High-performance artificial synaptic devices are key building blocks for developing efficient neuromorphic computing systems. However, the nonlinear and asymmetric weight update of existing devices has restricted their practical applications. Herein, floating gate nonvolatile memory (FG NVM) devices based on two-dimensional (2D) ${\mathrm{Hf}\mathrm{S}}_2$/$h$-BN/FG-graphene heterostructures have been designed and investigated as multifunctional NVM and artificial optoelectronic synapses. Benefiting from the FG architecture, the ${\mathrm{Hf}\mathrm{S}}_2$-based NVM device exhibits competitive performances, such as a high on:off ratio ($>{10}^5$), large memory window (approximately 100 V), excellent charge retention ability ($>{10}^4\mathrm{s}$), and robust durability ($>{10}^3$ cycles). Notably, the artificial optoelectronic synapses based on ${\mathrm{Hf}\mathrm{S}}_2$ FG NVM show an impressive large conductance margin and good linearity, owing to the ultrahigh photoresponsivity and photogain of ${\mathrm{Hf}\mathrm{S}}_2$. The energy consumption of per spike in our artificial synapse is as low as 0.2 pJ. Therefore, a high recognition accuracy up to 91.5% of the artificial neural network on the basis of our ${\mathrm{Hf}\mathrm{S}}_2$-based optoelectronic synapse at the system level has been achieved, which is superior to other reported 2D artificial optoelectronic synapses. This work paves the way forward for all 2D material-based memory for developing efficient optogenetics-inspired neuromorphic computing in brain-inspired intelligent systems.

Aspect-level sentiment classification (ASC), detecting and predicting the sentiment polarity of the given aspecs, has attracted increasing attention in the field of Natural Language Processing (NLP). Recent studies in ASC leveraged the graph based on the dependency tree of the context to incorporate the syntactic information and structure of a sentence for better relation extraction. Some researchers noted that existing methods ignored semantic relations or failed to consider affective dependency information, and then proposed several state-of-art methods tackling the above two limitations. However, these approaches failed to consider both informative relations simultaneously. Therefore, we explore and propose a novel solution based on semantic latent graph and SenticNet to leverage semantic and affective information. Specifically, we build a latent semantic graph based on self-attention networks to parse semantic relations within the contexts. In addition, we utilize affective knowledge from SenticNet to enhance the dependency graphs of sentences. Moreover, we use the gate mechanism to dynamically combine information from both the enhanced dependency graphs and latent semantic graphs. Experimental results on three benchmark datasets illustrate the effectiveness and state-of-the-art performance of our model.

The accuracy of OCR is usually affected by the quality of the input document image and different kinds of marred document images hamper the OCR results. Among these scenarios, the low-resolution image is a common and challenging case. In this paper, we propose the cascaded networks for document image super-resolution. Our model is composed by the Detail-Preserving Networks with small magnification. The loss function with perceptual terms is designed to simultaneously preserve the original patterns and enhance the edge of the characters. These networks are trained with the same architecture and different parameters and then assembled into a pipeline model with a larger magnification. The low-resolution images can upscale gradually by passing through each Detail-Preserving Network until the final high-resolution images. Through extensive experiments on two scanning document image datasets, we demonstrate that the proposed approach outperforms recent state-of-the-art image super-resolution methods, and combining it with standard OCR system lead to signification improvements on the recognition results.