Search papers, labs, and topics across Lattice.
Graz University of Technology
1
0
3
WTA bottlenecks in multi-task learning aren't just another activation function; they're a surprisingly effective way to force the emergence of disentangled, symbolic representations.