-
For the moment, the nc_benchmark generator function allows for a nc_first_task option, which is good for pre-training in the class-incremental learning scenario. However, the same kind of option is no…
-
I have all my data in an H5 file and it is large 400,000 data points and 50,000+ features. (all float 32) I have access to google colab with 51 gb of memory, and I have been looking for solutions to …
-
Hi @brjathu, thanks for sharing your work! In Table 2 you provide the experiment for ImageNet-100 and ImageNet-1000. However, I can not find details about how you construct the ImageNet-100 dataset fo…
-
I copy this post from [my blog](https://haoyu.love/blog608.html).
-----
What is the most distressing thing in life?
Not money, not paper, but the author released their fantastic code while yo…
-
Hi,
Thank you very much for anwsering my problem, I find that the accuracy of the results is only about 0.8, how can I improve the accuracy of the result?
-
Hi QueuQ,
I recently tried to reproduce the result you made,but i got a question. The last class acc was always 0% when i trained it. Can you help me to solve this problem?
-
In the codes, the encoder for CIFAR100 has one more layer and each layer has large feature dimensions. (4 layers with [64,128,256,512] feature dimensions)
But all of the previous works use much light…
-
@fmcp Thanks for the help before on the cifar dataset.
Now I am trying to use the code to reproduce the imagenet results. I notice that in the readme.md, it mentioned that:
```
For ImageNet, you…
-
As the training process is costly every time to take the whole data, is it possible to perform incremental learning on CRF based on users' annotation in real time?
Do you think that it is a good i…
-
Hello
How are you?
Thanks for contributing this project.
I have a question.
Does this project support class-incremental training for saving training time without catastrophic forgetting?
Let's su…