been corrected. If the reviewers require it, this will be sent to a native speaker after this revision because new changes in the document have been performed.
Point 1. “Add some photos about sensors, measurement”
Response 1: Figure 1 has been replaced by a figure explaining our approach (hypothesis, methodology, so on), so none photos from others papers will be incorporated. We think that with the references is enough.
Point 2. “Please add a block diagram of the proposed method”
Response 2: Figures 3 and 5 have been improved. Now, we think they explain the proposed method better. Additionally, we add explanations for arrows and others elements in them. We add the following explanation: “Additionally, our proposed method is presented and explained in Figures 3 and 5 by using block diagrams [23].”
Point 3. “Please add a block diagram of the proposed research, what is the result of the paper?”
Response 3: We add a new figure explaining hypothesis, objective, methodology, result and evaluation. We add a new paragraph about this: “More formally, our research (hypothesis, main objective, methodology, result and evaluation) is explained and formalized in the Figure 2. Additionally, our proposed method is presented and explained in Figures 3 and 5 by using block diagrams [23].”
Point 4. “Please add photos of the application of the proposed research, 2-3 photos”
Response 4. We add new 9 photos of our application in the appendix section.
Point 5. “Please add sentences about future analysis”
Response 5. I think in future work we explain future analysis
Point 6. “Fonts of figures should be the same/bigger”
Response 6 Done
Point 7. “the paper should be more scientific, there are too many algorithms (code)”
Response 7. All codes have been explained in scientific way by using comments in natural language. In the computer science field, sometimes algorithms and pseudocodes are very useful in order to reply the experimentation. We have improved explanations of these pseudocodes. Also, we add a new section (appendix)
Point 8. add arrows, axes, labels what is what
Response 8 All figures have been improved and arrows, axes and labels have been explained.
Point 9. formulas and fonts should be formatted
Response 9 Formulas and fonts have been revisited. If you detect formulas and fonts which need to change, please tell us.
Point 10. references should be 2019-2021 Web of Science about 50% or more;; 30-40 at least show knowledge
Response 10 We add 11 new references from 2019-2021 (see References with color blue)
Point 11. is there a possibility to use the proposed methods for other problems?
Response 11 In the abstract you can read: “work constitutes a novel approach to get explainable artificial intelligence systems in general”. However, in order to clarify this point we add a new paragraph: “It is important to highlight, although we employ our video generation method for explaining HNN’s execution, it could be applied to any problem whose solution involves the implementation of a certain algorithm (classical or machine learning ones) because we use the execution traces to automatically obtain the sequence of frames that will compose the future explainer video. More formally, our research (hypothesis, main objective, methodology, result and evaluation) is explained and formalized in the Figure 2. Additionally, our proposed method is presented and explained in Figures 1 and 4 by using block diagrams [28].
Point 12. please compare different methods advantages/disadvantages
Response 12 I add a paragraph in which I compare my approach with others related approaches: “In the literature, we can found several works about automatic video generation which differ quite a lot from our proposal [21,59]. In [21] a framework for machine learning to generate videos called ArrowGAN is presented and in [59] a multi-modal approach for automatically generating hierarchical tutorials from instructional makeup videos is proposed. The main difference is that our case videos are automatically generated from execution traces (symbolic information) to get a better understanding of machine learning algorithms. In the mentioned approaches videos are employed to generating others videos with different objectives. On the other hand, there is another line of work which aims to automatically generate textual explanations of software projects (code) [49]. This paper is focused in the generation of function descriptions from example software projects and it is applied to study the related problem of semantic parser induction. It employs mainly code as main resource of communication, none video is generated. To the best of our knowledge, this is the first time that data-to-text based on fuzzy logic and software visualization techniques are employed to automatically generate videos from execution traces.”
Point 13. add Fault diagnosis of electric impact drills using thermal imaging, Measurement, Volume 171, 2021, https://doi.org/10.1016/j.measurement.2020.108815
Response 13. We add the reference about Fault diagnosis of electric impact drills using thermal imaging
Point 14. “Conclusion: point out what are you done”
Response 14 In conclusion a new details have been incorporated paragraph points out what I did (new verbs have been employed, now we think is more clear what are we done”
Reviewer 2 Report
The paper is interesting and well-written and, in general, I think it addresses a relevant topic. However, the methodology, hypotheses, experiments, results and conclusions are totally missing. In those conditions the paper cannot be accepted.
What is the objetive of your research? Please provide some hypotheses. Later, describe a methodology with all details about how you plan to validate your hypotheses in relation with the proposed new approach or technology. This methodology must include some experiments with quantintative results, so different variables and indicator can be studied. These results must show if the initial hypotheses were true or not.
Author Response
Response to Reviewer 2 Comments
Point 0. About English.
Response about English: Erratas and typos have been corrected. If the reviewers require it, this will be sent to a native speaker after this revision because new changes in the document have been performed.
The paper is interesting and well-written and, in general, I think it addresses a relevant topic.
Point 1: However, the methodology, hypotheses, experiments, results and conclusions are totally missing. In those conditions the paper cannot be accepted. What is the objective of your research? Please provide some hypotheses. Later, describe a methodology with all details about how you plan to validate your hypotheses in relation with the proposed new approach or technology. This methodology must include some experiments with quantitative results, so different variables and indicator can be studied. These results must show if the initial hypotheses were true or not.
Response 2: Thank you for your review. You are right. Now, we have rewritten some parts of the version 1 (marked in blue color). We add a new section 7 about the Evaluation of our hypothesis which has been also explained in the section 1. We add special figure in which hypothesis, objective, methodology, result and evaluation have been explained. We think now:
- Objective of our research is explained
- Hypothesis is provided.
- Methodology is described
- A new section validate our hypothesis by using an evaluation instrument
Round 2
Reviewer 1 Report
The paper is good enough to publish
Reviewer 2 Report
Although I think the paper could be still improved in order to increase its quality, at this point I do not have any concern. All my previous comments have been addressed and I think the paper may be accepted.