Search for: All records

Abstract Reinforcement learning (RL) systems can be complex and non-interpretable, making it challenging for non-AI experts to understand or intervene in their decisions. This is due in part to the sequential nature of RL in which actions are chosen because of their likelihood of obtaining future rewards. However, RL agents discard the qualitative features of their training, making it difficult to recover user-understandable information for “why” an action is chosen. We propose a techniqueExperiential Explanationsto generate counterfactual explanations by traininginfluence predictorsalong with the RL policy. Influence predictors are models that learn how different sources of reward affect the agent in different states, thus restoring information about how the policy reflects the environment. Two human evaluation studies revealed that participants presented with Experiential Explanations were better able to correctly guess what an agent would do than those presented with other standard types of explanation. Participants also found that Experiential Explanations are more understandable, satisfying, complete, useful, and accurate. Qualitative analysis provides information on the factors of Experiential Explanations that are most useful and the desired characteristics that participants seek from the explanations. 

We experimentally demonstrate slow light photonic crystal waveguide (PCW) and subwavelength waveguide (SWWG) loop terminated Mach-Zehnder interferometer (LT-MZI) sensors in a foundry-fabricated silicon-on-insulator (SOI) platform. We compare the experimental results on sensitivity and limit of detection (LOD) on the interferometer sensors with microcavity-type sensors. We show experimentally that 2-D PCW interferometers have higher phase sensitivities than SWWGs of the same length. Based on experimental results, 20- μ m-long 2-D PCW LT-MZI sensors and 200- μ m-long SWWG LT-MZI sensors achieve an LOD of 3.4×10−4 and 2.3×10−4 RIU, respectively, with nearly the same insertion losses in foundry-fabricated devices. We show that by considering the various sources of loss in our benchtop fiber-to-fiber photonic integrated circuit measurement system, it will be possible to reach 10−7 LOD in both slow light PCW and SWWG-based LT-MZI sensors with on-chip integrated light sources and detectors. We show via simulations and experiment that the LOD of a 20- μ m-long slow light PCW LT-MZI is equivalent to that of a 100- μ m-long SWWG LT-MZI, thus enabling more compact LT_MZI sensors when using slow light PCWs versus SWWGs 

We experimentally demonstrated slow wave enhanced phase and spectral sensitivity in asymmetric Michelson interferometer sensors with a phase sensitivity 277,750 rad/RIU-cm and theoretical phase sensitivity as high as 461,810 rad/RIU-cm. In the context of low-cost chip integrated photonic packaged sensors, in this paper we will experimentally demonstrate a method for active tuning of interferometer fringes using phase change materials that will potentially overcome fabrication induced variation of interference fringe wavelengths, thus allowing sensor chip packaging with a fixed wavelength laser and available integrated photodetectors.