Deep Reinforcement Learning for Beyond-Conceivable Forecasting
Contact person: Glenn Kristiansen
Keywords: Deep Learning, Reinforcement Learning, Forecasting, Digital Technology, Economics
Research group: Entrepreneurship (ENT)
Department of Informatics
We seek to determine to what extent deep reinforcement learning (DRL) could help regulatory- and corporate decision makers in imagining possible, but not immediately obvious, distant future development and application of digital technologies outlined within regulatory sandboxes. Is it possible for DRL to show currently unknown pathways toward such a discontinuous future? Such pathways would need to be modelled by combing domain knowledge in digital technology, governance mechanisms, and economics. DRL is a machine learning approach allowing for meta learning, meaning that algorithms potentially can self-learn from experience. Implementation of DRL within digital-economic environments is relatively new, with significant promise since the algorithms interact directly with its novel environment. The question raises itself why we would need DRL in the first place? Our ability to imagine alternative future pathways is highly bound by our present everyday lives and previous experiences, which may not represent anything like the outcome from Industry 4.0.
We seek a postdoc who can assist in developing DRL as a method for simulating alternative pathways in the intersection between new digital technologies and possible economic contexts:
- To what extent can DRL operate as a methodological toolkit for modelling pathways for developing new digital technologies for a discontinuous future, i.e. a future that is markedly different from today?
- What kind of visions of the future are constructed conversationally in the regulatory sandbox, and can these be superimposed on DRL to create viable pathways for applying new digital technologies?
- How do we create effective reward functions for short- and long-term decision making within DRL for a digital-economic context?
- How do we balance exploitation and exploration in decision making within DRL (for the RL-agent) for obtaining a desirable digital future that are economically sustainable and in equilibrium?
Mentoring and internship will be offered by a relevant external partner.