Search

Offline Reinforcement Learning (RL) defines a framework for learning from previously collected static buffer. However, offline RL is prone to approximation errors caused by out-of-distribution (OOD) data and particularly inefficient for pixel-based learning tasks compared with state-based input control methods. Several pioneer efforts have been made to solve this problem; some use pessimistic Q-values approximation for unseen observation while others train a model to simulate the environment to train a model on previously collected data to learn policies. However, these methods require accurate and time-consuming estimation of the Q-values or the environment models. Based on this observation, we present offline RL methods with augmented data (ORAD), a handy but non-trivial extension to offline RL algorithms. We show that simple data augmentations, e.g. random translation and random crop, significantly elevate the performance of the state-of-the-art offline RL algorithms. Besides, we find that regularization of the Q-values can also enhance performance. Extensive experiments on the pixel-based input control-Atari demonstrate the superiority of ORAD over SOTA offline RL methods considering both performance and data efficiency, and reveal that ORAD is more effective for the pixel-based control. [ABSTRACT FROM AUTHOR]

Generalized entropy is a flexible risk measure to characterize indeterminacy of uncertain random variables. As important roles of risk measures in finance, this paper proposes the concept of generalized entropy for uncertain random variables. Also, an approach for computing generalized entropy is derived based on inverse uncertainty distributions. As an application in finance, portfolio selection problems of uncertain random returns are solved by minimizing generalized entropy in mean-entropy models. Finally, for better understanding, some examples and tables are provided. [ABSTRACT FROM AUTHOR]

Heterogeneous graph representation learning is designed to learn meaningful representation vectors from heterogeneous networks in few dimensions to extract the structure and features of the attributes of these networks. The embedding vector is the basis of and crucial to complex network analysis, and can be used in such downstream tasks as classification, clustering, link prediction, and recommendation. Key issues in heterogeneous graph neural networks pertain to ways to define heterogeneous neighbors and ways to aggregate them. Although considerable research has been devoted to homogeneous and heterogeneous network representation, the effective combination of information on the network structure and the attributes of nodes, especially effective use of meta-paths containing specific semantic information, remains rare. Here a meta-path-based heterogeneous graph neural network model is proposed. The meta-path is applied to sample the heterogeneous neighbors of each node in the network, and aggregate features of the same types of nodes to form type-related embedding. A multi-head attention mechanism is then applied to aggregate information on neighbors of different types of nodes and the model is trained by reducing context loss. Experiments on classification, clustering, link prediction, and recommendation tasks verified the validity of this model, which significantly improved the results of baseline methods. [ABSTRACT FROM AUTHOR]

Plectasin is a promising and potent antimicrobial peptide isolated from the fungus Pseudoplectania nigrella which has been heterologously expressed in various hosts. In this study, a four-copy cassette of plectasin was constructed via 2A peptide assembly to further increase its expression level in recombinant Pichia pastoris. The yeast transformant 4Ple-61 harboring four-copy cassette of plectasin could secrete 183.2 mg/L total protein containing 60.8% of plectasin at the flask level within 120 h, which was 2.3 times higher than that of the yeast transformant Ple-6 carrying one-copy cassette of plectasin. Western blot confirmed the significant peptide expression level in the transformant 4Ple-61. Furthermore, it yielded as high as 426.3 mg/L total protein within 120 h during a 5-L fermentation. The purified plectasin shows superior stability and good antimicrobial activity against conventional Staphylococcus aureus ATCC 26,001 and some food-borne antibiotic-resistant S. aureus strains with the MICs ranging from 8 to 32 μg/mL. Therefore, the strategy based on 2A peptide assembly can enhance the expression of plectasin and further expand its application prospect. Key points: • A yeast transformant 4Ple-61 with four-copy cassette of plectasin was constructed. • The plectasin level yield by the transformant 4Ple-61 was boosted by 2.3 times. • The plectasin showed good activity against food-borne antibiotic-resistant S. aureus. [ABSTRACT FROM AUTHOR]

Network embedding technology transforms network structure into node vectors, which reduces the complexity of representation and can be effectively applied to tasks such as classification, network reconstruction and link prediction. The main concern of network embedding is to keep the local structural features while effectively capturing the global features of the network. The "shallow" network representation models cannot capture the deep nonlinear features of the network, and the generated network embedding is usually not the optimal solution. In this paper, a new graph autoencoder-based network representation model combines the first- and second-order proximity to evaluate the performance of network embedding. Aiming at the shortcomings of existing network representation methods in weighted and directed networks, on one hand, the concepts of receiving vector and sending vector are introduced with a simplification of decoding part of the neural network which reduces computation complexity; on the other hand, a measurement index based on node degree is proposed to better emphasize the weighted information in the application of network representation. Experiments including directed weighted networks and undirected unweighted networks show that the proposed method achieves better results than the baseline methods for network reconstruction and link prediction tasks and is of higher computation efficiency than previous graph autoencoder algorithms. Besides, the proposed weighted index is able to improve performances of all baseline methods as an external assistance. [ABSTRACT FROM AUTHOR]

Pangolins are threatened placental mammals distributed in Africa and Asia. Many efforts have been undertaken in the last century to maintain pangolins in captivity, but only a few of them succeeded in maintaining and keeping this species in a controlled environment. This study reports the first systematic breeding of the Critically Endangered Malayan pangolin (Manis javanica) in captivity. Our captive breeding approach successfully improved the reproductive rate for both wild and captive-born female pangolins. From 2016 to 2020, we had 33 wild pangolins and produced 49 captive-born offspring spanning three filial generations. The female offspring further bred 18 offspring, of which 14 (78%) were conceived during the first time of cohabitation with males, and four offspring were conceived during the second cohabitation event, suggesting that they may practice copulation-induced ovulation. We observed that captive-born female pangolins could reach sexual maturity at 7–9 months (n = 4), and male pangolins could mate and successfully fertilise females at nine months age (n = 1). We also observed a female pangolin conceiving on the eighth day after parturition (the fifth day after the death of its pup). Our captive pangolins had a female-biased sex ratio of 1:0.5 at birth, unlike other known captive-born mammals. Also, captive-born pangolins were generally more viable after successful weaning and had a similar gestation length (~185 days) to wild pangolins. Most importantly, we report the first self-sustaining captive population of Malayan pangolins, and this species has an efficient reproduction strategy. These advances provide more comprehensive information for people to understand pangolins, and have implications for conserving endangered Malayan pangolins and providing scientific guidance to the management of other pangolin species. Yan et al. report on a successful captive breeding program for Critically Endangered Malayan pangolins, Manis javanica. Breeding pangolins in captivity is historically challenging and unsuccessful, highlighting the importance of this self-sustaining population. [ABSTRACT FROM AUTHOR]

Knot-free timber cultivation is an important goal of forest breeding, and lateral shoots affect yield and stem shape of tree. The purpose of this study was to analyze the molecular mechanism of lateral bud development by removing the apical dominance of Pinus massoniana young seedlings through transcriptome sequencing and identify key genes involved in lateral bud development. We analyzed hormone contents and transcriptome data for removal of apical dominant of lateral buds as well as apical and lateral buds of normal development ones. Data were analyzed using an comprehensive approach of pathway- and gene-set enrichment analysis, Mapman visualization tool, and gene expression analysis. Our results showed that the contents of auxin (IAA), Zea and strigolactone (SL) in lateral buds significantly increased after removal of apical dominance, while abscisic acid (ABA) decreased. Gibberellin (GA) metabolism, cytokinin (CK), jasmonic acid, zeatin pathway-related genes positively regulated lateral bud development, ABA metabolism-related genes basically negatively regulated lateral bud differentiation, auxin, ethylene, SLs were positive and negative regulation, while only A small number of genes of SA and BRASSINOSTEROID, such as TGA and TCH4, were involved in lateral bud development. In addition, it was speculated that transcription factors such as WRKY, TCP, MYB, HSP, AuxIAA, and AP2 played important roles in the development of lateral buds. In summary, our results provided a better understanding of lateral bud differentiation and lateral shoot formation of P. massoniana from transcriptome level. It provided a basis for molecular characteristics of side branch formation of other timber forests, and contributed to knot-free breeding of forest trees. [ABSTRACT FROM AUTHOR]

Photo-controlled release hydrogel provides a new strategy for treating tumours. Under the stimulation of external light sources, the ability to release the entrapped drug on time and space on demand has outstanding advantages in improving drug utilisation, optimising treatment, and reducing toxicity and side effects. In this study, a photo-controlled drug delivery system for disulphide cross-linked polyaspartic acid (PASP-SS) hydrogels encapsulating proteinase K (ProK) adsorbed with platinum nanoparticles (PtNPs) was designed. The injectable cysteamine-modified polyaspartic acid (PASP-SH) sol and PtNPs adsorbed by ProK (ProK-PtNPs) as regulatory factors were prepared. Then, ProK-PtNPs and lentinan were dissolved in the sol, and the oxidant was added to the matrix to form the gel in situ quickly after injection. Finally, the degradation of PASP-SS hydrogel by ProK and the controllability of drug release under near-infrared (NIR) light irradiation were elucidated. In vitro degradation of hydrogels and drug release experiments showed that the degradation rate of PASP-SS hydrogel significantly increased and the drug release rate increased significantly under near-infrared radiation. The results of cytotoxicity test showed that PASP-SS, ProK-PtNPs, and lentinan all had more than 90% cell survival rate on NIH3T3, and the lentinan released from the carrier obviously inhibited the proliferation of MCF7. PASP hydrogel has the potential to respond to on-demand light control. [ABSTRACT FROM AUTHOR]