Can JL lemma be a replacement for Spectral embedding?

[igaloly]

Because of the lack of ability of computing eigenvalues/vector efficiently in js, and with the guarantees of the JL lemma and the increased speed of its "random projection", might it be worth using the JL lemma instead of Spectral embedding?