Issues with Fig.5

[EmanueleFrancazi]

Description

Figure 5 is not shown correctly: - The big image on the left should not be present. - The caption has some problems: - "at initialization" should be underlined, not circled. - Appendix reference is not shown correctly in the last sentence (before the blue comment). - The blue comment at the end should be removed.

(Optional:) Please add any files, screenshots, or other information here.

No response

(Required) What is this issue most closely related to? Select one.

Figures

Internal issue ID

d24d0b5b-012b-4a9e-a2a0-f417b7ee021c

Paper URL

https://arxiv.org/html/2306.00809v4

Browser

Firefox/127.0

Device Type

Desktop

[html-feedback-bot[bot]]

Location in document: ltx_tag

Selected HTML: igure 5: Left: IGB and Performance Bound: The accessible performance range conditioned on the behavior of $\max${i}{f{i}}roman_max start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT { italic_f start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT } in a balanced binary dataset, in accordance with Eq. (14). Right: Comparison of the trend of $\max${i}{f{i}}roman_max start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT italic_f start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT with that of accuracy during the learning dynamics, varying with the level of guessing bias at initialization (IGB). The curves show a consistent pattern with the diagram on the right. They also demonstrate that the time for IGB absorption increases with the level of IGB itself. The simulations were conducted on an MLP-mixer using a binary dataset (dogs vs cats from CIFAR) as input; more details on the setting and additional experiments with more architectures are provided in Appendix []. [Are the colors indicating the amount of IGB at initialization? I would write that explicitly]

[github-actions[bot]]

Hello @EmanueleFrancazi, thanks for the issue report! We are reviewing your report and will address it as soon as possible.