I develop a novel method for computing globally accurate solutions to recur- sive macro-epidemic models featuring aggregate uncertainty and a potentially large number of state variables. Compared to the previous literature which either restricts attention to perfect-foresight economies amendable to sequence- space representation or focuses on highly simplified, low dimensional models that could can be analyzed using standard dynamic programming and linear projection techniques, I develop a deep learning-based algorithm that can han- dle rich environments featuring both aggregate uncertainty and large numbers of state variables. In addition to solving for particular model equilibria, I show how the deep learning method could be extended to solve for a whole set of models, indexed by the parameters of government policy. By pre-computing the whole equilibrium set, my deep learning method greatly simplifies compu- tation of optimal policies, since it bypasses the need to re-solve the model for many different values of policy parameters. 1
| Identifer | oai:union.ndltd.org:nusl.cz/oai:invenio.nusl.cz:453160 |
| Date | January 2021 |
| Creators | Žemlička, Jan |
| Contributors | Slavík, Ctirad, Kapička, Marek |
| Source Sets | Czech ETDs |
| Language | English |
| Detected Language | English |
| Type | info:eu-repo/semantics/masterThesis |
| Rights | info:eu-repo/semantics/restrictedAccess |
Page generated in 0.0019 seconds