@randall_balestr@ylecun Great work!If I understand correctly, using multiple descriptions of the same scene as positive pairs allows smaller models to achieve better crossModalalignment. Is the main challenge obtaining enough of these highquality pairs, which is why we rely on scaling up models instead?
๐๐ฟ๐ฒ ๐๐ผ๐ ๐ฎ ๐ฟ๐ฒ๐๐ฒ๐ฎ๐ฟ๐ฐ๐ต๐ฒ๐ฟ ๐ถ๐ป ๐ฝ๐ฟ๐ผ๐ฐ๐ฒ๐๐ ๐บ๐ถ๐ป๐ถ๐ป๐ด, ๐บ๐ฎ๐ฐ๐ต๐ถ๐ป๐ฒ ๐น๐ฒ๐ฎ๐ฟ๐ป๐ถ๐ป๐ด, ๐ผ๐ฟ ๐ฑ๐ฎ๐๐ฎ ๐บ๐ถ๐ป๐ถ๐ป๐ด looking for a
healthcare dataset? ๐ ๐๐ ๐๐-๐๐ฉ-๐๐ ๐-๐๐๐๐ is now publicly available on PhysioNet! https://t.co/SfKbXqpeHT
Happy New Year to all of you! We start 2022 with two contributions from @PAGroupTUe at the ProcessMining in #Healthcare HICSS Minitrack. A paper by @Renata_MdeC and a paper with @fmannhardt as contributor.
Unfortunately, both without an in-person presentation in Hawaii.
#icpm21 is kicked-off with an early Sunday morning start of the Doctoral Consortium. More than 20 participants joined, among them nine PhD researchers that get to pitch their research plan to the community. Thanks to @MiekeJans and @FabrizioMaggi to organizing the DC!
Identifying valuable insights by using a multi-entity event data for care pathways of multi-morbid patients! https://t.co/4qRRXkEPLL. By supervision of @fmannhardt & Pieter Toussaint and representation of multiple entities in event logs as event graphs by @DblQlik & @dfahland.