1 paper · 1 filter
Joshua Hanson, Maxim Raginsky
We show how continuous-depth neural ODE models can be framed as single-layer, infinite-width nets using the Chen--Fliess series expansion for nonlinear ODEs. In this net, the outpu…