I address one criticism that has been leveled at subregular analyses of phonology, which is that the presence of optional phonological processes implies that phonological processes are non-deterministic, contra the hypothesis that phonological processes, at least for segmental phonology, are deterministic (Heinz and Lai 2013, Jardine 2016, McCollum et al. 2019). I show how in fact optional processes can be analyzed with deterministic transducers where the outputs are drawn from the monoid of finite languages and concatenated. I then present continuing research, which builds on Jardine et al. (2014) and Beros and de la Higuera (2016), of a learning algorithm which can learn classes of optional processes from examples.