Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

feedback on the graph visualization #106

Open
percyliang opened this issue Mar 8, 2023 · 0 comments
Open

feedback on the graph visualization #106

percyliang opened this issue Mar 8, 2023 · 0 comments
Assignees

Comments

@percyliang
Copy link
Contributor

percyliang commented Mar 8, 2023

Dimitris:

  • it would be nice if different nodes had different sizes. Maybe datasets should be scaled by size or out-degree?
  • i guess at a higher-level, maybe it would be worth thinking what takeaways we want from the graph (e.g., the disconnected components are cool) and then thinking about how to visualize them
  • it would be nice to be able to quickly zoom in to specific assets, maybe a sidebar listing models/datasets (as tabs) where I can click on a model and visually jump to it’s neighborhood (I was trying to find the Pile and it took me a fat minute)
  • it would be nice if you could isolate subgraphs: what lies downstream of the Pile?
  • maybe it’s worth precomputing the graph arrangement? having things jump around was fun the first time, but annoying when I reloaded the page
  • is there a way to initially display a coarse structure? maybe “PaLM and friends” so that i can take away something at a glance instead of having to parse the entire graph?

Ansh:

  • Seems like the canvas size has a fixed height of 600, it looks particularly small on bigger screens. This should be a percentage value.
  • Having a legend for node types and edge types to understand the relationships between the nodes (since graph is heterogenous) would be nice.
  • Agree with Dimitris’s comment about the coarse structure, since the graph looks a bit dense and spread out. The coarse structure can be expanded into the main graph upon double-clicking (like it happens in TensorFlow computation graphs)
  • I think it would be nice to have a quick navigation side-bar with the important anchor nodes. These could be two-level, dataset/model/application and then a search-able list of corresponding nodes.
  • The action for panning around the graph and moving the nodes and edges around is same - click and drag. This can make navigation cumbersome sometimes if your cursor is accidentally over the node.
  • Instead of the click opening the page corresponding to the entity - it would be nice to have a flashcard pop-up with the core information and a link for more details. It would be nice to have some basic info while checking out the graph.

Yifan:

  • Works well on mobile; that's good!
  • Not sure why it tries to do a force directed layout upon first load, before snapping to the final layout (which seems not force directed)
  • Not sure if moving nodes should even be supported; I haven't seen this in other graph visualizations
  • Some graph layout ideas: maybe try to group nodes of the same type, or maybe try to lay out most directed arrows from top to bottom
  • Would be good to have zoom and and out buttons, since scroll wheel is not always available on all systems
  • Consider not using red on the color scheme (has a strong "danger" connotation)
  • Clicking should open a infobox instead of going to another page
@percyliang percyliang changed the title Dimitris's feedback on the graph visualization feedback on the graph visualization Mar 8, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants