HIGHLIGHTS
SUMMARY
Flexibility requires thinking ahead, which naturally suggests simulation-based solutions (Kita et_al, 2014; Li et_al, 2015). The use of FMNNs for control tasks has been explored on various domains (Ebert et_al, 2018; Henaff et_al, 2017; Janner et_al, 2019; Lesort et_al, 2018; Wahlström et_al, 2015), albeit in most cases on tasks of lower dimensionality than CM requires. A FMNN can replace the task environment in model-free RL to improve sample efficiency (Henaff et_al, 2017; Wahlström et_al, 2015), or be used as part of a model-based RL algorithm (Clavera et_al, 2020). The . . .
If you want to have access to all the content you need to log in!
Thanks :)
If you don't have an account, you can create one here.