Interactive Guiding Sparse Auto-Encoder with Wasserstein Regularization for Efficient Classification
Round 1
Reviewer 1 Report
There is no performance analysis of proposed parallel approach. There is no information about obtained speedups, parallel computation efficiency and scalability for different datasets sizes and different number of used CPUs and GPUs. Paper must be updated with this analysis because there is "parallel" in article title. Reader doesn't know how parallel computations impacts on quality of obtained results and performance.
Research is relevant, interesting and proposed approach is original but relevant part about parallel computation is missing. Authors proposed parallel algorithm but discussion about impact of parallel computation on quality of results is missing. Also there is no information about parallel algorithm performance - speedup, efficiency and scalability. Why authors used parallel algorithm? I assumed that for better results, possibility to solve larger problems and shorter computation times - there is no evidences and arguments for that in paper.
Author Response
Reply to comment is uploaded.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments of applsci-2393331
A parallel guiding sparse autoencoder (PGSAE) to guide the information by two parallel guiding layers and sparsity constraints is proposed in the article.
Major revision questions,
1. The significant contributions of the research do no doubt shown in the research report, though it is an interesting issue for study and the reviewer agrees with the results of the investigation.
2. Please check whether the variables are suitable for the rightness of each equation, for example, the transform of a matrix shown in Eq. (1) should be defined. Checking up all of them carefully is necessary.
3. Certainly, the presence of all different color lines in Figure 1, should be explained in advance. By the way, the “Classifier” shown in Figure should also explain in detail. That is, the ELM adopted in the Figure. 1 is ambiguous, how to do the operation?
4. Moreover, “? is a amount of input samples” is defining the sample number of Eq. (3), but why it becomes “? is the number of input samples” used in Eq. (5)?
5. The “activation function” deployed in the operation of machine learning models analyzed in “Experimental result of PGSAE” section should be definitely announced. It is believed that this function seriously affects the performance of the proposed PGSAE.
6. The reviewer thinks that the run times of the test will also impact the outcomes from the proposed algorithms, that is, the authors said “We ran tests five times with randomly shuffled samples.” Which 5 times are held by the investigation, is it enough? Discussion, please.
7. Furthermore, the detailed analysis of datasets (which has been listed later) adopted as the training components for the classifier algorithms is very significant. There has an extra suggestion in the discussion of the amount of the dataset, that is, is it providing enough for the analysis employed to the simulation help in the work? Please has a brief discussion.
8. One more suggestion is to provide one or two flow charts for the proposed Classifier Algorithms put in the suitable place of the thesis which can make clear the significant illustration.
9. Please respond to what is wasting time for simulation and processing of the dataset and dataset learning.
10. Providing some up-to-date citations is necessary, the nearest year of the current references is “2020”?
11. The proof of validation and certification that can allow certain developments on the SAE-based machine learning algorithms are believed very interesting. Therefore, the authors are suggested to propose suitable reasons to ensure that the development is appropriate and efficient of the proposed PGSAE.
Moderate editing of English language required
Author Response
Reply to comment is uploaded.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
I have no further comments.
Reviewer 2 Report
1.Most of the illustrated flaws are revised appropriately.
2.The editorial format should follows up the rules of the Journal.
