-
thanks for the great work. I see that the loss of the netwok is a self implemented binary cross entropy loss parameterized by class weights. (I guess the weights are used to adress the imbalanced dat…
-
Reported in https://stackoverflow.com/a/70178087/17100540 by @DahnJ.
Given imbalanced data, you want the same value to be split across multiple output partitions:
```python
In [1]: import dask.da…
-
It seems that this is pretty difficult to make some grid-search on `OneClassSVM`
I wanted to search the `C` parameter with a `RandomizedSearchCV` and give the `balanced_accuracy_score`. However, `O…
-
Did not configure the model at all, simply ran
```
> from keras_fcn import FCN
> fcn_vgg19 = FCN_VGG19(input_shape=(500, 500, 3), classes=21,
> weights='imagenet', traina…
-
## Bug Description
I have used your template before (approx 1 year ago) without issues. Now, I re-cloned your project, installed it's dependencies, and I am trying to run the training job: `poetry …
-
In the paper, you say that "we keep groups with a normalized entropy smaller than a threshold empirically set to T =0.9."
However, in your code, why set the imbalanced_threshold as "mean - std" ?
-
how you treat unbalanced data for splitting and gini ?
any algorithm description pls
-
Why Ethnologue codes? May be it is better to use glottolog or ISO-639-3? Now there are some Ethnologue codes, 24 `0`, 26 `-` and some codes that appeared more than once in database. In glottolog you h…
-
#### Learning Goals
1. Learning different techniques that one needs to adapt if there is an imbalance in the dataset.
### Exercise Statement
1. Different Resampling techniques to handle the datas…
-
Hello,
I have a process per GPU and the amount of data distributed across all the processes could be imbalanced so a process could complete training faster than other processes.
If a process co…