Skip to content

Thanksyy/DeepEye-APIs

master
Switch branches/tags
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
Oct 8, 2019
Oct 8, 2019
Oct 8, 2019
Jul 10, 2018
Oct 8, 2019
Oct 8, 2019

DeepEye - APIs

This is an ongoing project that is aiming to recommend good visualizations from a relational dataset. This repository share the APIs of the DeepEye's visualization recommendation module. A live demo of DeepEye: http://deepeye.tech

Description

Visualization Ranking:

the task is, given two visualization nodes, to decide which one is better. Hence we need to rank multiple visualization nodes. We support three ranking approaches:

  1. Learning-to-rank: We use a learning-to-rank model. Roughly speaking, it is a supervised learning task that takes the input space X as lists of feature vectors, and Y the output space consisting of grades (or ranks). The goal is to learn a function F (·) from the training examples, such that given two input vectors x1 and x2, it can determine which one is better, F (x1 ) or F (x2 ). We used the LambdaMART algorithm. More Details
  2. Parial Order-based approach: the basic idea is that we define some partial orders which are used to decide which visualization node is better. Then we build a graph based on the partial orders, where each vertex is a visualization node and the directed edge between two nodes is decided by the partial order. At last, we can use the graph to compute a score for each visualization node based on topology sorting, i.e., the smaller the topology order is, the larger the score is. More Details
  3. Diversified Ranking: We aim to select diversified top-k visualization nodes since there may be many similar visualizations showing redundant information. For example, v1 > v2 and v2 > v3 do not necessarily mean that v1 + v2 > v1 + v3, since v1 and v2 might be very “similar". Our basic idea is to construct a graph in which nodes are visualizations, and weight of the edge between two nodes denotes the distance between them. Then we use the graph, our defined relevance and diversity measurement to calculate diversified top-k visualizations.

Platforms

DeepEye - APIs has been tested on OS X, CentOS, Linux, and Windows 10.

Usage

Dependencies

How to use

If you want to get the latest source code, please clone it from Github repo with the following command.

https://github.com/Thanksyy/DeepEye-APIs.git
cd DeepEye-APIs
python3 test.py

Publications

Contributors

# Contributor Affiliation Contact
1 Guoliang Li Professor, Tsinghua University LastName+FirstName@tsinghua.edu.cn
2 Nan Tang Senior Scientist, Qatar Computing Research Institute ntang@hbku.edu.qa
3 Xuedi Qin PhD Candidate, Tsinghua University qxd17@mails.tsinghua.edu.cn
4 Yuyu Luo Master Student, Tsinghua University luoyy18@mails.tsinghua.edu.cn
5 Wenbo Li Undergraduate, Tsinghua University li-wb17@mails.tsinghua.edu.cn
If you have any questions or feedbacks about this project, please feel free to contact Yuyu Luo (luoyy18@mails.tsinghua.edu.cn).

About

APIs of DeepEye. DeepEye: Towards Automatic Data Visualization [ICDE 2018]

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published