“…Since standard ABC is notoriously slow and requires tuning of some hyperparameters, there has been considerable research in making likelihood-free inference more efficient, using, for example, ideas from Bayesian optimisation and experimental design (Gutmann and Corander, 2016;Järvenpää et al, 2019Järvenpää et al, , 2020, conditional density estimation (Papamakarios and Murray, 2016;Lueckmann et al, 2017;Greenberg et al, 2019), classification (Gutmann et al, 2018), indirect inference (Drovandi et al, 2015), optimisation (Meeds and Welling, 2015;Ikonomov and Gutmann, 2020), and more broadly surrogate modelling with Gaussian processes (Wilkinson, 2014;Meeds and Welling, 2015) and neural networks (Blum and Francois, 2010;Chen and Gutmann, 2019;Papamakarios et al, 2019).…”