Experimental Design Using Interlacing Polynomials

Experimental design is a classical problem in statistics [ Puk06 ], which recently found wide applications from machine learning (e.g., active learning, feature selection, data summ arization) to numerical linear algebra (e.g., column subset selection, sparse least squares regression) t o graph algorithms (e.g., total effective resistance minimization, algebraic connectivity maximization). We ref er the reader to [ SX20, AZLSW21, NST22, LZ22b, LZ22a, LWZ23 ] and the references therein for additional background and related applications.