yashaektefaie
commented
3 weeks ago Apologies for the delayed reply! Thanks for your kind words, I am glad you found the perspective helpful.
An example of this approach is Holoprot, which simultaneously learns surface and sequence representations of a protein within a unified architecture. Rather than processing these modalities separately and combining them at a later stage, Holoprot jointly optimizes over both, allowing it to reason across surface and sequence features in an integrated manner for improved protein modeling.
If this is still confusing I can provide more examples.
Hello Yasha Ektefaie, I saw one of your Multimodal learning with graphs reviews on Nature Machine Intelligence. I have read this review very carefully. It is really an excellent article, which gives me a very good inspiration for my subsequent research. Here, I am very grateful to you. I am contacting you to inquire about the integrated multimodal architecture you mentioned in your overview of Figure 2(b). I really want to know about this part, but I can't find any other information. Could you give me some suggestions? Thank you very much!