ellisv

33 ms·

361.
by ellisv 5y ago
This is interesting to me but the motivation behind this is unclear. Since FastAI.jl uses Flux, and not PyTorch, functionality has to be reimplemented. FastAI.jl has vision support but no text support yet. What does this mean for the develo
362.
by ellisv 5y ago
this works fine for me import Dates Dates.today()
363.
by ellisv 5y ago
It’s in outer space
364.
by ellisv 5y ago
This article strikes me as odd. The premise is based on a "pattern of behavior" but to me it feels the author is attributing malice where there likely is none. > For the last nine-months I have noticed that the term Tidyverse i
365.
by ellisv 5y ago
Ah CV not ML? No CNNs in there?
366.
by ellisv 5y ago
and some of those nails don’t need to be hammered!
367.
by ellisv 5y ago
Is that because the ML models aren’t useful or because of integration problems?
368.
by ellisv 5y ago
It’s elegant. It doesn’t overreach.
369.
by ellisv 5y ago
In many situations, few or none of the observations will have the average value. It’s been discussed here before — search for the “When U.S. air force discovered the flaw of averages” story
370.
by ellisv 5y ago
Agreed. I’ve seen that figure before (or some variation) and it is a bad way to try to argue his statement.
371.
by ellisv 5y ago
The Fourier transform layer might not be the immediate hidden layer.
372.
by ellisv 5y ago
That's a fair point. My assumption was that the neural network isn't going to be as efficient at the FFT but the article doesn't include that type of assessment.
373.
by ellisv 5y ago
Neural networks can learn many things but I can't imagine a practical use-case of this.
374.
by ellisv 6y ago
I participated in IRB for a UMN campus and that’s pretty much how these things worked back then.
375.
by ellisv 6y ago
This sort of incident is difficult to set expectations for reporting back. I expect UMN to move swiftly on this but it’s a large university and may move more slowly than we’d like.
376.
by ellisv 6y ago
Most of published statistics are not performed by statisticians.
377.
by ellisv 6y ago
Seems like an oddly specific critique. The paper here doesn’t concern itself with timeseries specifically (nor spatial, spatiotemporal, etc).
378.
by ellisv 6y ago
Soft Skills is my new favorite podcast. I've been going through all the old episodes.
379.
by ellisv 6y ago
> but I think it's a solid principle that you shouldn't assume your database will retain its schema indefinitely. this is not a hard thing to anticipate and yet my coworkers thought otherwise...
380.
by ellisv 6y ago
I feel very funnel/inverse-triangle shaped.
381.
by ellisv 6y ago
It boggles my mind that in 2021 IPython/Jupyter notebooks _still_ don't have inline code evaluation or sensible parameterization without using extensions.
382.
by ellisv 6y ago
Any of it ready to share?
383.
by ellisv 6y ago
it looks like it redacts all the text?
384.
by ellisv 6y ago
and I'm sitting where wondering, why would you use Airflow over Argo Workflows? Everyone should evaluate the options for their own needs.
385.
by ellisv 6y ago
I disagree with Geyer on this. Burn-in may not be necessary but often is for the type of inference that most people are familiar with. But these days, who cares? I can fit a highly parameter used model on my 6 year old MacBook Pro with Stan
386.
by ellisv 6y ago
I think this is a port of the dplyr syntax/verbs only. None of the dplyr C++ code is used, it's all syntactic sugar on top of Pandas. So any SQL translation is going to be similar to what Pandas does.
387.
by ellisv 6y ago
> What I think would be a nice approach would be someone creating a new Dataframe class that merely inherits from a Pandas dataframe but just implementing .filter(), .summarise(), .select() etc. as methods, using the dplyr syntax. If the
388.
by ellisv 6y ago
How's the performance? I'm certainly willing to give up a little computational performance for being able to write my code faster.
389.
by ellisv 6y ago
From what I can glean from the docs, you run the UI locally which reads from experiments from `.aim/` The experiments results could be populated by running the training job remotely and committing the `.aim/` directory afterwards.
390.
by ellisv 6y ago
It's not really that different from deploying any other sort of RESTful API. If you want a framework you could use Seldon but could probably just as easily write your own with Flask/fastapi.

More ›