3 ms·

So, basically a NOT operator on a neural network. Does this even require a differentiation?

• nl 4y ago
  A NN is trained to make something indistinguishable from reality as possible - so it can't tell the difference. The inverse of that will just claim reality is fake too. What you need a a deep fake and a real video of the same thing, then train on the difference. Clearly this is impossible - which is what makes the problem hard.
  • sigmoid10 4y ago
    You actually don't need that. You only need a set of real videos and a generator for fake ones. Then train the discriminator to tell these two classes apart and make use of its differentiability to update the generator in tandem with the discriminator.
    • anankaie 4y ago
      Would you run into class balance issues if you do not have a growing pool of initial videos?
      • sigmoid10 4y ago
        Not if you have a balanced dataset. If not, you're already getting into trouble with GANs, which is why they were superseded by diffusers on image generation tasks.
• rusticpenn 4y ago
  Not really a not operator. It’s more like tuning the generated models until the fake detector cannot detect it.
• anankaie 4y ago
  It depends on how you set it up. You can use a very non-expressive valuation that assigns a score, but then you need to use reinforcement learning techniques to transform that score to a model update. Alternatively, if your valuation represents a differentiable metric function from your target you have a way of going directly from your output to a model update. The second way requires dramatically fewer update steps (usually) than the first. Thus - having your adversarial target be differentiable definitely helps, though it is possible to do even absent such a criterion.