RDMMFET: Representation of Dense Multimodality Fusion Encoder Based on Transformer

Zhang, Xu; Han, Dong‐Guk; Chang, Chin‐Chen

doi:10.1155/2021/2662064

Cited by 4 publications

(4 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As mentioned above, we investigate how to utilize the deep learning model for the English translation in order to better assist the establishment of the machine translation system [27]. We design an English translation method based on an encoder-decoder structure and an attention mechanism.…”

Section: ⅲ Methodsmentioning

confidence: 99%

Online Automatic English Translation Based on Mobile Edge Computing

Qiu¹,

Qiu²

2023

J Multimed Inf Syst

View full text Add to dashboard Cite

The English translation is an important means of converting language as the basis of communication, is gradually becoming an indispensable part of people's daily life. With the rapid progress of science, cultural exchanges between countries in the world have become more and more common. And the advent of the Internet era has led to more frequent exchanges between different languages, and the demand for language translation has gradually increased. How to translate English quickly and accurately is the current challenge of machine translation tasks. In this paper, we investigate how to use the neural network model for the English translation in order to better assist the establishment of the machine translation system. We design an English translation method based on an encoder-decoder structure and an attention mechanism. First, we analyze the characteristics of the LSTM model. Second, we design an English translation framework using the seq2seq model. Third, we combine the attention mechanism to build a more robust translation model to improve the translation performance of the model. Finally, we validate the translation performance of our model on two public datasets and experimental results prove that the method proposed in this paper has good evaluation performance.

show abstract

Section: ⅲ Methodsmentioning

confidence: 99%

Online Automatic English Translation Based on Mobile Edge Computing

Qiu¹,

Qiu²

2023

J Multimed Inf Syst

View full text Add to dashboard Cite

show abstract

“…In addition, as mentioned before, the model should adequately study both local and global features. Given these concerns, we constructed a model comprised of tandem Gated Recurrent Units (GRU) [25][26] and Transformer encoder [27][28] for sequence modelling (Figure 2a). The data sequence was split into short fragments and fed into GRU cells, whose outputs were concatenated and processed by Transformer encoder (Figure 2a).…”

Section: Standalone Neural Network For Pattern Extractionmentioning

confidence: 99%

FECDO-Flexible and Efficient Coding for DNA Odyssey

Sun,

Qian

2024

Preprint

View full text Add to dashboard Cite

SummaryDNA has been pursued as a compelling medium for digital data storage during the past decade. While large-scale data storage and random access have been achieved in artificial DNA, the synthesis cost keeps hindering DNA data storage from popularizing into daily life. In this study, we proposed a more efficient paradigm for digital data compressing to DNA, while excluding arbitrary sequence constraints. Both standalone neural networks and pre-trained language models were used to extract the intrinsic patterns of data, and generated probabilistic portrayal, which was then transformed into constraint-free nucleotide sequences with a hierarchical finite state machine. Utilizing these methods, a 12%-26% improvement of compression ratio was realized for various data, which directly translated to up to 26% reduction in DNA synthesis cost. Combined with the progress in DNA synthesis, our methods are expected to facilitate the realization of practical DNA data storage.

show abstract

“…y i is its corresponding ground-truth intent label. RNNbased encoder-decoder captures the temporal correlation through model parameters (like memories), transformerbased model (Vaswani et al 2017a;Han et al 2021;Xu et al 2021;Yi and Qu 2022;Wu et al 2023) design attention modules to capture all the possible relationships. In other words, for a trained model, the attention mechanism relies on the data itself explicitly to capture the temporal correlation, while LSTM/RNN memorizes the temporal information implicitly through model parameters.…”

Section: Framework Overviewmentioning

confidence: 99%

TrEP: Transformer-Based Evidential Prediction for Pedestrian Intention with Uncertainty

Zhang

Tian

Ding

2023

AAAI

View full text Add to dashboard Cite

With rapid development in hardware (sensors and processors) and AI algorithms, automated driving techniques have entered the public’s daily life and achieved great success in supporting human driving performance. However, due to the high contextual variations and temporal dynamics in pedestrian behaviors, the interaction between autonomous-driving cars and pedestrians remains challenging, impeding the development of fully autonomous driving systems. This paper focuses on predicting pedestrian intention with a novel transformer-based evidential prediction (TrEP) algorithm. We develop a transformer module towards the temporal correlations among the input features within pedestrian video sequences and a deep evidential learning model to capture the AI uncertainty under scene complexities. Experimental results on three popular pedestrian intent benchmarks have verified the effectiveness of our proposed model over the state-of-the-art. The algorithm performance can be further boosted by controlling the uncertainty level. We systematically compare human disagreements with AI uncertainty to further evaluate AI performance in confusing scenes. The code is released at https://github.com/zzmonlyyou/TrEP.git.

show abstract

RDMMFET: Representation of Dense Multimodality Fusion Encoder Based on Transformer

Cited by 4 publications

References 23 publications

Online Automatic English Translation Based on Mobile Edge Computing

Online Automatic English Translation Based on Mobile Edge Computing

FECDO-Flexible and Efficient Coding for DNA Odyssey

TrEP: Transformer-Based Evidential Prediction for Pedestrian Intention with Uncertainty

Contact Info

Product

Resources

About