You can ask any question here.

[pykt-team]

[123kristin]

Thanks to the pykt library for your great contribution to the knowledge tracing field!. What is the parameter ‘emb_type’ in the At-dkt code? It's not the default value 'qid', is it?

[sonyawong]

hi~exactly, not all the values of emb_type are set to "qid" in our repo. For more detail about the parameters setting of our model, you can visit https://drive.google.com/drive/folders/1MWYXj73Ke3zC6bm3enu1gxQQKAHb37hz?usp=drive_link.

[123kristin]

You've been a great help

[123kristin]

About rkt model, How to get phi_array_1_2_3_4.pkl file?

[linchuanghong]

What is the meaning of "is_repeat" data in the data set generated after the original data is preprocessed?

[sonyawong]

It is a flag to denote whether the interaction is expanded. Since there is a question contains multiple knowledge components (KCs), for those questions, the interaction will be expand based on KC level one by one.

[cwsy92]

I understand that there are various scenarios e.g. Questions Level (All-in-One), KC Level (ALL-in-One), KC Level (One-by-One). When running the results, there are also various outputs that are generated, may I check if the Questions Level (All-in-One) scenario results is matched to the oriaucxxx results? (xxx is the fusion mechanism)

[sonyawong]

I understand that there are various scenarios e.g. Questions Level (All-in-One), KC Level (ALL-in-One), KC Level (One-by-One). When running the results, there are also various outputs that are generated, may I check if the Questions Level (All-in-One) scenario results is matched to the oriaucxxx results? (xxx is the fusion mechanism)

hi, thank you for ur interest in our work. The results ori+auc/acc+xxx(fusion mechanism) are the evaluation results in question-level without slide window setting. For the results of window setting, the results are denotes as window+auc/acc+xxx(fusion mechanism). Moreover, the KC Level test+auc/acc and window_test+auc/acc.

[cwsy92]

so, may I know which setting (if any) would be the best to examine in knowledge tracing?

[Xemorr]

Hi, what is the reason for breaking after 5000 items in the ednet preprocess?

[BojunZhan]

Because the ednet dataset is too large, due to the training time factor we only intercept part of it, you can control the size of the intercept as needed.

[Echo628999]

pykt团队您好，我先后在akt和simplekt上得到了在数据集as2009上auc，发现明显高于论文数据
同时我又在测akt在al2005上的表现，simplekt 和 akt的auc同时高达0.9左右，同样远高于论文数据，
因为数据集都是官方下载，我怀疑是论文里提到的标签泄露问题，请问如何修改，解决这个问题，复现出平台论文里介绍的数据

[Echo628999]

我们组想要用pykt完成一些工作，不知道佬能不能留下联系方式邮箱之类的咨询一些问题

[BojunZhan]

pykt团队您好，我先后在akt和simplekt上得到了在数据集as2009上auc，发现明显高于论文数据 同时我又在测akt在al2005上的表现，simplekt 和 akt的auc同时高达0.9左右，同样远高于论文数据， 因为数据集都是官方下载，我怀疑是论文里提到的标签泄露问题，请问如何修改，解决这个问题，复现出平台论文里介绍的数据