Paper Review: Multi-Aspect Streaming Tensor Completion

  • This article is all about a recently-viewed paper, Multi-Aspect Streaming Tensor Completion in proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD ‘17).
    • #tensor_completion, #MAST, #dynamic_tensor_decomposition,
      #online_algorithm, #streaming_data, #CP_factorization
  • Abstract
    • Completion of incremental tensors without sacrificing effectiveness remains a challenging task due to the uncertainty of tensor mode changes and complex data structure of multi-aspect streaming tensors. To bridge this gap, we propose a Multi-Aspect Streaming Tensor completion framework based on CANDECOMP/PARAFAC decomposition to track the subspace of general incremental tensors for completion.
  • Relevance to My Research
    • MAST is the 1st approach to online analysis that solved the multi-aspect streaming problem.
    • I’m also working on streaming tensor and trying to develop my method. You can see the progress of my research on GitHub.

More other topics on https://datalab.snu.ac.kr/seminar.

Desktop Screen Capturing Module for MopReM

Desktop Screen Capturing Module for MopReM

This is a part of image pre-processing module for MopReM: Moiré Pattern Removal for Mobile, Texts/Diagrams on Single-colored Background. Click the link to visit the project page of SNU-CV-PIP team.

We first select the window to screen-capture and crop the shared screen with range selector. The module saves the screenshot surrounded with target frame as target.png. Re-capture the target photo and save as source.png. Only executable on desktop!

DAO-CP - Data-Adaptive Online CP decomposition for tensor stream

PONE-D-21-39294R1
I’m pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

How can we accurately and efficiently decompose a tensor stream? The problem of efficiently decomposing tensor streams has been of great interest because many real-world data dynamically change over time. Therefore, we propose DAO-CP, an accurate and efficient online CP decomposition method which adapts to data changes.