科研成果

Figure 1  shows humanists’ academic activities and procedures in traditional Chinese studies steps as follows: (a) Forefathers recorded faithfully what had happened around them, what they thought, talked about and wrote down originally. Some of their records nowadays we call “historical materials”, and others we call philosophy, fiction, prose, drama, poetry and so on. (b) With passage of time, some records have been passing down, while others went missing or were buried with nobles as burial objects. (b.1) For the burial objects underground, archeologists dug them out, classified those objects , recognized the ancient characters on them, sorted them chronologically，compiled and published them as primary literatures. (b.2) For the records passing down: (b.2.1) Some deliberate or unintentional mistakes were made on them for unknown reasons. Different versions were produced and remained to be distinguished and corrected by younger generation. (b.2.2) Authors died, their thoughts and analects became hard to understand. Many activities conducted by younger generations: ·they tried to translate, interpret, comment, make notes in their own way based on their own knowledge, and accordingly some of them formed their academic circles or factions; ·some of them didn’t know which one was right, so they attempted to argue, debate, correct, or just collected their forefathers’ interpretations, comments, notes, and compiled them into a book; ·others of them proofread, annotated, compiled, rewrote, or excerpted according to their own requirements and purposes. (b.2.3) During such procedures, some people recorded faithfully what had happened around them, what they thought, talked about and wrote down originally, and also some of the literatures went missing or were buried underground. (c) With time going on, above academic activities carries out repetitively while outputs are different.

Feature-based time series representations have attracted substantial attention in a wide range of time series analysis methods. Recently, the use of time series features for forecast model averaging has been an emerging research focus in the forecasting community. Nonetheless, most of the existing approaches depend on the manual choice of an appropriate set of features. Exploiting machine learning methods to extract features from time series automatically becomes crucial in state-of-the-art time series analysis. In this paper, we introduce an automated approach to extract time series features based on time series imaging. We first transform time series into recurrence plots, from which local features can be extracted using computer vision algorithms. The extracted features are used for forecast model averaging. Our experiments show that forecasting based on automatically extracted features, with less human intervention and a more comprehensive view of the raw time series data, yields highly comparable performances with the best methods in the largest forecasting competition dataset (M4) and outperforms the top methods in the Tourism forecasting competition dataset.

Lead-based organic-inorganic hybrid perovskite materials are widely used in optoelectronic devices due to their excellent photophysical properties. However, the main issues which hinder its commercialization are the toxicity caused by lead and the intrinsic instability of the material. Recently, many lead-free halide materials with good intrinsic stability have been reported, among which bismuth-based halide materials have attracted extensive research due to their structure and promising optoelectronic properties. In this review, bismuth-based materials are divided into binary BiX3 (X = I, Br, Cl), ternary A(a)Bi(b)X(a)(+3)(b) (A = Cs, Rb, MA, Ag, etc.), and quaternary A(2)AgBiX(6) (A = Cs, Rb, MA, etc.) according to its elemental composition. The structure and optoelectronic properties of bismuth-based halide materials, which may be helpful for the development of bismuth-based halide materials and lead-free perovskites in the future, are summarized and highlighted.

Automatic transistor sizing is a challenging problem in circuit design due to the large design space, complex performance tradeoffs, and fast technology advancements. Although there have been plenty of work on transistor sizing targeting on one circuit, limited research has been done on transferring the knowledge from one circuit to another to reduce the re-design overhead. In this paper, we present GCN-RL Circuit Designer, leveraging reinforcement learning (RL) to transfer the knowledge between different technology nodes and topologies. Moreover, inspired by the simple fact that circuit is a graph, we learn on the circuit topology representation with graph convolutional neural networks (GCN). The GCN-RL agent extracts features of the topology graph whose vertices are transistors, edges are wires. Our learning-based optimization consistently achieves the highest Figures of Merit (FoM) on four different circuits compared with conventional black box optimization methods (Bayesian Optimization, Evolutionary Algorithms), random search and human expert designs. Experiments on transfer learning between five technology nodes and two circuit topologies demonstrate that RL with transfer learning can achieve much higher FoMs than methods without knowledge transfer. Our transferable optimization method makes transistor sizing and design porting more effective and efficient.

High-efficiency perovskite solar cells (PSCs) have experienced rapid development and attracted significant attention in recent years. The PSCs based on doctor bladed or slot-die coated perovskite films usually have lower power conversion efficiency (PCE) than that based on spin-coated perovskite films. In this work, we have developed an effective method, called glass rod-sliding and low pressure assisted solution processing composition engineering (GRS-LPASP), to manufacture high quality perovskite film in air. GRS-LPASP composition engineering effectively increases the grain size and thickness of perovskite films and reduces the defect density by increasing the contact area between the perovskite layer and the hole transport layer, thus leading an increased current density (Jsc) of perovskite solar cells. The device with GRS-LPASP composition engineering achieves a maximum PCE of 19.78%. The experimental results demonstrates that GRS-LPASP composition engineering is a feasible method to prepare high-efficiency PSCs. Moreover, GRS-LPASP composition engineering also provides a potential approach for the commercial production of PSCs.