Is agile the right approach for cross-functional teams of AI researchers and software developers? Here's what I learned from 10 years of leading such teams
Very cool to see this – but will only be successful with great terminal plotting tools. The ones the author mentions like the matplotlib interface clearly won't do[0] due to lack of resolution
A perfect use case for unicode plotting [1] (shameless plug)
I spent a hot second last year thinking about sparklines in a terminal, and it seemed to me that unicode could use some more glyphs to improve this situation of treating a single glyph as a modest array of pixels. I don't know what I googled for, but for some reason I found only about 1/3rd of the characters listed in the final two lines of
Unless I'm wrong about your demo, it seems like you are using some of these? There's a 2x2 pixel grid, and some characters that would be good for bar charts, both horizontal and vertical, and also with negative numbers. That's better than I thought we had, but not a big palette to work with.
Are you aware of any proposals to expand the pixel art capabilities of Unicode? Grids have the problem of requiring 2^(X x Y) characters to represent all states, but diacritical marks don't have that scaling problem. For example it would take 64 characters to represent a 2x3 grid of all on/off states, whereas a 6x8 grid of rectangular diacritical marks would only take 48 glyphs.
That's right, I'm using the 2x2 characters and some off-center horizontal line characters.
Note aware of Unicode expansions, though more and more fonts seem to support the above.
Sparklines in the terminal sounds very cool, did you end up coding something?
Note that one other popular alternative (see for example the Unicode plotting lib for Julia) is using Braille characters, which probably have an even better support across fonts
> Sparklines in the terminal sounds very cool, did you end up coding something?
It seemed to me that the existing tools were about as good as I was going to achieve on my own. I got distracted from my use case and will have to circle back, hopefully soon.
I am a big fan of https://github.com/Evizero/UnicodePlots.jl . It is unexpected how useful these tools can be to track the evolution of numerical experiments/developments right on the terminal.
PS: Is the general case of a background image handled by Uniplot, and does Uniplot allow for the idea of taking a PDF, putting the plot data into it, and then creating a new PDF file ?
Thanks for the question! So the labels only depend on the limits of the plotting window, which you can supply as a keyword option. In that sense, and together with using the same gridlines options, you could have the same background in some sense.
What uniplot cannot do today is to have multiple vertical axis (like weight and head circumference in your example).
Hope this helps!
P.S.: Regarding your PDF question, for that in fact I think that some plotting library to graphics (rather than the terminal) might be better
So, the question isn't really about the axes. The question is about the curved lines that are part of the background. These lines allow a medical person to know what the "normal" curves are for a child, and then the plotted points are compared to how well they match the curves (by observation) This tells the medical person if the child is growing normally, or if they are malnourished or have some other problem that should be watched to explain the unusual lack or presence of growth.
Since the data points may be gathered at any time, the possible creation of a combined PDF will allow the medical record to include which chart was used and when, so that folks in the future have a better handle on the history of that child.
Kind of agree, but I often use Jupyter just to capture the command history and its (textual) output. For someone like me this is still useful even without the graphical part of Jupyter.
Speaking as a pytorch user, many of the steps in your Readme example remind me of the usual setup except the pipeline.run() handoff which is replaced by eager evaluation in pytorch.
Are you seeing something like an eager mode for your library, or perhaps a pytorch plugin that might use your apis?
Definitely! It might sound like a sales-thing to say, but PyTorch in general is a huge point on our roadmap, but we’re not 100% sure what the most logical approach is. To keep the APIs consistent will involve some fiddling on our end.
I'm guessing that's because they measured the return rate of institutions:
"Wallets were returned to one of five societal institu-
tions: (i) banks, (ii) theaters, museums, or other cultural establishments, (iii) post offices, (iv) hotels, and (v) police stations, courts of law, or other public offices."
That's not very intuitive for me. Let's do some limit analysis: imagine the groups were one million sessions each, but the convertions in the groups were only one and two people respectively. Wouldn't this result seem like the result of random chance?
The conversion rate is basically one in a million in both cases.
This is a form of philosophy specifically known as causal processes.
[1] contains a good overview of this that presents Russel and critics (Salmon et. al). (I prefer that philosophy be taken in context rather than from the mouth of a particular philosopher).
