Time-series Clustering with Jointly Learning Deep Representations, Clusters and Temporal Boundaries

  title={Time-series Clustering with Jointly Learning Deep Representations, Clusters and Temporal Boundaries},
  author={Panagiotis Tzirakis and Mihalis A. Nicolaou and Bj{\"o}rn Schuller and Stefanos Zafeiriou},
  journal={2019 14th IEEE International Conference on Automatic Face \& Gesture Recognition (FG 2019)},
Clustering and segmentation of temporal data is an important task across several fields, with prominent applications in computer vision and machine learning such as face and gesture segmentation. [] Key Method In this paper, we propose the first methodology that simultaneously discovers suitable deep representations, as well as clusters and temporal boundaries, with the clustering process providing supervisory cues for updating temporal boundaries and training the proposed deep learning architecture. We…

Figures and Tables from this paper

Clustering of Time Series Data with Prior Geographical Information

This paper proposes a variation of a Deep Embedded Clustering (DEC) model, where it clusters time series data based on spatial and temporal contexts, and shows that the proposed Spatial-DEC can find more desired spatio-temporal clusters.

Deep Multivariate Time Series Embedding Clustering via Attentive-Gated Autoencoder

A deep-learning based framework for clustering multivariate time series data with varying lengths, namely DeTSEC (Deep Time Series Embedding Clustering), includes two stages: firstly a recurrent autoencoder exploits attention and gating mechanisms to produce a preliminary embedding representation; then, a clustering refinement stage is introduced to stretch the embedding manifold towards the corresponding clusters.

A Review of Motion Segmentation: Approaches and Major Challenges

Several popular and state-of-the-art methods in motion segmentation were reviewed, with the focus on the most important attributes, namely Image Difference, Optical Flow, Wavelet, Statistical, Layers, Manifold Clustering, Template Matching, and Deep Learning.

Spatio-Temporal Clustering of Traffic Data with Deep Embedded Clustering

The analysis shows that the temporal and temporal clusters have meaningful relationships with temporal and spatial patterns in traffic data, and the clustering method effectively finds similarities in trafficData.

Static Seeding and Clustering of LSTM Embeddings to Learn from Loosely Time-Decoupled Events

This work improves on the use of Recurring Neural Networks (RNN) in particular Long Short-Term Memory (LSTM) networks to enable AI solutions that generate better timeseries predictions for LDT, and shows an improvement in 10-day moving average predictions of disease propagation at the US County level.

ARMA-Based Segmentation of Human Limb Motion Sequences

An unsupervised segmentation algorithm based on limb-bone partition angle body structural representation and autoregressive moving average (ARMA) model fitting is proposed and compared with the principle component analysis (PCA), K-means clustering algorithm (K-me means), and back propagation (BP) neural-network-based segmentation algorithms, which shows higher segmentation accuracy.



Joint Unsupervised Learning of Deep Representations and Image Clusters

A recurrent framework for joint unsupervised learning of deep representations and image clusters by integrating two processes into a single model with a unified weighted triplet loss function and optimizing it end-to-end can obtain not only more powerful representations, but also more precise image clusters.

Human Motion Segmentation via Robust Kernel Sparse Subspace Clustering

This paper converts the segmentation of motion capture data into a temporal subspace clustering problem, and proposes a new segmentation method, which is robust to non-Gaussian noise, since correntropy is a localized similarity measure.

Aligned Cluster Analysis for temporal segmentation of human motion

Aligned cluster analysis (ACA) is proposed, a robust method to temporally segment streams of motion capture data into actions and a dynamic time warping kernel is used to achieve temporal invariance.

Hierarchical Aligned Cluster Analysis for Temporal Clustering of Human Motion

This work poses the problem of learning motion primitives as one of temporal clustering, and derives an unsupervised hierarchical bottom-up framework called hierarchical aligned cluster analysis (HACA), which finds a partition of a given multidimensional time series into m disjoint segments such that each segment belongs to one of k clusters.

Structured Time Series Analysis for Human Action Segmentation and Recognition

By combining the temporal segmentation algorithm and the alignment algorithm, online human action recognition can be performed by associating a few labeled examples from motion capture data and its ability to handle noisy and partially occluded data, in the transfer learning module is demonstrated.

Towards K-means-friendly Spaces: Simultaneous Deep Learning and Clustering

A joint DR and K-means clustering approach in which DR is accomplished via learning a deep neural network (DNN) while exploiting theDeep neural network's ability to approximate any nonlinear function is proposed.

Efficient Unsupervised Temporal Segmentation of Motion Data

A method for automated temporal segmentation of human motion data into distinct actions and compositing motion primitives based on self-similar structures in the motion sequence is introduced, which requires no assumptions about the motion sequences at hand and no user interaction for the segmentation or clustering.

A Proximity-Aware Hierarchical Clustering of Faces

An unsupervised face clustering algorithm called “Proximity-Aware Hierarchical Clustering” (PAHC) that exploits the local structure of deep representations that can be applied to curate a set of large-scale and noisy training dataset while maintaining sufficient amount of images and their variations due to nuisance factors.

Unsupervised Learning of Deep Feature Representation for Clustering Egocentric Actions

This work proposes a robust and generic unsupervised approach for first person action clustering that surpasses the supervised state of the art accuracies without using the action labels and demonstrates that clustering of features leads to the discovery of semantically meaningful actions present in the video.

Statistical analysis of dynamic actions

A simple statistical distance measure between video sequences which captures the similarities in their behavioral content is designed, which is nonparametric and can handle a wide range of complex dynamic actions.