Learning Fair Canonical Polyadical Decompositions using a Kernel Independence Criterion
This addresses fairness in tensor decompositions for applications like recommendation systems or data analysis, but it is incremental as it builds on existing methods with a new regularization approach.
The paper tackled the problem of learning fair low-rank tensor decompositions by regularizing the Canonical Polyadic Decomposition with the kernel Hilbert-Schmidt independence criterion (KHSIC), showing that a small KHSIC ensures approximate statistical parity and surpassing the state-of-the-art FATR algorithm in balancing fairness and fit on synthetic and real data.
This work proposes to learn fair low-rank tensor decompositions by regularizing the Canonical Polyadic Decomposition factorization with the kernel Hilbert-Schmidt independence criterion (KHSIC). It is shown, theoretically and empirically, that a small KHSIC between a latent factor and the sensitive features guarantees approximate statistical parity. The proposed algorithm surpasses the state-of-the-art algorithm, FATR (Zhu et al., 2018), in controlling the trade-off between fairness and residual fit on synthetic and real data sets.