Mingjie Zeng - Update week 10

Mingjie Zeng - week10.md

@@ -0,0 +1,28 @@
+#### Reflection 10 - 03/21/2022
+#### Mingjie Zeng (671222265)
+#### Email:mzeng2@wpi.edu
+----
+
+This week's paper is about a method called Text-to-Viz of automatically generating visualizations from proportion-related natural language statements: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8813126
+
+First of all, here are some examples created by Text-to-Viz:
+![image](https://github.com/JasmineZZZ9/reflections-research/blob/main/pics/r10-exp.jpg)
+For example, (e) and (f) are generated from the statement: “40 percent of USA freshwater is for agriculture.” We can see that the visualizations can effectively deliver messages and information about numbers in an outstanding and memorable way. Although there are many tools to create vary kinds of infographics, it takes time to know how to create these visualizations.
+In this paper, the authors explore a method to automatically generate these visualizations from natural language statements for those who are either unwilling to take time to learn the tools or lacking in proper
+design expertise to create a professional infographic.
+
+The authors first conducted a preliminary study to explore the design space of infographics. The goal of this survey was to better understand how infographics are used in real life and identify a specific type on which to build this proof-of-concept system. 
+And the authors categorized all the infographic into four main types: statistical-based, timeline-based, process-based and location-based. And also they further discovered diverse patterns in terms of graphic designs and the underlying messages, which were categorized into four major sub-categories, namely proportion, quantity, change, and rank.
+
+Based on the preliminary study, they built a proof-of-concept system that automatically converts statements about simple proportion-related statistics to a set of infographics with pre-designed styles. 
+The two main part of this system are text analyzer and visual generator. In the text analyzer part, a textual statement is identified by the analyzer and is segmented into 5 basic segments such as modifier, whole, part, number and others.
+Then the original statement and the extracted segments are fed into the visual generator for infographic construction for each dimension including layout, description, graphic, and color. Then, the system enumerates all combinations of
+these elements, to synthesize valid infographic candidates. Finally, all the synthesized results are evaluated and ranked, and the ones with high scores are recommended to users.
+
+Finally, they demonstrated the usability and usefulness of the system through sample results, exhibits, and expert reviews.
+
+But this work may not always be successful. For some complicated and long statements, the texutal analyzer may not segment correctly and provide correct tags, which will eventually lead to incomprehensible infographics. The design also has some limitations in terms of capability. The foremost limitation is that the current approach
+can only handle a relatively small set of information. Also this approach lacks human creativity and is based on a set of pre-designed infographic styles.
+
+
+Although this method has many limitations, I think it's creative and a good combination of natural language processing and data visualization. The idea of automatically generating the visualizations by some textual information is not only cool but also has a lot of usage, especially for today's situation about information explosion, people need to use visualization to make the info more straightforward to let people spend less time to take that info in.

Mingjie Zeng - week11.md

@@ -0,0 +1,29 @@
+#### Reflection 11 - 03/28/2022
+#### Mingjie Zeng (671222265)
+#### Email:mzeng2@wpi.edu
+----
+
+This week's paper is DeepDrawing: A Deep Learning Approach to Graph Drawing: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8807275
+
+This paper proposes a method to directly map network structures to graph drawing using a graph-LSTM-based approach. The authors train the proposed
+graph-LSTM-based model to capture their layout characteristics given a set of layout examples as the training dataset. Then, the trained model is used to generate graph drawings in a
+similar style for new networks. Furthermore, the authors evaluated the proposed approach on grid layouts and star layouts in both qualitative and quantitative ways and also a time cost assessment on the drawings of small graphs with 20 to 50 nodes were conducted.
+
+Here is the workflow of graph drawing algorithms, (a) is traditional graph drawing algorithms and (b) is the proposed deep learning based approach.
+![image](https://github.com/JasmineZZZ9/reflections-research/blob/main/pics/r11-workflow.jpg)
+
+When using a drawing algorithm to visualize a specific graph using the traditional way, users also need to tune the algorithm parameters through
+trial and error to achieve a suitable graph drawing result, which is tedious and time-consuming. But for the deep learning based approach, given a set of graph drawing examples with desirable aesthetic properties and their structures, the deep learning
+model is trained to learn the mapping and corresponding algorithmspecific parameters for determining the desirable graph drawings. Once the deep learning model is successfully trained, when given a new graph, it can automatically analyze the graph
+structure and directly generate a layout that carries the common visual properties of the drawing examples.
+
+Besides the proposed DeepDrawing method, the authors also implement wonderful experiment for evaluation. The picture below is their qualitative evaluation results on star graphs with different number of nodes.
+![image](https://github.com/JasmineZZZ9/reflections-research/blob/main/pics/r11-evaluzation.jpg)
+
+They also conduct experiment on general graphs, here is a small part of the experiment results:
+![image](https://github.com/JasmineZZZ9/reflections-research/blob/main/pics/r11-general.jpg)
+
+Specially, the authors evaluate the time cost of the proposed approach in comparison with both the original graph drawing techniques and the baseline model.
+
+This paper does not only propose a deep learning based method but also conduct wonderful experiments for evaluation which should be learned from. In the evaluation part, the authors use a lot of comparisons for the result which is clear and meaningful and the comparisons are from vary different aspects.
+The combination of deep learning and graph drawing is full of inspiration but there are still limitations causing failure cases. Automatically creating visualizations is always a chanllenging but meaningful work, there are a lot more to explore.

Mingjie Zeng - week12.md

@@ -0,0 +1,26 @@
+#### Reflection 12 - 04/04/2022
+#### Mingjie Zeng (671222265)
+#### Email:mzeng2@wpi.edu
+----
+
+This week's paper is GUIRO: User-Guided Matrix Reordering: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8807245
+
+Matrix representations are very effective visualization techniques for relational or network data. However, it's very difficult to dicide which matrix reordering algorithm should be applied basee on different dateset. What's worse is that different reordering algorithms applied to the same dateset may let significantly different visual matrix patterns emerge. 
+This paper presents GUIRO, a Visual Analytics system that helps novices, network analysts, and algorithm designers to investigate the usefulness and expressiveness of 70 accessible matrix reordering algorithms. GUIRO helps
+to increase the transparency of matrix reordering algorithms, thus helping a broad range of users to get a better insight into the complex
+reordering process, in turn supporting data and reordering algorithm insights.
+
+This is a simple example demonstrating the basic design of GUIRO. This picture shows a specific dataset with three reodering options in the left part. And in the right part, the projection space shows that distinct groups of rows/columns can be formed.
+![image](https://github.com/JasmineZZZ9/reflections-research/blob/main/pics/r12-emp.jpg)
+
+The system provides 71 different choices of reordering algorithms which shows the comparisons between the different matrix reodering results:
+
+![image](https://github.com/JasmineZZZ9/reflections-research/blob/main/pics/r12-choice.jpg)
+
+Based on the matrix reodering results, then the system represents matrix reoderings in the projection space and then interprets reodering results in the projection space.
+
+The system proposed in this paper guides the user to submatrices, which could potentially be improved by showing row/column similarity in a novel projection view. And the user may choose to apply an arbitrary automatic matrix reordering algorithm on selected submatrices.
+This system also helps the user by showcasing thumbnails of the local reorderings to help to anticipate the operation’s outcome. Furthermore, the user is free to rearrange the rows and columns—or groups thereof—manually. 
+In addition, the approach allows the hierarchical construction of new matrix reordering algorithms, by applying local optimizations on a global matrix reordering result.
+
+For me, it's very interesting to solve the relationship between network data in this way and I'm still curious how to deal with the relationship or order data in a network graph. The matrix reodering method mentioned in the paper may give up some inspirations.

Mingjie Zeng - week13.md

@@ -0,0 +1,21 @@
+#### Reflection 13 - 04/10/2022
+#### Mingjie Zeng (671222265)
+#### Email:mzeng2@wpi.edu
+----
+
+This week's paper is Animated Exploration of Graphs with Radial Layout: http://zesty.ca/pubs/infovis-2001-submit.pdf
+
+In this paper, the authors describe a new animation technique for supporting interactive exploration of a graph, building on the wellknown radial tree layout method and they apply this
+technique to visualizations of social networks and of the Gnutella file-sharing network.
+
+I found this paper interesting because our project has something about the network graphs and we found that the network structure is important and somehow complecated. And also, visualizing a network structure need the data to be well structured. So given the data, generating the network visualization is a challenging task.
+In this paper, the authors consider visualizations in which the view of the graph is determined by the selection of a single node as the center of interest, or focus. The main contribution of this work is a new technique for animating the transitions from one view to the next, in a smooth, appealing manner that minimizes confusion.
+And here is the visualization of the Gnutella network:
+
+![image](https://github.com/JasmineZZZ9/reflections-research/blob/main/pics/r13%20-%20g.jpg)
+
+And here is an good point that this method can change the focused node with smooth animation, both keep the original orientation of the edges between the nodes and the ordering of the nodes' neighbors.
+![image](https://github.com/JasmineZZZ9/reflections-research/blob/main/pics/r13%20-%20change1.jpg)
+![image](https://github.com/JasmineZZZ9/reflections-research/blob/main/pics/r13-change2.jpg)
+
+The new method for animating the transitions from one view to the next in an appealing manner that minimizes confusion is well performed and useful.

Mingjie Zeng - week14.md

@@ -0,0 +1,21 @@
+#### Reflection 14 - 04/18/2022
+#### Mingjie Zeng (671222265)
+#### Email:mzeng2@wpi.edu
+----
+
+The title of this week's paper is The Value of Visualization: https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.75.6547&rep=rep1&type=pdf
+
+This course has almost come to an end, and we have learned the basics of data visualization, including the various types of charts, applications, etc. And at the end of this semester, I'd like to come back to the most basic and essential part, the value of visualization.
+
+As the authors said, the field of visualization is getting mature. Many problems have been solved, and new directions are sought for. In order to make
+good choices, an understanding of the purpose and meaning of visualization is needed. And it's very important to to assess what a good visualization is.
+
+In this paper, an economic model of visualization is presented where the adopted viewpoint is that the value of visualization is measured based on effectiveness and efficiency. The authors use the model to judge the methods that visualization use and to determine the reason why these methods are used. And they also evaluate the visualization on both art and scientic points of view.
+
+Here is the basic model of visualization. The authors use a visualization way to describe the systematical process of visualization:
+![image](https://github.com/JasmineZZZ9/reflections-research/blob/main/pics/r14%20-%20model.jpg)
+
+And with this model and its abstract parameters, by assessing values to it, there is a way to numerically determine whether a visualization method is worthwhile.
+
+But there are many things are affecting the value of a visualization, there is not a single answer to tell the value of a visualization, it depends on the point of view one adopts. However, there are still some wonderful points that can determine whether a visualization is of value, including the provable effectiveness and efficiency, elegance and beauty,
+and generic laws with predictive power.

Mingjie Zeng - week15.md

@@ -0,0 +1,21 @@
+#### Reflection 15 - 04/25/2022
+#### Mingjie Zeng (671222265)
+#### Email:mzeng2@wpi.edu
+----
+
+The title of this week's paper is Data Changes Everything: Challenges and Opportunities in Data Visualization Design Handoff: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8816695
+
+There are gaps between data characterization tools, visualization design tools, and development platforms which will pose challenges for designer-developer teams working to create new data visualizations. It requires collaboration between people with different visualization-related
+skills. The authors report observations and reflections from five large multidisciplinary visualization design
+projects and highlight six data-specific visualization challenges for design specification and handoff. These challenges include adapting
+to changing data, anticipating edge cases in data, understanding technical challenges, articulating data-dependent interactions,
+communicating data mappings, and preserving the integrity of data mappings across iterations. And based on these observations, the authors give some advice on future tools for prototyping, testing, and communicating data-driven designs, which may make visualization design easier to be successful.
+
+There are five visualization design and development projects in which the authors observed:
+
+![image](https://github.com/JasmineZZZ9/reflections-research/blob/main/pics/r15%20-%20design.jpg)
+
+For every project that picked up, there should be a variety of design-related roles to complete the progress of design. And the authors documented the full design and discussion process, judging and analyzing the difficulties in the process through various aspects of reflection. 
+
+We all know that there would be many challenges when it comes to design and publish a new visualization, I think the methods that the authors used to record these challenges in this paper are actrually we can learn from. To make progress, we need to know where the limitations are. 
+It is not only necessary to do so in the process of research and development of visualization, but also in the path of research and development in all fields, this method of determining the direction of development through experimentation and analysis is worth learning.