<ul><li>A study on the complexity of learning real-valued Multi-Index Models (MIMs) under the Gaussian distribution was conducted.</li><li>An algorithm for PAC learning a broad class of MIMs with respect to the square loss, even in the presence of adversarial label noise, was introduced.</li><li>A nearly matching Statistical Query (SQ) lower bound was established, suggesting the optimality of the algorithm's complexity regarding the dimension.</li><li>An efficient learner for the class of positive-homogeneous $L$-Lipschitz $K$-MIMs was developed, providing a new PAC learning algorithm for Lipschitz homogeneous ReLU networks with improved complexity.</li></ul>

Algorithms and SQ Lower Bounds for Robustly Learning Real-valued Multi-index Models

Discover more