43 Matching Annotations
  1. Jul 2019
  2. May 2019

      This is an interesting fact, usually when I think of visualization and data I go to the classic default charts and data. I'll have to keep this iin mind.

  3. Feb 2019
    1. To help us get better comprehension of the structure of an argument, we can also call forth a schematic or graphical display

      I might be getting ahead of what's to come, since I am annotating as I am reading, but this gets me thinking about some visualization approaches I saw in the 1990s by the brilliant and forgotten Roy Stringer working on what he called "Navihedra" - while they were often seen as navigational, his ideas seemed to be rooted in better representations of the kinds of structures Engelbart is telling us

      In brief, however, Navihedra are 3D models based on Platonic solids and relationships between pieces of information are articulated in terms of the spatial relationships represented by the vertices of the polyhedron. That is, units of information (of any kind, media, size or complexity) are attached to a specific vertex and bi-directionally hyperlinked to all the immediately adjacent vertices. The overall structure being determined by some perceived relevance reflected in proximity. Proximate vertices are understood to locate units of information/argument that are more closely related to one another than units of information that are not directly hyperlinked. Furthermore, this 3 dimensional arrangement can be rotated in space so that differing patterns of inter-relatedness can be viewed. Creating such an arrangement is much more difficult than it might appear and requires an author to consider the structure/presentation of even a simple argument like the one contained in this article with at least as much care as a more conventional presentation.

      Sadly these were produced in a media form hardly displayable now (Macromedia Shockwave), remnants are in the Internet Archive.

    1. Deep Learning Multidimensional Projections


      其中有好些是与 UMAP 和 t-sne 做的对比。

  4. Jan 2019
    1. Nyhan and Reifler also found that presenting challenging information in a chart or graph tends to reduce disconfirmation bias. The researchers concluded that the decreased ambiguity of graphical information (as opposed to text) makes it harder for test subjects to question or argue against the content of the chart.

      Amazingly important double-edged finding for discussions of data visualization!

  5. Dec 2018
    1. A Tutorial on Distance Metric Learning: Mathematical Foundations, Algorithms and Software

      很不错的 Distance Metric Learning 综述性材料,富含概念,如何设计DML算法,DML 算法的数学理论是怎样的(凸优化、矩阵分析、信息论)等等。最后开源了Python 库 pyDML 以方便研究此 paper 中的算法。

    2. How convolutional neural network see the world - A survey of convolutional neural network visualization methods

      果断收藏并且要细读下。。。Paper Summary 准备!

      这可是对 CNN 可视化方法的 review 啊!

      一篇很棒的综述,专门说 CNN 的可视化的!要好好读读了!

      Paper Summary 准备!

  6. Nov 2018
    1. Why scatter plots suggest causality, and what we can do about it

      看了半天我真是不明白,转了45度再把图捏成方形的,就可以写篇 paper 宣传了?。。。[哼]

    2. UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction

      此文提供了一个和 t-sne 非常类似的降维可视化算法。效果相当不错!也开源了算法代码。

      按照作者的说法,UMAP 比 T-SNE 算法更好的优点有二:更快!更准!

    3. Local Explanation Methods for Deep Neural Networks Lack Sensitivity to Parameter Values

      This paper shows that local explanations for DNNs with random-initialized weights are qualitatively and quantitatively similar to explanations produced by DNNs with learned weights.

      • Pros:

      The paper is clear, the problem is well stated and the method is sound.

      • Cons:

      The impact of the findings in this paper is unclear. Perhaps the most important point made in the paper is the importance of the architecture over fine-tuning of the weights for explanation tasks (and more in general).

      其实 goodfellow 这文章篇幅很短,可视化图像的效果是很棒的!

    4. Sanity Checks for Saliency Maps

      专门探讨对各种 Saliency methods (显著图方法)的。

      Goodfellow 署名的该文章内含有大量很棒的可视化效果。

    5. Using Machine Learning to Predict the Evolution of Physics Research


  7. Oct 2018
  8. idyll-lang.org idyll-lang.org
    1. A toolkit for creating data-driven stories and explorable explanations.

      Markup language for creating data driven stories

  9. Jun 2018
  10. May 2018
    1. Thus, the digital object ossifies out of two histories, one virtual and another visual.Within computation, the object arises out of a desire to create a model of the worldwithin the computer but at the same time out of an attempt to create a whole new visualworld native to the compute



  11. Mar 2018
  12. Oct 2017
  13. Sep 2017
    1. After we were done, pictures were taken of the group and distributed online to groups in other cities performing similar activities, contributing to the spectacle of the day.

      En los eventos locales se toman fotos durante el evento, al margen de los resultados. En el Data Week en cambio, las fotos con pocas en comparación (a veces nulas), particularmente en consideración a la privacidad. La lógica del espectáculo/impacto está más centrada en las visualizaciones mismas.

    1. That “hackers” can model beneficial process disrupts the often presumed subversive nature of hacking as much as it does easy assumptions about a Foucaultian notion of governmentality. Prototypes act as working evidence to lobby for changing government process, particularly those that improve digital infrastructure or direct communication with citizens. The capa-bility of code to act as a persuasive argument has long been noted, and modeling can produce charged debates about the very meaning of “civic.”

      [...] On a level of hackathons, prototypes can be speculative (Lodato and DiSalvo, in press) rather than an “outcome,” revealing conflicting notions of “civic tech” (Shaw, 2014).

      Nuestro enfoque ha estado centrado más en la modelación, que es requerida para la visualización, pero también en la idea de construir capacidad en la infraestructura y en la comunidad, lo cual va más allá del prototipo volátil, que se abandona después.

  14. Aug 2017
  15. Apr 2017
    1. opml2json A simple tool to convert opml files exported by Mindnode Pro to JSON consumable by D3 Javascript library.
  16. Mar 2017
    1. Prophet : Facebook에서 오픈 소스로 공개한 시계열 데이터의 예측 도구로 R과 Python으로 작성되었다.

      python statics opensource, also can use R

  17. Feb 2017
  18. Jan 2017
    1. Haiku for Clouds

      The collective noun for a plural of haiku is a 'visualization'. See below:

  19. Dec 2016
    1. sites such as Facebook and Twitter automatically and continuously refresh the page; it’s impossible to get to the bottom of the feed.

      Well is not. A scrapping web technique used for the Data Selfies project goes to the end of the scrolling page for Twitter (after almost scrolling 3k tweets), which is useful for certain valid users of scrapping (like overwatch of political discourse on twitter).

      So, can be infinite scrolling be useful, but not allowed by default on this social networks. Could we change the way information is visualized to get an overview of it instead of being focused on small details all the time in an infitite scroll tread mill.

  20. Sep 2016
    1. If efficiency incentives and tools have been effective for utilities, manufacturers, and designers, what about for end users? One concern I’ve always had is that most people have no idea where their energy goes, so any attempt to conserve is like optimizing a program without a profiler.
    2. This is aimed at people in the tech industry, and is more about what you can do with your career than at a hackathon. I’m not going to discuss policy and regulation, although they’re no less important than technological innovation. A good way to think about it, via Saul Griffith, is that it’s the role of technologists to create options for policy-makers.

      Nice to see this conversation happening between technology and broader socio-political problems so explicit in Bret's discourse.

      What we're doing in fact is enabling this conversation between technologist and policy-makers first, and we're highlighting it via hackathon/workshops, but not reducing it only to what happens there (an interesting critique to the techno-solutionism hackathon is here), using the feedback loops in social networks, but with an intention of mobilizing a setup that goes beyond. One example is our twitter data selfies (picture/link below). The necesity of addressing urgent problem that involve techno-socio-political complex entanglements is more felt in the Global South.

      ^ Up | Twitter data selfies: a strategy to increase the dialog between technologist/hackers and policy makers (click here for details).

  21. Jun 2016
    1. Also, the more complex a software project becomes, the more work you have to put into and it grows exponentially. So, keep it simple and make it fast. It's much easier to write software, throw it away and start over again quickly, than having this huge generic system that tries to do everything. It doesn't make sense. It's just too much work. You'd get this huge software system with thousand dependencies and, in the end, it's really hard to innovate, get new stuff in there, or, the worst case, to change the concept. Almost every software that we have published is not generic but is used only for one case. So, keep it simple and get a prototype in under three days.

      Agile visualization its a worthy exception to this trend. It is generic while being flexible and moldable. My first projects start with an easy prototype in a week and became full projects in a couple of months average. Then I can reuse the visual components by using abstraction and making visual builders.

      The couple of months average included the learning of the programming language and environment, the data cleaning and completion. With the builders the time has started to decrease exponentially.

    2. What type of team do you need to create these visualisations? 
OpenDataCity has a special team of really high-level nerds. Experts on hardware, servers, software development, web design, user experience and so on. I contribute the more mathematical view on the data. But usually a project is done by just one person, who is chief and developer, and the others help him or her. So, it's not like a group project. Usually, it's a single person and a lot of help. That makes it definitely faster, than having a big team and a lot of meetings.

      This strengths the idea that data visualization is a field where a personal approach is still viable, as is shown also by a lot of individuals that are highly valuated as data visualizers.

  22. Feb 2016
  23. Jan 2016
    1. UT Austin SDS 348, Computational Biology and Bioinformatics. Course materials and links: R, regression modeling, ggplot2, principal component analysis, k-means clustering, logistic regression, Python, Biopython, regular expressions.

  24. Nov 2015
    1. The effectiveness of infographics, or any other form of communication, can be measured in terms of whether people:

      • pay attention to it
      • understand it
      • remember it later

      Titles are important. Ideally, the title should concisely state the main point you want people to grasp.

      Recall of both labels and data can be improved by using redundancy -- text as well as images. For example:

      • flags in addition to country names
      • proportional bubbles in addition to numbers.
  25. Aug 2015
  26. Jun 2015
  27. Mar 2015
  28. Nov 2014
    1. This is an ongoing attempt at an algorithmically-generated, readability-adjusted scatter-plot of the musical genre-space, based on data tracked and analyzed for 1306 genres by The Echo Nest. The calibration is fuzzy, but in general down is more organic, up is more mechanical and electric; left is denser and more atmospheric, right is spikier and bouncier.