@baadealan
CS PhD Student at Stanford
AI
The poster says image representations converge to features whose dot product with another image's representation equals their pointwise mutual information.