Stochastic microstructure reconstruction has become an indispensable part of computational materials science, but ongoing developments are specific to particular material systems. In this paper, we address this generality problem by presenting a transfer learning-based approach for microstructure reconstruction and structure-property predictions that is applicable to a wide range of material systems. The proposed approach incorporates an encoder-decoder process and feature-matching optimization using a deep convolutional network. For microstructure reconstruction, model pruning is implemented in order to study the correlation between the microstructural features and hierarchical layers within the deep convolutional network. Knowledge obtained in model pruning is then leveraged in the development of a structure-property predictive model to determine the network architecture and initialization conditions. The generality of the approach is demonstrated numerically for a wide range of material microstructures with geometrical characteristics of varying complexity. Unlike previous approaches that only apply to specific material systems or require a significant amount of prior knowledge in model selection and hyper-parameter tuning, the present approach provides an off-the-shelf solution to handle complex microstructures, and has the potential of expediting the discovery of new materials.

Microstructure characterization and reconstruction have become indispensable parts of computational materials science. The main contribution of this paper is to introduce a general methodology for practical and efficient characterization and reconstruction of stochastic microstructures based on supervised learning. The methodology is general in that it can be applied to a broad range of microstructures (clustered, porous, and anisotropic). By treating the digitized microstructure image as a set of training data, we generically learn the stochastic nature of the microstructure via fitting a supervised learning model to it (we focus on classification trees). The fitted supervised learning model provides an implicit characterization of the joint distribution of the collection of pixel phases in the image. Based on this characterization, we propose two different approaches to efficiently reconstruct any number of statistically equivalent microstructure samples. We test the approach on five examples and show that the spatial dependencies within the microstructures are well preserved, as evaluated via correlation and lineal-path functions. The main advantages of our approach stem from having a compact empirically-learned model that characterizes the stochastic nature of the microstructure, which not only makes reconstruction more computationally efficient than existing methods, but also provides insight into morphological complexity.

In this paper, we develop a convolutional neural network for stance detection in tweets. According to the official results, our system ranks 1 st on subtask B (among 9 teams) and ranks 2 nd on subtask A (among 19 teams) on the twitter test set of SemEval2016 Task 6. The main contribution of our work is as follows. We design a "vote scheme" for prediction instead of predicting when the accuracy of validation set reaches its maximum. Besides, we make some improvement on the specific subtasks. For subtask A, we separate datasets into five sub-datasets according to their targets, and train and test five separate models. For subtask B, we establish a two-class training dataset from the official domain corpus, and then modify the softmax layer to perform three-class classification. Our system can be easily re-implemented and optimized for other related tasks.

No abstract