msseibel
msseibel
https://github.com/intel-isl/MultiObjectiveOptimization/blob/1c6d0d503ccf33cc83d5b6c356ca2fc2bf255606/multi_task/losses.py#L79
Currently I am using my ground truth labels of the target dataset for searching the optimal loss weights. In your code they are called `sloss, tloss, jdot_alpha`. Are there other...
Not important, I am just curious about your code style in the Deepjdot class. Why do you assign functions to the class with the following structure: class Deepjdot(object): .... def...
The old soft-dtw implementation was changed to the new soft-dtw-divergence implementation. This makes it also easy to test the new soft-DTW-divergence approach, which corrects the 'bias introduced by entropic regularization'....
In a paper by Chuang et al. 2020 (Debiased Contrastive Learning), it has been shown that false negative samples can deteriorate the performance of contrastive learning. Are you aware of...
What does the 150 mean here? Did you find that value by trial and error or does one epoch consist of 150 iterations? I assume, that the value should be...
Hello together, thank you for your repository. It seems to work well for extracting .vol data. We plan to export our data to another file format such as Nifti, Dicom...
What is the reason of using Dense(2,activation='softmax') instead of Dense(1, activation='sigmoid')? Is it related to the Gradient Reversal Layer? If so, can you explain?
In the "LE-UDA: Label-Efficient Unsupervised Domain Adaptation for Medical Image Segmentation" paper, it is rightly mentioned that target labels are not available for hyperparameter selection. I just want to clarify...
Many layers are not supported by this lib, even if they don't affect the resulting receptive field. Please support s.t. those layers can be quickly replaced. or add something like...