How meaningful are similarities in deep trajectory representations?

Taghizadeh, Saeed; Elekes, Abel; Schäler, Martin; Böhm, Klemens

doi:10.1016/j.is.2019.101452

Cited by 8 publications

(4 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Li et al (2018b) proposed a Seq2Seq learning method (t2vec) to compute the similarity and enhance its robustness based on a new spatial proximity loss function. Taghizadeh et al (2019) further verified the effectiveness of t2vec in achieving good similarity measurement results. To reduce the unexpected influence of noise points in the process of similarity measurement, a novel auto-encoder model is proposed by introducing an attention mechanismto realize the feature representations of noisy vessel trajectories in a low dimensional space (Zhang et al, 2019a).…”

Section: Deep Learning Methods In Maritime Knowledge Discoverymentioning

confidence: 54%

Unsupervised hierarchical methodology of maritime traffic pattern extraction for knowledge discovery

Lam

Yang

et al. 2022

Transportation Research Part C: Emerging Technologies

View full text Add to dashboard Cite

Section: Deep Learning Methods In Maritime Knowledge Discoverymentioning

confidence: 54%

Unsupervised hierarchical methodology of maritime traffic pattern extraction for knowledge discovery

Lam

Yang

et al. 2022

Transportation Research Part C: Emerging Technologies

View full text Add to dashboard Cite

“…It could achieve significant similarity computation results which are robust to non-inform, low-sampling rates and random noisy data. The meaningfulness of learned similarity values in t2vec have been extensively assessed in the literature (Taghizadeh et al, 2019). To overcome the influence of unwanted noise on similarity measure, Zhang et al (2019a) presented a robust auto-encoder model with attention mechanism to learn the low-dimensional representations of noisy vessel trajectories.…”

Section: Learning-based Trajectory Similarity Computationmentioning

confidence: 99%

“…However, these traditional methods often suffer from several drawbacks, such as high computational cost and sensitivity to unwanted noise and non-uniform sampling rate. In the case of large-scale vessel trajectories in maritime applications, it is intractable to efficiently measure the trajectory similarities or cluster trajectories (Taghizadeh et al, 2019). The original vessel trajectories can be regarded as a combination of different sub-trajectories, which commonly have different geometrical structures.…”

Section: Introductionmentioning

confidence: 99%

An Unsupervised Learning Method with Convolutional Auto-Encoder for Vessel Trajectory Similarity Computation

Liang,

Liu,

et al. 2021

Preprint

View full text Add to dashboard Cite

To achieve reliable mining results for massive vessel trajectories, one of the most important challenges is how to efficiently compute the similarities between different vessel trajectories. The computation of vessel trajectory similarity has recently attracted increasing attention in the maritime data mining research community. However, traditional shape-and warping-based methods often suffer from several drawbacks such as high computational cost and sensitivity to unwanted artifacts and non-uniform sampling rates, etc. To eliminate these drawbacks, we propose an unsupervised learning method which automatically extracts low-dimensional features through a convolutional auto-encoder (CAE). In particular, we first generate the informative trajectory images by remapping the raw vessel trajectories into two-dimensional matrices while maintaining the spatio-temporal properties. Based on the massive vessel trajectories collected, the CAE can learn the low-dimensional representations of informative trajectory images in an unsupervised manner. The trajectory similarity is finally equivalent to efficiently computing the similarities between the learned low-dimensional features, which strongly correlate with the raw vessel trajectories. Comprehensive experiments on realistic data sets have demonstrated that the proposed method largely outperforms traditional trajectory similarity computation methods in terms of efficiency and effectiveness. The high-quality trajectory clustering performance could also be guaranteed according to the CAE-based trajectory similarity computation results.

show abstract

“…Therefore, the representation learning techniques used in NLP cannot be straightforwardly applied to CDR trajectories. On the other hand, existing techniques for the vector representation of trajectories, such as [16,26], focus on spatial trajectories, thus overlooking the discrete, symbolic dimension of movement, while, by contrast, techniques focusing on symbolic locations typically organize the set of locations visited by users in graphs, and not in temporal sequences, e.g [4,27], or rather interpret behavioral similarity in terms of location similarity, e.g. [29].…”

Section: Introductionmentioning

confidence: 99%

Learning Behavioral Representations of Human Mobility

Damiani

Acquaviva

Hachem

et al. 2020

Proceedings of the 28th International Conference on Advances in Geographic Information Systems

View full text Add to dashboard Cite

In this paper, we investigate the suitability of state-of-the-art representation learning methods to the analysis of behavioral similarity of moving individuals, based on CDR trajectories. The core of the contribution is a novel methodological framework, mob2vec, centered on the combined use of a recent symbolic trajectory segmentation method for the removal of noise, a novel trajectory generalization method incorporating behavioral information, and an unsupervised technique for the learning of vector representations from sequential data. mob2vec is the result of an empirical study conducted on real CDR data through an extensive experimentation. As a result, it is shown that mob2vec generates vector representations of CDR trajectories in low dimensional spaces which preserve the similarity of the mobility behavior of individuals.

show abstract

How meaningful are similarities in deep trajectory representations?

Cited by 8 publications

References 16 publications

Unsupervised hierarchical methodology of maritime traffic pattern extraction for knowledge discovery

Unsupervised hierarchical methodology of maritime traffic pattern extraction for knowledge discovery

An Unsupervised Learning Method with Convolutional Auto-Encoder for Vessel Trajectory Similarity Computation

Learning Behavioral Representations of Human Mobility

Contact Info

Product

Resources

About