The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
A previsão de trajetória semântica multimodal tornou-se um novo desafio devido ao rápido crescimento de trajetórias semânticas multimodais com mensagens de texto. Os métodos tradicionais de previsão de trajetória RNN têm os seguintes problemas para processar trajetória semântica multimodal. A distribuição de amostras de trajetórias multimodais muda gradualmente com o treinamento. Isso leva a uma convergência difícil e a um longo tempo de treinamento. Além disso, cada recurso modal muda em direções diferentes, o que produz múltiplas distribuições do conjunto de dados. Para resolver os problemas acima, é proposto o MNERM (Modelo Recorrente Aprimorado de Normalização de Modo) para trajetória semântica multimodal. MNERM incorpora vários recursos modais e combina a rede LSTM para capturar a dependência de trajetória de longo prazo. Além disso, ele projeta um mecanismo de normalização de modo para normalizar amostras com múltiplas médias e variâncias, e cada distribuição normalizada cai na área de ação da função de ativação, de modo a melhorar a eficiência da previsão e, ao mesmo tempo, melhorar significativamente a velocidade de treinamento. Experimentos em conjuntos de dados reais mostram que, comparado ao SERM, o MNERM reduz a sensibilidade da taxa de aprendizagem, melhora a velocidade de treinamento em 9.120 vezes, aumenta a FC@1 em 0.03 e reduz o ADE em 120 metros.
Shaojie ZHU
China University of Mining and Technology,Ministry of Education
Lei ZHANG
China University of Mining and Technology,Ministry of Education
Bailong LIU
China University of Mining and Technology,Ministry of Education
Shumin CUI
China University of Mining and Technology,Ministry of Education
Changxing SHAO
China University of Mining and Technology,Ministry of Education
Yun LI
China University of Mining and Technology,Ministry of Education
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copiar
Shaojie ZHU, Lei ZHANG, Bailong LIU, Shumin CUI, Changxing SHAO, Yun LI, "Mode Normalization Enhanced Recurrent Model for Multi-Modal Semantic Trajectory Prediction" in IEICE TRANSACTIONS on Information,
vol. E103-D, no. 1, pp. 174-176, January 2020, doi: 10.1587/transinf.2019EDL8130.
Abstract: Multi-modal semantic trajectory prediction has become a new challenge due to the rapid growth of multi-modal semantic trajectories with text message. Traditional RNN trajectory prediction methods have the following problems to process multi-modal semantic trajectory. The distribution of multi-modal trajectory samples shifts gradually with training. It leads to difficult convergency and long training time. Moreover, each modal feature shifts in different directions, which produces multiple distributions of dataset. To solve the above problems, MNERM (Mode Normalization Enhanced Recurrent Model) for multi-modal semantic trajectory is proposed. MNERM embeds multiple modal features together and combines the LSTM network to capture long-term dependency of trajectory. In addition, it designs Mode Normalization mechanism to normalize samples with multiple means and variances, and each distribution normalized falls into the action area of the activation function, so as to improve the prediction efficiency while improving greatly the training speed. Experiments on real dataset show that, compared with SERM, MNERM reduces the sensitivity of learning rate, improves the training speed by 9.120 times, increases HR@1 by 0.03, and reduces the ADE by 120 meters.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2019EDL8130/_p
Copiar
@ARTICLE{e103-d_1_174,
author={Shaojie ZHU, Lei ZHANG, Bailong LIU, Shumin CUI, Changxing SHAO, Yun LI, },
journal={IEICE TRANSACTIONS on Information},
title={Mode Normalization Enhanced Recurrent Model for Multi-Modal Semantic Trajectory Prediction},
year={2020},
volume={E103-D},
number={1},
pages={174-176},
abstract={Multi-modal semantic trajectory prediction has become a new challenge due to the rapid growth of multi-modal semantic trajectories with text message. Traditional RNN trajectory prediction methods have the following problems to process multi-modal semantic trajectory. The distribution of multi-modal trajectory samples shifts gradually with training. It leads to difficult convergency and long training time. Moreover, each modal feature shifts in different directions, which produces multiple distributions of dataset. To solve the above problems, MNERM (Mode Normalization Enhanced Recurrent Model) for multi-modal semantic trajectory is proposed. MNERM embeds multiple modal features together and combines the LSTM network to capture long-term dependency of trajectory. In addition, it designs Mode Normalization mechanism to normalize samples with multiple means and variances, and each distribution normalized falls into the action area of the activation function, so as to improve the prediction efficiency while improving greatly the training speed. Experiments on real dataset show that, compared with SERM, MNERM reduces the sensitivity of learning rate, improves the training speed by 9.120 times, increases HR@1 by 0.03, and reduces the ADE by 120 meters.},
keywords={},
doi={10.1587/transinf.2019EDL8130},
ISSN={1745-1361},
month={January},}
Copiar
TY - JOUR
TI - Mode Normalization Enhanced Recurrent Model for Multi-Modal Semantic Trajectory Prediction
T2 - IEICE TRANSACTIONS on Information
SP - 174
EP - 176
AU - Shaojie ZHU
AU - Lei ZHANG
AU - Bailong LIU
AU - Shumin CUI
AU - Changxing SHAO
AU - Yun LI
PY - 2020
DO - 10.1587/transinf.2019EDL8130
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E103-D
IS - 1
JA - IEICE TRANSACTIONS on Information
Y1 - January 2020
AB - Multi-modal semantic trajectory prediction has become a new challenge due to the rapid growth of multi-modal semantic trajectories with text message. Traditional RNN trajectory prediction methods have the following problems to process multi-modal semantic trajectory. The distribution of multi-modal trajectory samples shifts gradually with training. It leads to difficult convergency and long training time. Moreover, each modal feature shifts in different directions, which produces multiple distributions of dataset. To solve the above problems, MNERM (Mode Normalization Enhanced Recurrent Model) for multi-modal semantic trajectory is proposed. MNERM embeds multiple modal features together and combines the LSTM network to capture long-term dependency of trajectory. In addition, it designs Mode Normalization mechanism to normalize samples with multiple means and variances, and each distribution normalized falls into the action area of the activation function, so as to improve the prediction efficiency while improving greatly the training speed. Experiments on real dataset show that, compared with SERM, MNERM reduces the sensitivity of learning rate, improves the training speed by 9.120 times, increases HR@1 by 0.03, and reduces the ADE by 120 meters.
ER -