Miracle-2001
commented
9 months ago Hello~ I appreciate it for your interest on our work! Frankly speaking, Het-EmotionNet is indeed designed for a multi-modal dataset. ( i.e. the dataset is expected to have eeg, eog, ecg signals, etc.) And you are right! FACED is a unimodal dataset so we only use eeg signals in our experiments. So if I do not misunderstand your question, there is no need to "dismantling", in Graph Transform Network, we can only use one single signal. (And if you wonder: Is that "heterogeneous" if we only use one type of signal? Or : Why do you choose HetEmotionNet considering that FACED is a unimodal dataset? I have to say : Het-EmotionNet is an influential and effective structure, so we choose this modal even if there is no "MULTI-MODAL")
Hope I can solve your question~ If I somehow misunderstand ( for my poor English :( or there is still something I do not mention, please feel free to give a feedback~ Best wishes
Kaiyuan Zhang
Miracle @.***
------------------ 原始邮件 ------------------ 发件人: @.>; 发送时间: 2024年2月29日(星期四) 晚上7:12 收件人: @.>; 抄送: @.***>; 主题: [Miracle-2001/GNN4EEG] On the dismantling of Het_emotionnet's multimodal graph neural network structure into a unimodal model on FACED (Issue #4)
Hello, from my superficial understanding, Het-Emotionnet describes a two-stream heterogeneous recurrent graph neural network, a graph neural network based on a multimodal data implementation. According to my understanding, I think you, the author, must have used some method to achieve the dismantling and stripping of the multimodal recurrent neural graph neural network into a graph neural network under a unimodal dataset FACED?
My understanding may be superficial, if you have any documentation of your research on the Het model, I sincerely hope that I can make friends（神交） with you!
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: @.***>
Hello, from my superficial understanding, Het-Emotionnet describes a two-stream heterogeneous recurrent graph neural network, a graph neural network based on a multimodal data implementation. According to my understanding, I think you, the author, must have used some method to achieve the dismantling and stripping of the multimodal recurrent neural graph neural network into a graph neural network under a unimodal dataset FACED?
My understanding may be superficial, if you have any documentation of your research on the Het model, I sincerely hope that I can make friends（神交） with you!