Why Deep Learning Failed on Tables for a Decade - Frank Hutter
Frank Hutter, co-founder of Prior Labs, on why deep learning struggled with tabular data for a decade: tables are messy and heterogeneous, hyped models like TabNet did not generalise to new datasets, and there was no ImageNet of tables.
Proof1 independent outlet
Key points
- The breakthrough came from learning to transfer at the level of patterns across many different tables; TabPFN-3.5 now tops the TabArena benchmark.
Sources (1)
- [1]Why Deep Learning Failed on Tables for a Decade - Frank HutterMachine Learning Street Talk (YouTube) · Sep 29, 09:50 AM
Frank Hutter, co-founder of Prior Labs, on why deep learning struggled with tabular data for a decade: tables are messy and heterogeneous, hyped models like TabNet did not generalise to new datasets, and there was no ImageNet of tables.
The breakthrough came from learning to transfer at the level of patterns across many different tables; TabPFN-3.5 now tops the TabArena benchmark.
Extractive summary: sentences quoted from the sources.
