Brief

Researchers have developed a new method for matching correlated Vector Autoregressive (VAR) time series databases. The approach introduces a probabilistic framework to recover hidden permutations between two time series, generalizing the problem of matching point clouds to time series. The study derives a maximum likelihood estimator (MLE) and analyzes an estimator based on linear assignment, providing recovery guarantees based on noise thresholds. Additionally, the paper proposes using convex relaxations of permutation matrices, such as the Birkhoff polytope, to efficiently estimate parameters via alternating minimization, with empirical results showing linear assignment often performs comparably or better than MLE relaxation methods. AI

Researchers have introduced KromHC, a novel method for improving neural network training stability and scalability. KromHC addresses limitations in existing Hyper-Connections (HC) by using Kronecker products of smaller doubly stochastic matrices to parameterize the residual matrix. This approach guarantees exact double stochasticity while significantly reducing the number of trainable parameters compared to previous methods. Experiments demonstrate that KromHC performs comparably to or better than state-of-the-art variants with substantially fewer parameters. AI

Researchers have introduced Transportation Birkhoff Polytope (TBP) parameterizations as a novel method for constructing exactly doubly stochastic mixing matrices in hyper-connections. This approach offers full expressivity of the Birkhoff polytope with significantly reduced degrees of freedom compared to previous methods. TBP parameterizations have demonstrated competitive performance in language model pre-training, showing improved stability and scalability. AI