NHẬN DẠNG CỬ CHỈ TAY BẰNG RA-ĐA FMCW DỰA TRÊN MẠNG NƠ RON TÍCH CHẬP KẾT NỐI CHÉO

Nghiên cứu này đề xuất sử dụng mạng nơ-ron tích chập kết nối chéo (Cross-CNN: Cross-connection Convolutional Neural Network) để nhận dạng cử chỉ tay dựa trên dữ liệu phân tích phổ Doppler vi mô (micro-Doppler) của ra-đa FMCW (Frequency Modulated Continuous Wave). Thêm vào đó, các mức nhiễu khác nhau được thêm vào dữ liệu để đánh giá đầy đủ hơn đối với mô hình đề xuất. Kết quả thử nghiệm cho thấy, mô hình được huấn luyện với dữ liệu có nhiễu cho chất lượng nhận dạng tốt hơn mô hình được huấn luyện với dữ liệu không có nhiễu. Sau đó, mô hình Cross-CNN được phân tích, đánh giá khi thay đổi các tham số cấu trúc, từ đó chọn ra bộ tham số phù hợp nhất cho bài toán đề ra. Cuối cùng, mô hình Cross-CNN được đánh giá, so sánh với các mô hình hiện có khác khi thực hiện trên cùng một tập dữ liệu. Kết quả là, mạng Cross-CNN cho khả năng nhận dạng cử chỉ tay tốt hơn các mô hình khác nhờ vào các kết nối chéo cho phép sử dụng kết hợp các đặc tính cũ với đặc tính mới trong quá trình huấn luyện mạng.

Reusing Monolingual Pre-Trained Models by Cross-Connecting Seq2seq Models for Machine Translation

This work uses sequence-to-sequence (seq2seq) models pre-trained on monolingual corpora for machine translation. We pre-train two seq2seq models with monolingual corpora for the source and target languages, then combine the encoder of the source language model and the decoder of the target language model, i.e., the cross-connection. We add an intermediate layer between the pre-trained encoder and the decoder to help the mapping of each other since the modules are pre-trained completely independently. These monolingual pre-trained models can work as a multilingual pre-trained model because one model can be cross-connected with another model pre-trained on any other language, while their capacity is not affected by the number of languages. We will demonstrate that our method improves the translation performance significantly over the random baseline. Moreover, we will analyze the appropriate choice of the intermediate layer, the importance of each part of a pre-trained model, and the performance change along with the size of the bitext.

A parallel cross-connection recovery scheme for dual link failure in elastic optical networks

In this paper, we proposed a fast recovery strategy for a dual link failure (DLF) in elastic optical network (EON). The EON is a promising solution to meet the next generation higher bandwidth demand. The survivability of high speed network is very crucial. As the network size increases the probability of the DLF and node failure also increases. Here, we proposed a parallel cross connection backup recovery strategy for DLF in the network. The average bandwidth blocking probability (BBP), bandwidth provisioning ratio (BPR), and recovery time (RT) for our proposed Intermediate node cross-connect backup for shared path protection (INCB-SPP) for ARPANET are 0.38, 2.71, 4.68 ms, and for DPP 0.70, 6.02, 8.71 ms and for SPP 0.40, 2.87, and 16.33 ms respectively. The average BBP, BPR, and RT of INCB-SPP for COST239 are 0.01, 1.71, 3.79 ms and for DPP are 0.39, 3.50, 8.20 ms and SPP are 0.04, 1.75, and 12.47 ms respectively. Hence, the proposed strategy shows lower BBP, fast connection recovery, and BPR when compared with the existing shared path protection (SPP) and dedicated path protection (DPP) approaches. Simulation is performed on ARPANET and COST239 topology networks.

CostNet: A Concise Overpass Spatiotemporal Network for Predictive Learning

Predicting the futures from previous spatiotemporal data remains a challenging topic. There have been many previous works on predictive learning. However, mainstream models suffer from huge memory usage or the gradient vanishing problem. Enlightened by the idea from the resnet, we propose CostNet, a novel recursive neural network (RNN)-based network, which has a horizontal and vertical cross-connection. The core of this network is a concise unit, named Horizon LSTM with a fast gradient transmission channel, which can extract spatial and temporal representations effectively to alleviate the gradient propagation difficulty. In the vertical direction outside of the unit, we add overpass connections from unit output to the bottom layer, which can capture the short-term dynamics to generate precise predictions. Our model achieves better prediction results on moving-mnist and radar datasets than the state-of-the-art models.

Cross-connection structure of concordant semigroups

Cross-connection theory provides the construction of a semigroup from its ideal structure using small categories. A concordant semigroup is an idempotent-connected abundant semigroup whose idempotents generate a regular subsemigroup. We characterize the categories arising from the generalized Green relations in the concordant semigroup as consistent categories and describe their interrelationship using cross-connections. Conversely, given a pair of cross-connected consistent categories, we build a concordant semigroup. We use this correspondence to prove a category equivalence between the category of concordant semigroups and the category of cross-connected consistent categories. In the process, we illustrate how our construction is a generalization of the cross-connection analysis of regular semigroups. We also identify the inductive cancellative category associated with a pair of cross-connected consistent categories.

“THERE IS NO PROGRESS, CHANGE IS ALL WE KNOW.” NOTES ON DUCHAMP’S CONCEPT OF PLASTIC DURATION

Henri Bergson is generally recognized as one of the most influential philosophers in the history of historical avant-gardism. Nevertheless, it has been widely neglected that Bergson’s philosophy also played a crucial role for the radically new concept of art that Marcel Duchamp developed based on his critical attitude towards the avant-gardes. First and foremost, this is apparent in view of Duchamp’s paintings The Passage from Virgin to Bride and Bride of 1912, as they both feature an idea of transition laying the foundation for his Large Glass and associated works. But there is also another cross-connection that one wouldn’t expect at the first glance. As this paper argues, Duchamp paradoxically also draws on Bergson’s ideas with his ready-mades, pointing to that productive interplay of intuition and intellect, which Bergson defined as a vital source for any kind of imagination and agency. Thus, Duchamp’s idea of choosing his ready-mades in terms of a “rendezvous with fate,” which he also reflected in his writing experiments The and Rendezvous, can be closely linked to his declared interest in Bergson’s “primacy of change,” leading him to explore the idea of “plastic duration.”