LqgOpt learns optimal control in unknown LQG systems with minimal regret.
problem Adaptive control in partially observable linear quadratic Gaussian systems with unknown dynamics.
method Optimism in the face of uncertainty, predictor state evolution, closed-loop system identification, confidence bounds.
result Proves a regret upper bound of i l d e O ( T ) ilde{\mathcal{O}}(\sqrt{T}) i l d e O ( T ) for LQG systems. Study cost-driven state representation learning for control from partial observations.
problem Learning state representation for control from partial and high-dimensional observations.
method Cost-driven state representation learning via predicting cumulative costs.
result Established finite-sample guarantees for near-optimal representation and controller.
Study learns state representations from observations for control, proving guarantees.
problem Learning state representations from high-dimensional observations for control.
method Cost-driven approach, learning latent state model to predict costs.
result Proves finite-sample guarantees for near-optimal state representation and controller.
Study capacity constraints in continual learning with a simple model.
problem Understanding optimal resource allocation for agents with limited memory and compute resources.
method Analyzes a capacity-constrained linear-quadratic-Gaussian (LQG) sequential prediction problem and demonstrates optimal capacity allocation strategies.
result Derives a solution to the capacity-constrained LQG sequential prediction problem and shows how to optimally allocate capacity across sub-problems in the steady state.
Paper tackles sim-to-real transfer in continuous domains with partial observations.
problem Lack of theoretical foundation for sim-to-real transfer in continuous domains with partial observations.
method Developed a new algorithm for infinite-horizon average-cost LQGs and established a regret bound.
result A popular robust adversarial training algorithm can learn competitive policies from simulation to real-world environments.
The study sets limits on how well systems can be controlled adaptively.
problem Learning to control unknown linear Gaussian systems with quadratic costs.
method Combining ideas from experiment design, estimation theory, and perturbation bounds of information matrices.
result Regret lower bounds of the order of T \sqrt{T} T in the time horizon T T T accurately capture control-theoretic parameters. We study the performance of the certainty equivalent controller on Linear Quadratic (LQ) control problems with unknown transition dynamics. We show that for both the fully and partially observed settings, the sub-optimality gap between the cost incurred by playing the certainty equivalent controller on the true system …
Non-bilinear observations make optimal control harder, showing non-convex costs and non-affine optimal controllers.
problem Optimal control from bilinear observations in linear systems is challenging.
method Analytical and numerical methods to study the non-convex cost-to-go and non-affine optimal controllers.
result The Separation Principle does not hold for bilinear observations, leading to non-convex costs and non-affine optimal controllers.
Reinforcement learning for continuous-time risk-sensitive asset allocation
problem Continuous-time risk-sensitive asset allocation
method Free energy-entropy duality reformulation and q q q -learning actor-critic method result Optimal policy learning with high accuracy
Paper introduces a new method for risk-sensitive investment management using RL.
problem Risk-sensitive portfolio management with unknown model parameters.
method Combines RL and risk-sensitive stochastic control with Gaussian perturbations for exploration.
result Endogenous relative-entropy regularization and optimal investment strategy derived.
Survey of theoretical foundations for policy optimization in control.
problem Understanding the theoretical properties of gradient-based methods in control and reinforcement learning.
method Interdisciplinary review of optimization landscape, convergence, and sample complexity for various control problems.
result Recent theoretical results on stability and robustness in learning-based control.
The trade-off between the cost of acquiring and processing data, and uncertainty due to a lack of data is fundamental in machine learning. A basic instance of this trade-off is the problem of deciding when to make noisy and costly observations of a discrete-time Gaussian random walk, so as to minimise the posterior var…
New framework learns policies for partially observable systems.
problem Learning policies in partially observable dynamical systems.
method Partially Observable Bilinear Actor-Critic framework.
result Algorithm can learn against optimal policies in certain cases.
Solves risk-sensitive investment via duality, entropic regularization, and RL.
problem Risk-sensitive portfolio management in a factor-based setting.
method Free energy-entropy duality, Kuroda-Nagai change-of-measure, RL algorithm.
result Direct analytical solution, explicit controls, two interpretations of optimal allocation.
Policy gradient methods are very attractive in reinforcement learning due to their model-free nature and convergence guarantees. These methods, however, suffer from high variance in gradient estimation, resulting in poor sample efficiency. To mitigate this issue, a number of variance-reduction approaches have been prop…
Study optimizes interbank lending and borrowing to reduce systemic risk.
problem Optimizing lending and borrowing in interbank markets to mitigate systemic risk.
method Risk-sensitive mean field games with common noise, convex analysis, Fokker-Planck equations, first hitting time method.
result Risk-averse behavior reduces individual and systemic bank risks.
New method infers human sensorimotor costs from behavior.
problem Inferring human sensorimotor costs from observed behavior.
method Inverse optimal control with signal-dependent noise.
result Recovering costs and benefits in sensorimotor behavior.
Survey explores geometric aspects of policy optimization in control systems.
problem Understanding the geometric relationships between control design and optimization.
method Geometric perspective on policy optimization, focusing on parameterization and topology.
result Implications of policy geometry on stability and performance of local search algorithms.
AdaptOn achieves logarithmic regret in adaptive control of unknown partially observable linear systems.
problem Adaptive control in partially observable linear dynamical systems.
method AdaptOn algorithm that estimates system dynamics through online learning and gradient descent.
result AdaptOn achieves a logarithmic regret bound of polylog(T) after T steps.
New controller reduces regret in non-stochastic control with adversarial perturbations.
problem Non-stochastic control with adversarial perturbations and partially observed states.
method Denoised observations and online gradient descent.
result Sublinear regret bounds, optimal for known and unknown systems.
Paper optimizes WGAN parameters for non-Gaussian data.
problem Optimizing parameters for non-Gaussian data in WGAN.
method Characterization of optimal solutions for population WGAN beyond LQG setting, using sliced Wasserstein framework.
result Closed-form optimal parameters for non-linear activation functions and non-Gaussian data derived.
Develops variational framework for LQG risk-sensitive MFGs with major-minor interactions.
problem Risk-sensitive optimal control in LQG systems with major-minor interactions.
method Variational approach, nonlinear necessary and sufficient condition of optimality, equivalent risk-neutral measure, Markovian closed-loop best-response strategies.
result Derives optimal control strategies for LQG risk-sensitive MFGs with major-minor interactions, establishing Nash and ε \varepsilon ε -Nash equilibria. MF-PID uses interacting samples to efficiently transport probability mass.
problem Efficiently transporting probability mass in generative models.
method Introducing Mean-Field Path-Integral Diffusion (MF-PID) where samples become interacting agents.
result MF-PID achieves 19-24% reductions in control energy for demand-response control of energy systems.
Optimal liquidation strategy reduces risk and improves performance.
problem Risk-averse trader liquidating assets in volatile markets.
method Conditional Value-at-Risk (CVaR) objective, dynamic trading strategies, closed-form solutions.
result Dynamic policies outperform static and VWAP policies by 15-25%.