AN Active Learning sYstem Trajectory Classification (ANALYTiC)

An AN Active Learning sYstem Trajectory Classification (ANALYTiC) is a web-based Active Learning System in which the user can interact with a learning trajectory annotation algorithm.

• Context:
  • It was developed by Junior et al. (2017).
  • It can solve a Active Learning Trajectory Annotation Task by implementing a Active Learning Trajectory Annotation Algorithm.
  • It uses the following libraries:
    • Mapbox and Leaflet for map visualization and user interaction;
    • Google Chart for data visualization;
    • scikit-learn for implemeting active learning and Slor database query algorithms.
    • Bottle Python Web Framework for implementing a RESTful Web Services.
• Example(s):
  • …
• Counter-Example(s):
  • NEXT.
  • a Constructive Active Learning System,
  • a Pool-Based Active Learning System,
  • a Stream-Based Active Learning System,
  • a Statistical Active Learning System.
  • an ASE-Progol System.
• See: Interactive Entity Record Disambiguation System, Active Learning Theory, Machine Learning System.

---

---

References

2017

• (Junior et al., 2017) ⇒ Amilcar Soares Junior, Chiara Renso, and Stan Matwin. (2017). “ANALYTiC: An Active Learning System for Trajectory Classification.” In: IEEE Computer Graphics and Applications Journal, 37(5). doi:10.1109/MCG.2017.3621221
  • QUOTE: ... we propose a web-based visual interactive annotation tool named ANALYTiC (AN Active Learning sYstem Trajectory Classification) supporting the user in the active learning trajectory annotation task. By designing an effective visual support we can enable fast, accurate, and hopefully trustworthy semantic annotation of the learning training set. The basic idea of active learning is to interactively submit to the annotator the best set of trajectories to annotate to improve the classifiers’ performance. In machine learning, having a good sample of labeled trajectories is essential to reach good performance values, thus the ANALYTiC tool is a step towards this direction.

    To the best of our knowledge, this is the first attempt to exploit active learning techniques in the trajectory semantic classification field. At the same time, this is the first annotation tool for trajectories tailored to the active learning task.

    ...we propose an answer to research question RQ3 - How can the user be assisted in labeling trajectories? The ANALYTiC visual interactive system is a possible answer to this question. The architecture is shown in Figure 4.

     

    Figure 4: The ANALYTiC architecture.

---