3 ms·
How do you stop piracy on something like this? Surely I could take a model, refine it slightly to get distinct weights (ie. not a flat clone) and unload it as
by wokwokwok 5y ago
How do you stop piracy on something like this?
Surely I could take a model, refine it slightly to get distinct weights (ie. not a flat clone) and unload it as my own novel variant?
...and I mean, is the problem not utterly fundamental?
When is a model “yours” or sell?
When you trained it? When you’re not using transfer learning? When you wrote the paper? When you own the training data? When you can recreate it from scratch?
I see this being problematic to look after...
- belval 5y agoThe value of these could be in private datasets. Model architectures are not really patentable and ownership of a trained model has always been very muddy water. That being said, if I am selling a bird classifier chances are that to train it I had to combine multiple datasets and someone can't easily finetune it to get better results.
- wokwokwok 5y agoThey dont need to fine tune it for better results, they only need to fine tune it to give different results and sell it for $5 less than you are.
- pepe-sad 5y agoI believe it has a kind of pre-moderation.
- belval 5y agoI misunderstood your point. The "attack" that you are describing could probably be mitigated by doing activation analysis. Given random initialization even with fine-tuning the activation distribution will be similar. Same method should be used to weed out people just repackaging the ImageNet pretrained model from torchvision. That being said you can still distill a network into another with a teacher-student approach and that would be undetectable.
- aledalgrande 5y agoThe marketplace should probably be on datasets.
- Moosdijk 5y agoI'd rather buy a model than a dataset that I have to spend money to train a model with.
- aledalgrande 5y agoYou are limited in expanding your dataset and making your model better, if you only have the model. And I would rather see how clean/balanced the dataset my model was built from was.
- Moosdijk 5y agoWhen using models becomes more mainstream, I think it's better to strive for transparency than expandability.
- bonoboTP 5y agoRight, you could steal the model by querying lots of images and using teacher-student model distillation.
- Moosdijk 5y agoDisclaimer: I'm not very familiar with the current state of CV and the way models are trained. What authors could do is train the model on a dataset, but add a specific image that it flags as a key. That way, if you keep the image private, you can check if a model is a clone of yours if recognises that specific image.