Residual Policy Learning (RPL) google
We present Residual Policy Learning (RPL): a simple method for improving nondifferentiable policies using model-free deep reinforcement learning. RPL thrives in complex robotic manipulation tasks where good but imperfect controllers are available. In these tasks, reinforcement learning from scratch remains data-inefficient or intractable, but learning a residual on top of the initial controller can yield substantial improvement. We study RPL in five challenging MuJoCo tasks involving partial observability, sensor noise, model misspecification, and controller miscalibration. By combining learning with control algorithms, RPL can perform long-horizon, sparse-reward tasks for which reinforcement learning alone fails. Moreover, we find that RPL consistently and substantially improves on the initial controllers. We argue that RPL is a promising approach for combining the complementary strengths of deep reinforcement learning and robotic control, pushing the boundaries of what either can achieve independently. …

Named-Entity Linking (NEL) google
In the legal domain it is important to differentiate between words in general, and afterwards to link the occurrences of the same entities. The topic to solve these challenges is called Named-Entity Linking (NEL). Current supervised neural networks designed for NEL use publicly available datasets for training and testing. However, this paper focuses especially on the aspect of applying transfer learning approach using networks trained for NEL to legal documents. Experiments show consistent improvement in the legal datasets that were created from the European Union law in the scope of this research. Using transfer learning approach, we reached F1-score of 98.90\% and 98.01\% on the legal small and large test dataset. …

Rainforest Plots google
Research has shown that forest plots are a gold standard in the visualization of meta-analytic results. However, research on the general interpretation of forest plots and the role of researchers’ meta-analysis experience and field of study is still unavailable. Additionally, the traditional display of effect sizes, confidence intervals, and weights have repeatedly been criticized. The current work presents an online statistical cognition experiment in which a total of 279 researchers with experience in meta-analysis from 36 countries evaluated conventional forest plots and two novel versions of forest plots, namely, thick forest plots and rainforest plots. The results indicate certain biases in the interpretation of forest plots, especially with regard to heterogeneity, the distribution of weights, and the theoretical concept of confidence intervals. Although the two novel displays (thick forest plots and rainforest plots) are associated with slightly longer viewing times, they are at least as well-suited and esthetically and perceptively pleasing as the conventional displays while facilitating the correct and exhaustive interpretation of the meta-analytic information. Furthermore, it is advisable to combine conventional forest plots with distribution information of the individual effects, make confidence lines more visually striking, and to display a background grid in the graph. …

Advertisements