Course talk:CPSC522/Generative Adversarial Networks

From UBC Wiki
Jump to navigation Jump to search

Contents

Thread titleRepliesLast modified
Critque100:18, 15 March 2016
suggestions100:17, 15 March 2016
critique100:14, 15 March 2016
Critique100:12, 15 March 2016

Hi Ricky,

First of all, let me apologize. I totally mixed up the due date of critiquing with the due date of Final Draft. Let me thank you for contributing in our Wiki pages, and what a fantastic job you have done. There are parts that I did not understand. In these cases hyperlinks that guide me towards some useful sources would really come handy. I strongly suggest to add these. For this draft the material is good, but maybe add more for the final draft (the content part of the page is short and perhaps less than minimal that is necessary to completely understand for a reader like me). But other than this, it is a great page. Thank you again!

Cheers,

Mehrdad Ghomi

MehrdadGhomi (talk)20:05, 12 March 2016

Right, the problem with adding hyperlinks to parts that you don't understand is.. I don't know what you don't understand. :/

A bit more elaboration would help, ie. is it the background section that is not clear or the introduction to GAN, etc.

Thanks for feedback!

TianQiChen (talk)00:13, 15 March 2016
 

suggestions

Hi Ricky,
Good work! Your page is very well written and organized. You clearly described the incremental contribution of one paper over the other. However, I did find the page a bit difficult to understand (maybe because I'm completely new to the topic). As Jordon said, you could probably add more links in cases where you use terminology that many of us don't know about. Other than that, excellent page!

AdnanReza (talk)18:22, 12 March 2016

Right, I'll re-read this wiki page and see if there are some terminology-heavy areas that require a bit more elaboration. I'll add more explanations as soon as I figure out where to add them!

Thanks!

TianQiChen (talk)00:17, 15 March 2016
 

Hi Ricky,

A solid draft. I don't think I fully understand it at this point, but the writing style and organization are good, and I understood enough that I'm confident the rest would come in time. You also make it evident what the contribution of the second paper is, which is good.

General comments:

  • I think having more links would be good in cases where you use terms another student may not know, but that is outside the scope of the paper to explain.
  • It's good that you have the sources for the figures when I click on them, but it would probably be good to cite them directly in the page as well.

Section-specific comments:

  • Laplacian pyramid
    • I think a figure would be really useful here, even though there's one in the next section.

Nicely done!

Clear skies,
Jordon

JordonJohnson (talk)08:05, 11 March 2016

Great; I'll add figures for the Laplacian pyramid. (Switched some arrows for the existing figure.)

The problem with adding links beside terms that other students may not know is that.. I don't know which terms others may not know.

Thanks for feedback!

TianQiChen (talk)23:55, 14 March 2016
 

Hi Ricky,

Great job. I appreciate your work on this topic although there are still many contents I do not understand. I have two suggesions for you.

  • I hope you can reorganize the hierarchy of your page so that it is clear and coherent. For example, Section 1 and Section 3 have the same title. Maybe you can change one of them.
  • I think you can add your own thoughts on how successful Laplacian Generative Adversarial Network was and how it can be improved.

Sincerely,

Ke Dai

KeDai (talk)01:48, 14 March 2016

Regarding the second bullet point, since this is a wiki I've tried to reduce the amount of subjective information that gets put on the page, and only include points that are either objective or that most people would agree about..

But my own thoughts are that the LAPGAN paper is a great read because it introduces a framework that uses independent (but structured) GAN models. The use of Laplacian pyramid is a great example of this structural idea, but it is only an example (too restrictive) and likely will not be developed much further. Further improvements would be to the GAN approach, with LAPGAN as a comparison/motivation.

For the first bullet point, GAN is more of a general idea rather than a concrete algorithm. It just turns out that the paper that introduces it is named GAN... I'm not quite sure what to change the titles to, but if I think of something good I'll do it.

Thanks!

TianQiChen (talk)00:12, 15 March 2016