search

wzzheng / HDML

Implementation of Hardness-Aware Deep Metric Learning (CVPR 2019 Oral) in Tensorflow.
151 stars 28 forks source link

[Maths] Symbol in the paper? #4

Closed stoneyang closed 5 years ago

stoneyang commented 5 years ago

Hi, @wzzheng ,

Thanks for your contribution of this repo, and the original work.

In reading your paper, I found something that might be typo:

  1. In the 3rd paragraph in Sec. 3.2, you name the augmented harder negative sample as z with a tilde above it and a minus as a script, while in the append maths formulation (ie, eq.4) and the remaining of this section, I could not find any other symbol like this.
  2. In the 1st paragraph in Sec. 3.3, there is another z with a tilde above it and a minus as a script, again, I still could not find any other symbol like this in the same section, but only z with a tilde above it without any script in the 4th paragraph in the same section.

Is this just a typo, or something not fully rendered?

wzzheng commented 5 years ago

Hi, thanks for your interests in our work! We apologize for the typos and will correct them in the next arxiv version. To answer your question,

  1. The symbol of augmented harder negative sample should be z with a hat above it and a minus as a script as in eq.4.
  2. It should also be z with a hat above it and a minus as a script. Also, in the 5th line of the 4th paragraph, the z with a tilde above it should be z with a hat above it.

You can see Figure 1 for an illustration of those symbols. Intuitively, symbols with a hat refer to augmented feature/embeddings, and symbols with a tilde refer to synthesized feature/embeddings.

I hope you can understand the symbols even though I admit they are a bit confusing:(. You are welcomed to ask any further questions.

Thanks again!

stoneyang commented 5 years ago

Hi, thanks for your interests in our work! We apologize for the typos and will correct them in the next arxiv version. To answer your question,

  1. The symbol of augmented harder negative sample should be z with a hat above it and a minus as a script as in eq.4.
  2. It should also be z with a hat above it and a minus as a script. Also, in the 5th line of the 4th paragraph, the z with a tilde above it should be z with a hat above it.

You can see Figure 1 for an illustration of those symbols. Intuitively, symbols with a hat refer to augmented feature/embeddings, and symbols with a tilde refer to synthesized feature/embeddings.

I hope you can understand the symbols even though I admit they are a bit confusing:(. You are welcomed to ask any further questions.

Thanks again!

Thanks for your timely reply! I'd rather to say after post this issue, I found it as you pointed out. :)

Another question, In Sec. 3.3, you described the objective of generator J_gen as eq.9, while in the code, it seems different from your paper. Is that a problem?

wzzheng commented 5 years ago

Can you please be more specific about the differences so I can check out?

stoneyang commented 5 years ago

Can you please be more specific about the differences so I can check out?

In eq.9, J_gen is the weighted sum of J_recon and J_soft, where lambda is the weight term of J_soft; while in the code, the weight term is with J_recon in the form of (1 - FLAGS._lambda).

wzzheng commented 5 years ago

I believe it is our mistake. Thanks for pointing it out! I will modify the codes and relevant parameters.

wanghao14 commented 4 years ago

@wzzheng When will you update the next arxiv version?

wzzheng commented 4 years ago

@wzzheng When will you update the next arxiv version?

Hi, in one or two days. And thanks for pushing me :)