Efficient Online Learning of Optimal Rankings: Dimensionality Reduction via Gradient Descent
NIPS 2020, 2020.
We consider a natural model of online preference aggregation, where sets of preferred items $R_1, R_2, \ldots, R_t$ along with a demand for $k_t$ items in each $R_t$, appear online. Without prior knowledge of $(R_t, k_t)$, the learner maintains a ranking $\pi_t$ aiming that at least $k_t$ items from $R_t$ appear high in $\pi_t$. This is...More
PPT (Upload PPT)