On projection based operators in l_p space for exact similarity search
The motivation of this work is given by the problem of exact similarity search in a high-dimensional vector space \(V\), where the exact search suffers from dimensionality. The authors introduce a new adaptive projection that satisfies the 1-Lipschitz property, the optimal projection being discovered during a learning process. Further on, it evaluates the dependency of this adaptive projection for the \(l_p\) norms. The study continues with a description of a projection-based method, motivated by the tree structure, which defines a family of projections for a sequence of subspaces.NEWLINENEWLINENEWLINEThe paper concludes with an evaluation of the adaptive projection against the orthogonal mapping, and a comparison of the \(l_1\), \(l_2\), \(l_4\) and \(l_\infty\) norms.
- Approximate nearest neighbor search for \(\ell_p\)-spaces \((2<p<\infty)\) via embeddings
- Index structures for fast similarity search for real-valued vectors. I
- Approximate nearest neighbor search for \(\ell_{p}\)-spaces \((2 < p < \infty)\) via embeddings
- Finite-dimensional subspaces of \(L_p\) with Lipschitz metric projection
- Indexability, concentration, and VC theory
This page was built for publication: On projection based operators in \(l_{p}\) space for exact similarity search
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2805415)
