Caring for your home is a lesson in the 2nd law: Cease working against entropy and nature decays it into a forest.
Michaël Defferrard
@m_deffScientist. ML and (computational) graphs at @Qualcomm AI Research. Previously @EPFL_en (PhD with @trekkinglemon), @BerkeleyLab.
- The supremacy of search as a paradigm—and the associated attention-selling business—is coming to an end. Virtual assistants will disrupt search like search disrupted web portals.
- The essential nature of convolutions to space is the backbone of the thesis I'll defend today. My contribution: generalized convolutions. They enable parameter sharing for non-transitive and unknown symmetry groups to efficiently learn on arbitrary domains. Looking forward!Took me years; not done either. Convolutions are essential to space: They are the linear operators that commute with symmetries (equivariance). Not only a tool that works™. They are basic: built on a space's topology (incidence structure) and geometry (inner product). No more.
- Took me years; not done either. Convolutions are essential to space: They are the linear operators that commute with symmetries (equivariance). Not only a tool that works™. They are basic: built on a space's topology (incidence structure) and geometry (inner product). No more.It is amazing how long it took me to grasp the versatility and richness of basic convolutions. And I am probably not done.
- Yet another impactful achievement of graph ML! 🎉Positive phase 1 trial of cancer vaccine developed with graph-based ML. NEC's neoantigen prediction system using AI methods, such as graph-based relational learning, trained on multiple sources of biological data to discover candidate neoantigen targets. nec.com/en/press/20211…
