Inverse statistical problems: from the inverse Ising problem to data science

摘要: Inverse problems in statistical physics are motivated by the challenges of `big data' different fields, particular high-throughput experiments biology. In inverse problems, usual procedure needs to be reversed: Instead calculating observables on basis model parameters, we seek infer parameters a based observations. this review, focus Ising problem and closely related namely how coupling strengths between spins given observed spin correlations, magnetisations, or other data. We review applications problem, including reconstruction neural connections, protein structure determination, inference gene regulatory networks. For equilibrium, number controlled uncontrolled approximate solutions have been developed mechanics community. A particularly strong method, pseudolikelihood, stems from statistics. also non-equilibrium case, where must reconstructed