We investigate the mapping of sensor-based sub-symbolic data onto abstract symbols with the goal to create a binding between the sub-symbolic and the symbolic level in a hierarchical information processing architecture. The approach is envisioned as an iterative procedure, where the sub-symbolic processing level produces hypotheses for symbols, which may qualify for semantic meaningful symbols. The symbolic processing level starts to reason and predict using these hypotheses. Both levels exchange information iteratively to either reinforce a given symbol hypothesis or to reject it.
Günther, Johannes, et al. : "Intelligent laser welding through representation, prediction, and control learning: An architecture with deep neural networks and reinforcement learning." Mechatronics 34: 1-11. 2016.
Passenberg, C., Meyer, D., Feldmaier, J., & Shen, H. "Optimal water heater control in smart home environments." In Energy Conference (ENERGYCON), 2016 IEEE International (pp. 1-6). IEEE. 2016.
Günther, J., Shen, H., Diepold, K.: "Neural Networks for fast sensor data processing in Laser Welding" in Jahreskolloquium - Bildverarbeitung in der Automation, Lemgo, Germany. 2015.
Günther, J., Pilarski, P., Helfrich, G., Shen, H., Diepold, K.: "First Steps Towards an Intelligent Laser Welding Architecture Using Deep Neural Networks and Reinforcement Learning," in 2nd International Conference on System-Integrated Intelligence: Challenges for Product and Production Engineering, pp. 474 - 483. 2014.
Meyer, D., Degenne, R., Omrane, A., & Shen, H. "Accelerated gradient temporal difference learning algorithms." In 2014 IEEE Symposium on Adaptive Dynamic Programming and Reinforcement Learning (ADPRL) (pp. 1-8). IEEE. 2014.
Meyer, D., Shen, H., & Diepold, K. " l1-regularized gradient temporal-difference learning." In Proceedings of the Tenth European Workshop on Reinforcement Learning. 2012.