4 ms·
The obvious answer is to learn the metrics itself too. OpenAI does interesting work in this area: "Deep reinforcement learning from human preferences", for exam
by sanxiyn 7y ago
The obvious answer is to learn the metrics itself too. OpenAI does interesting work in this area: "Deep reinforcement learning from human preferences", for example. https://arxiv.org/abs/1706.03741 https://arxiv.org/abs/1706.03741
- AlexCoventry 7y agoThat's basically what GANs do, too. The recent results are amazing. https://learning-to-paint.github.io https://learning-to-paint.github.io
- inimino 7y agoI'm not sure if this is obvious, if what it means in practice is a metric that may be the wrong metric, but which is visible, being replaced by a metric that is invisible. At least in the first case we can reason about the metric and why optimizing for it might go wrong.