search

zmy-9 / MF-DAKT

2021 CIKM: source code for the Multi-Factors Aware Dual-Attentional Knowledge Tracing (MF-DAKT)
16 stars 5 forks source link

How to calculate ednet_difficulty.csv? #2

Open meta-tabchen opened 2 years ago

meta-tabchen commented 2 years ago

Our team wants to implement the model by PyTorch, can you give more details about how to generate the ednet_difficulty.csv?

zmy-9 commented 2 years ago

Hi, MF-DAKT defines the correctly answer rate of a target question as its difficulty, i.e. the correctly answered number of the question / the total number of answered number of the question.

codingchild2424 commented 1 year ago

Your work is good. But I also have a question about difficulty attributes in ednet. I can not find the code for calculating the difficulty of ednet. So, my question is below.

  1. If the difficulty is calculated from the whole dataset, then the difficulty contains the whole datasets' difficulty representation.
  2. Then, when we split the datasets to train datasets and test datasets, the train datasets' difficulty was contaminated by the test datasets' representation (cheating), because the difficulty representation was calculated using the whole datasets.
  3. If you don't mind, can you give me the code to reproduce how to calculate the ednet difficulty?
icemy commented 1 year ago

Our team wants to implement the model by PyTorch, can you give more details about how to generate the ednet_difficulty.csv?

Hello, have you open sourced the implementation of the PyTorch version?