Include Hadley's embedded plots paradigm for analyzing large and complex datasets?

Apparently ggvis is leafrogging regular faceted plots to instead include “Embedded plots”. I’ll let hadley explain:

“This article describes a class of graphs, embedded plots, that are particularly useful for analyzing large and complex datasets. Embedded plots organize a collection of graphs into a larger graphic, which can display more complex relationships than would otherwise be possible. This arrangement provides additional axes, prevents overplotting, and allows for multiple levels of visual summarization. Embedded plots also preprocess complex data into a form suitable for the human cognitive system, which can facilitate comprehension. We illustrate the usefulness of embedded plots with a case study, discuss the practical and cognitive advantages of embedded plots, and demonstrate how to implement embedded plots as a general class within visualization software, something currently unavailable. This article has supplementary material online.”

Visual example here: http://blog.revolutionanalytics.com/2012/09/visualize-complex-data-with-subplots.html

Insofar as bokeh is meant to aid exploration of large and complex datasets, are there plans to include such facilities?

Thanks

Full text of the paper can be found here: http://vita.had.co.nz/papers/embedded-plots.html

···

On Thursday, April 2, 2015 at 2:07:19 PM UTC-4, datnamer wrote:

Apparently ggvis is leafrogging regular faceted plots to instead include “Embedded plots”. I’ll let hadley explain:

“This article describes a class of graphs, embedded plots, that are particularly useful for analyzing large and complex datasets. Embedded plots organize a collection of graphs into a larger graphic, which can display more complex relationships than would otherwise be possible. This arrangement provides additional axes, prevents overplotting, and allows for multiple levels of visual summarization. Embedded plots also preprocess complex data into a form suitable for the human cognitive system, which can facilitate comprehension. We illustrate the usefulness of embedded plots with a case study, discuss the practical and cognitive advantages of embedded plots, and demonstrate how to implement embedded plots as a general class within visualization software, something currently unavailable. This article has supplementary material online.”

Visual example here: http://blog.revolutionanalytics.com/2012/09/visualize-complex-data-with-subplots.html

Insofar as bokeh is meant to aid exploration of large and complex datasets, are there plans to include such facilities?

Thanks

Haven’t had a chance to read the paper, but skimmed over it. To me, this looks like the use of the more complete specification for faceting, as defined in grammar of graphics. Since faceting is just the plotting of plots into another coordinate system, this can be generalized. I did a bit of work towards implementing a more complete and modular faceting approach, and hope I get some more time to work on it. It is currently in a development branch.

Typically, we use some form of grid faceting, whether that is by crossing the two facets, or by wrapping a single facet into a grid. We do have implemented an alternative dimension for faceting, which is facet tab, implemented in crossfilter. You can combine the grid-based approach and tab approach to spread dimensions out into separate groups of graphs.

I think the only challenge is making sure the plotting capability supports the embedded graphs, but the general approach is implemented.

···

On Thursday, April 2, 2015 at 1:14:42 PM UTC-5, datnamer wrote:

Full text of the paper can be found here: http://vita.had.co.nz/papers/embedded-plots.html

On Thursday, April 2, 2015 at 2:07:19 PM UTC-4, datnamer wrote:

Apparently ggvis is leafrogging regular faceted plots to instead include “Embedded plots”. I’ll let hadley explain:

“This article describes a class of graphs, embedded plots, that are particularly useful for analyzing large and complex datasets. Embedded plots organize a collection of graphs into a larger graphic, which can display more complex relationships than would otherwise be possible. This arrangement provides additional axes, prevents overplotting, and allows for multiple levels of visual summarization. Embedded plots also preprocess complex data into a form suitable for the human cognitive system, which can facilitate comprehension. We illustrate the usefulness of embedded plots with a case study, discuss the practical and cognitive advantages of embedded plots, and demonstrate how to implement embedded plots as a general class within visualization software, something currently unavailable. This article has supplementary material online.”

Visual example here: http://blog.revolutionanalytics.com/2012/09/visualize-complex-data-with-subplots.html

Insofar as bokeh is meant to aid exploration of large and complex datasets, are there plans to include such facilities?

Thanks

Responded from my work address, but here are some of the faceting-related functionality in bokeh now, or are in work:

Faceting in CrossFilter: https://github.com/bokeh/bokeh/blob/master/bokeh/crossfilter/models.py

My faceting branch, which aims to factor out the faceting functionality: https://github.com/rothnic/bokeh/commits/facet

As of right now, my faceting branch can produce a stand-alone, faceted grid of plots. I struggled a bit at first with the notation to use, but ended up just implementing the grid-based approach for now, and plan to generalize it a bit later. FacetGroup was meant to be that generalized faceting implementation, but it currently has a 2d cartesian coordinate system built into it. I think it probably needs to just support any coordinate system to be able to implement the referenced behavior in the white paper.

You end up wanting to plot and possibly scale the plot, just as you would a point, so the thing doing the plotting doesn’t care that it is actually plotting a plot (instead of some other geom) into its coordinate system.

···

On Friday, April 3, 2015 at 9:54:08 PM UTC-5, [email protected]ems.com wrote:

Haven’t had a chance to read the paper, but skimmed over it. To me, this looks like the use of the more complete specification for faceting, as defined in grammar of graphics. Since faceting is just the plotting of plots into another coordinate system, this can be generalized. I did a bit of work towards implementing a more complete and modular faceting approach, and hope I get some more time to work on it. It is currently in a development branch.

Typically, we use some form of grid faceting, whether that is by crossing the two facets, or by wrapping a single facet into a grid. We do have implemented an alternative dimension for faceting, which is facet tab, implemented in crossfilter. You can combine the grid-based approach and tab approach to spread dimensions out into separate groups of graphs.

I think the only challenge is making sure the plotting capability supports the embedded graphs, but the general approach is implemented.

On Thursday, April 2, 2015 at 1:14:42 PM UTC-5, datnamer wrote:

Full text of the paper can be found here: http://vita.had.co.nz/papers/embedded-plots.html

On Thursday, April 2, 2015 at 2:07:19 PM UTC-4, datnamer wrote:

Apparently ggvis is leafrogging regular faceted plots to instead include “Embedded plots”. I’ll let hadley explain:

“This article describes a class of graphs, embedded plots, that are particularly useful for analyzing large and complex datasets. Embedded plots organize a collection of graphs into a larger graphic, which can display more complex relationships than would otherwise be possible. This arrangement provides additional axes, prevents overplotting, and allows for multiple levels of visual summarization. Embedded plots also preprocess complex data into a form suitable for the human cognitive system, which can facilitate comprehension. We illustrate the usefulness of embedded plots with a case study, discuss the practical and cognitive advantages of embedded plots, and demonstrate how to implement embedded plots as a general class within visualization software, something currently unavailable. This article has supplementary material online.”

Visual example here: http://blog.revolutionanalytics.com/2012/09/visualize-complex-data-with-subplots.html

Insofar as bokeh is meant to aid exploration of large and complex datasets, are there plans to include such facilities?

Thanks

This looks like great work! I think bokeh has great exploratory analysis potential and am looking forward to helping and seeing it be fleshed out.

···

On Monday, April 6, 2015 at 9:06:51 AM UTC-4, Nick Roth wrote:

Responded from my work address, but here are some of the faceting-related functionality in bokeh now, or are in work:

Faceting in CrossFilter: https://github.com/bokeh/bokeh/blob/master/bokeh/crossfilter/models.py

My faceting branch, which aims to factor out the faceting functionality: https://github.com/rothnic/bokeh/commits/facet

https://github.com/rothnic/bokeh/blob/b107e88f8d190ea0eb199b334a35e327cdc414cd/bokeh/crossfilter/facets.py

As of right now, my faceting branch can produce a stand-alone, faceted grid of plots. I struggled a bit at first with the notation to use, but ended up just implementing the grid-based approach for now, and plan to generalize it a bit later. FacetGroup was meant to be that generalized faceting implementation, but it currently has a 2d cartesian coordinate system built into it. I think it probably needs to just support any coordinate system to be able to implement the referenced behavior in the white paper.

You end up wanting to plot and possibly scale the plot, just as you would a point, so the thing doing the plotting doesn’t care that it is actually plotting a plot (instead of some other geom) into its coordinate system.

On Friday, April 3, 2015 at 9:54:08 PM UTC-5, [email protected] wrote:

Haven’t had a chance to read the paper, but skimmed over it. To me, this looks like the use of the more complete specification for faceting, as defined in grammar of graphics. Since faceting is just the plotting of plots into another coordinate system, this can be generalized. I did a bit of work towards implementing a more complete and modular faceting approach, and hope I get some more time to work on it. It is currently in a development branch.

Typically, we use some form of grid faceting, whether that is by crossing the two facets, or by wrapping a single facet into a grid. We do have implemented an alternative dimension for faceting, which is facet tab, implemented in crossfilter. You can combine the grid-based approach and tab approach to spread dimensions out into separate groups of graphs.

I think the only challenge is making sure the plotting capability supports the embedded graphs, but the general approach is implemented.

On Thursday, April 2, 2015 at 1:14:42 PM UTC-5, datnamer wrote:

Full text of the paper can be found here: http://vita.had.co.nz/papers/embedded-plots.html

On Thursday, April 2, 2015 at 2:07:19 PM UTC-4, datnamer wrote:

Apparently ggvis is leafrogging regular faceted plots to instead include “Embedded plots”. I’ll let hadley explain:

“This article describes a class of graphs, embedded plots, that are particularly useful for analyzing large and complex datasets. Embedded plots organize a collection of graphs into a larger graphic, which can display more complex relationships than would otherwise be possible. This arrangement provides additional axes, prevents overplotting, and allows for multiple levels of visual summarization. Embedded plots also preprocess complex data into a form suitable for the human cognitive system, which can facilitate comprehension. We illustrate the usefulness of embedded plots with a case study, discuss the practical and cognitive advantages of embedded plots, and demonstrate how to implement embedded plots as a general class within visualization software, something currently unavailable. This article has supplementary material online.”

Visual example here: http://blog.revolutionanalytics.com/2012/09/visualize-complex-data-with-subplots.html

Insofar as bokeh is meant to aid exploration of large and complex datasets, are there plans to include such facilities?

Thanks