domoritz

For another grammar-of-graphics-based visualization library (flexibly compose charts rather than simply pick a template), check out Altair https://altair-viz.github.io.

You might already do this, but great opportunity to support them with a donation.

Arxiv can recompile latex to support accessibility and html. Going to pdf submissions would be a major step backward.

The UX has many rough edges especially in the default element apps. UX was the primary reason my university department passed on switching to matrix from slack so far.

I suspect most use of columnar databases is OLAP, which is different from what people usually mean when they say time series data.

Google produced a podcast introducing ML concepts with the same game a few years ago. https://pair.withgoogle.com/thehardway/

So many data probes would be solved if everyone looked at a few outputs instead of only metrics.

Urg, the bar charts to not start at 0. It's making it impossible to compare across model sizes. That's a pretty basic chart design principle. I hope they can fix it. At least give me consistent y scales!

I don't understand why there isn't an obvious, visible watermark at all. Yes, one could remove it but let's assume 95% of people don't bother removing the visible watermark. It would really help with seeing instantly when an image was AI generated.

I like the idea of having types at runtime for parsing etc and generating validators in various languages. What stopped me from going there so far is that I already have TypeScript types provided for the various libraries I use. How good are the tools for importing TypeScript types into ArkType/Zod and working with types in various representations in parallel?

But they are heavier. Not saying it offsets the effect but it's just not straight forward to assume either way.

Author of this paper here. Cool to see this show up on HN. Little story about this paper: Danyel and I did this while I was on an internship at Microsoft Research. The algorithm was too simple to get it published at a conference, but the technique has been adopted in a few places now, including other papers. We did some follow-up work to more efficiently compute the density and also support interactive filtering. See https://www.domoritz.de/papers/2021-KDBox-VIS.pdf.

And also not UI layout generation.

Altair works with Polars. The built-in plotting library for Polars is Altair actually. https://docs.pola.rs/user-guide/misc/visualization/

I really hope most people are not just trying to satisfy regulatory requirements but genuinely want to allow people using screen readers to understand what an image is about. Having said that, AI can definitely be useful when there is no alt text but it shouldn’t be an excuse not to add alt texts as a creator.

Semantic scholar is pretty good so I keep using it more and more.

I love the simplicity of this. Hurray for small models for small tasks.

I'd be scared to spell it with one M or two Ts.

Aiegradient is great and they make indoor and outdoor monitors. Fully open source. They are the authors of this article and to a lot of research and advocacy which is great.

Awesome to hear that you like Vega Altair. With the recent integration of VegaFusion you don’t need to embed the data in the notebook anymore and I’ve found Altair to scale quite well. Give it a shot.

matrix.org is a neat open protocol/platform but clients have some room to improve.

This looks interesting but I already have TypeScript types for my APIs so I developed https://github.com/vega/ts-json-schema-generator which lets me generate JSON schema from the sources directly. Yes, it does have some oddities because the two languages have slightly different feature sets but it’s been working well for us for a few years. If I didn’t have TypeScript or a smaller API surface I’d be okay with typing again I would look at TypeSpec though. It definitely beats writing JSON schema by hand.

One of the ArrowJS committers here. We have fixed a few significant performance bottlenecks over the last few versions so try again. Also, I'm also ways curious to see specific use cases that are slow so we can make ArrowJS even better. Some limitations are fundamental and you may be better off converting to the corresponding JS types (which should be fast).

We are working on GPU accelerated rendering for Vega and Altair.

In https://github.com/vega/vega-lite/issues/408#issuecomment-50... Leland Wilkinson himself said “having gone through a lot of these GG-inspired systems, I believe yours is the most authentic implementation. I'm using it every day. Thanks for all the great work you've done.”

Still, generating energy even with solar has impact on natural resources and the environment.

Streamlit also has caching mechanisms but it's specified explicitly on calls: https://docs.streamlit.io/library/advanced-features/caching.

Only in the worst case and large N. For practical problems, it’s solvable for tens of thousands of nodes.

Vega like ggplot (which this is very close to) build on the grammar of graphics but differ in the targets they support (Vega is using web technologies), language design details (e.g., Vega and ggplot use different terminology for the graphical elements and the encodings), and supported features (Vega has a reactive dataflow and support for building custom interactions). There is a lot more to be said but that’s a start for how they compare.

It’s great that PPS can reduce PM2.5 and CO2 to extremely low levels. I live in an area with industrial pollution (coal etc) and am concerned about VOCs, NOXs, SO2s etc as well. Wouldn’t I bring in more of those if I constantly bring in outside air even when the pollution is high outside? Unfortunately I haven’t been able to exactly identify the chemical that’s causing issues for me yet. I’d love to hear your thoughts on this situation.