A probabilistic model for the numerical solution of initial value problems

@sam_power_825 Btw.: at some point, you actually get decreasing number of significant digits due to the individual floating-point operation problems. See, e.g., here: https://t.co/KoGZ44Q5dB However, in double precision, that typically shouldn't happen bef

@mschoberml @stochastician This work is super cool: https://t.co/STURMHeGoi