huyegn

While these things are not difficult to solve with some minor additions to the standard library and documentation that emphasizes some basic conventions, the fact that we're on Swift6 and things like this aren't just available speaks to how much further the community needs to evolve to get the language ergonomics right for non apple-y things.

Comment · 6 years ago

Meanwhile a man from the Westerdam cruise ship, where one passenger already tested positive, escapes quarantine, books ubers, and multiple flights, is arriving back in Seattle:

https://komonews.com/news/local/man-stranded-on-coronavirus-...

Some people can't get more selfish. Also, what a colossal failure on the part of customs and immigration.

Comment · 6 years ago

As someone who has done frontend dev on the web, I find SwiftUI refreshingly good.

Using SwiftUI is the first time ui programming on iOS has clicked for me.

For anyone getting into this, I suggest you take a look at apple's very well done tutorial series before using other sources:

https://developer.apple.com/tutorials/swiftui/tutorials

I spent about 3 days to go through the entire series and the time investment was well worth it.

Comment · 6 years ago

I stated that rather than explainability, for most use cases, people just want calibrated and robust uncertainty estimates NOT low quality and uncalibrated uncertainties estimates.

Your example points to models that provide low quality uncertainty estimates, but that's not true for all deep learning models.

I believe it's these low quality uncertainty estimates that lead people to look toward "explainability" as a solution, but for the majority of use cases, I think people just want better uncertainty estimates so that they can "know when they're model doesn't know".

There are techniques now to get higher quality, calibrated, uncertainty estimates that don't suffer from the problems you mentioned and I've outlined these solutions in my posted link above.

Additionally if you're interested, there is some nice recent research from google on the subject:

https://ai.googleblog.com/2020/01/can-you-trust-your-models-...

and from oxford:

https://arxiv.org/abs/1912.10481v1

Comment · 6 years ago

That's an interesting idea and I get the motivation. I often wonder if we're already able explain predictions. There are things like gradcam to determine which inputs caused which activations. It's just that the explanations aren't very understandable or satisfying from a human perspective.

Comment · 6 years ago

I agree with Yann here ... I think the demand for explainability is like a person asking for a "faster horse" when what they really need is a car.

When people ask for explainable models, what they really want (in my opinion) is calibrated and robust uncertainty estimates .

Good uncertainty estimates would let them know when to trust a model's prediction and when to ignore it.

For example, a model trained to predict dog breeds should know nothing about cat breeds, and there should be some way to quantify when it doesn't know!

I've been doing a review of techniques that are becoming more popular in this area:

https://blog.everyhue.me/posts/why-uncertainty-matters/

Comment · 12 years ago

Absolutely agree with this.

Another nice illustration of how when arrival rate is greater than departure rate we get overflowing queues:

http://setosa.io/blog/2014/09/02/gridlock/index.html

Comment · 12 years ago

Wow, I just read through the documentation for the Nix package manager and it's a breath of fresh air...Source based builds but with the convenience of pre-compiled binaries hashed with all the input parameters used to create those binaries:

https://nixos.org/nix

Has anyone tried using this for development?

Comment · 15 years ago

To get around the problem of child subprocesses spewing out too much output and blocking the parent process, one can provide an open file handle to the stdout/stderr arguments of the Popen call. I've ran into this many times and this solution has reliably worked for me every time. This could be documented better in the Python docs.

For quick tasks and scripts, I've found subprocess.check_call, and subprocess.check_output with shell=True are great tools for spawning subprocesses and quickly grabbing output. They're pretty straightforward to use.

Comment · 15 years ago

We're also looking into getting a designer for our site, I'd be interested in hearing how you managed the project.

When you started with your designer, how did you convey the type of look and feel that you wanted to give your business? Did you point them to other similar looking sites or was he/she the one that pushed forward the final vision and design (which BTW looks great)?

Comment · 15 years ago

We used this technique once in one of our twisted servers to quickly fix in-memory data structures that had been loaded from old sources.

Recently, my team has been developing most of our newer network services using gevent. It's probably the closest thing to erlang that python has when it comes to programming asynchronous servers.

One of the things I always missed was this manhole feature, but while rummaging around the other day, I found this:

http://www.gevent.org/gevent.backdoor.html

Haven't tried it yet, but it looks like it will achieve the same task :)

Comment · 15 years ago

If you liked Kernighan and Ritchie's "The C Programming Language", then you'll love "Programming In Lua". The first edition is available online here:

http://www.lua.org/pil/

Like K&R, it's style is colloquially terse, yet each sentence is chock full of information. It's a great book for someone who's already coming from another language like Python.

If you're into graphics programming, then http://love2d.org/ would be the perfect vehicle for you to get into Lua while you read PIL.