Institution profile

University of Latvia

Academic institutioneurope · lv
Official website
Research library15linked papers
Opportunities0open roles
Selected work

Representative Papers

Interpretable Policy Distillation for Power Grid Topology Control

May 30, 2026

This work addresses the challenges of high inference cost, deployment difficulty, and opaque decision-making in deep reinforcement learning for power grid topology control. The authors propose a stress-focused data collection strategy to train a Proximal Policy Optimization (PPO) teacher model and, for the first time, distill it into interpretable, lightweight agents—specifically decision trees and random forests—targeting high-load critical states. The distilled models not only surpass the original PPO policy in average reward and survival duration while significantly reducing inference overhead, but also maintain highly consistent action outputs, enabling human auditability. Furthermore, the study reveals fundamental differences in feature dependencies between neural policies and tree-based models, achieving a balanced trade-off among performance, real-time responsiveness, and interpretability.

0 citationsRead paper

A sharp interaction-degree threshold for simulating QAOA

May 21, 2026

This work investigates the classical simulability of the Quantum Approximate Optimization Algorithm (QAOA) under varying degrees of interaction graphs to delineate the boundary of its quantum advantage. Focusing on QAOA with two-local cost functions, the study integrates computational complexity theory, quantum sampling analysis, and graph-theoretic techniques to establish a sharp threshold between graph degrees 2 and 3. It demonstrates that for degree-2 graphs and circuit depth O(log n), an n-qubit QAOA instance admits efficient exact classical sampling. In contrast, for degree-3 graphs—even when the cost function is trivially optimizable—approximate sampling would imply a collapse of the polynomial hierarchy to its third level, indicating computational hardness for classical simulation. This result precisely characterizes the complexity boundary for classically simulating QAOA.

0 citationsRead paper
Recent publications

Latest Papers

Interpretable Policy Distillation for Power Grid Topology Control

May 30, 2026

This work addresses the challenges of high inference cost, deployment difficulty, and opaque decision-making in deep reinforcement learning for power grid topology control. The authors propose a stress-focused data collection strategy to train a Proximal Policy Optimization (PPO) teacher model and, for the first time, distill it into interpretable, lightweight agents—specifically decision trees and random forests—targeting high-load critical states. The distilled models not only surpass the original PPO policy in average reward and survival duration while significantly reducing inference overhead, but also maintain highly consistent action outputs, enabling human auditability. Furthermore, the study reveals fundamental differences in feature dependencies between neural policies and tree-based models, achieving a balanced trade-off among performance, real-time responsiveness, and interpretability.

0 citationsRead paper

A sharp interaction-degree threshold for simulating QAOA

May 21, 2026

This work investigates the classical simulability of the Quantum Approximate Optimization Algorithm (QAOA) under varying degrees of interaction graphs to delineate the boundary of its quantum advantage. Focusing on QAOA with two-local cost functions, the study integrates computational complexity theory, quantum sampling analysis, and graph-theoretic techniques to establish a sharp threshold between graph degrees 2 and 3. It demonstrates that for degree-2 graphs and circuit depth O(log n), an n-qubit QAOA instance admits efficient exact classical sampling. In contrast, for degree-3 graphs—even when the cost function is trivially optimizable—approximate sampling would imply a collapse of the polynomial hierarchy to its third level, indicating computational hardness for classical simulation. This result precisely characterizes the complexity boundary for classically simulating QAOA.

0 citationsRead paper