search

jiazhou-garland / ExACT

MIT License
22 stars 0 forks source link

AFE #5

Closed caomq123 closed 1 week ago

caomq123 commented 2 weeks ago

Thank you for your contributions. Could you please explain the relationship between the sample numbering after adaptive sampling and the corresponding sample paths in SeAct_sampled_train.txt? It seems that the number of samples after sampling does not match the number of samples in the txt file. Moreover, why are some of them in even numbers in the txt?

jiazhou-garland commented 2 weeks ago

Hello, the sample number will vary if you change the following two hyperparameters in the init function:

self.sample_event_num_min = 15000 self.sample_event_threshold = 30

Both parameters are recursive boundary conditions and they are different for different datasets.

jiazhou-garland commented 2 weeks ago

For Q1: 'The number of samples after sampling does not match the number of samples in the txt file.' It is very likely that your choice of the above two parameters is inconsistent with mine.

For Q2: 'why are some of them in even numbers in the txt?', The recursive function will stop when meet boundary conditions, and the final overall data structure of event frames is like a binary tree, and the number of leaf nodes in a binary tree is not necessarily even.