Search papers, labs, and topics across Lattice.
This paper investigates the separation capacity of scattering networks specifically designed for low-dimensional datasets, focusing on architectures that utilize fixed monomial nonlinearities without pooling. The authors establish a theoretical framework that characterizes and bounds the separation capacity of feature extractors based on the geometric properties of rectifiable sets. Key findings reveal that optimal filter design requires sufficient frequency coverage and well-conditioned coupling matrices to enhance performance in data representation.
Scattering networks can achieve optimal separation capacity by strategically tuning filter frequencies and ensuring well-conditioned geometric couplings.
We aim to identify scattering network architectures that maximize the separation capacity on data with low intrinsic dimension. The networks we consider employ a fixed monomial nonlinearity and no pooling, so that the only design variable is the frame generated by the network filters. For data modeled as rectifiable sets, we first characterize and bound the separation capacity of general feature extractors in terms of the geometry of the dataset. We then particularize to scattering networks and obtain two design criteria: (i) the filters should meet the data on sufficiently many frequencies, and (ii) the matrices coupling the frame to the geometry of the data should be well-conditioned.
ETH