Africa
EfficientLearningofDiscrete-Continuous ComputationGraphs
End-to-end learnable discrete-continuous models are compositional, tend togeneralize better,and are more interpretable.
Sub-LinearMemory: HowtoMakePerformersSLiM
Recent works proposed various linear self-attention mechanisms, scaling only asO(L)for serial computation. We conduct a thorough complexity analysis of Performers,aclass which includes most recent linear Transformer mechanisms.
Data Market Design through Deep Learning
In a setting with multiple buyers, a buyer's expected value for an experiment may
EstimatingRank-OneSpikesfrom Heavy-Tailed NoiseviaSelf-AvoidingWalks
The main question about the spiked Wigner model is: how large should the signal-to-noise ratio ฮป > 0 be in order to achieve constant correlation withx?
Network-to-NetworkRegularization: Enforcing Occam'sRazortoImproveGeneralization
What makes a classifier have the ability to generalize? There have been a lot of important attempts to address this question, but a clear answer is still elusive.
