kyllo

Because it's a power law distribution and because wealth inequality is getting even worse over time.

Are you familiar with the edge popup algorithm introduced in "What's Hidden in a Randomly Weighted Neural Network?" https://arxiv.org/abs/1911.13299v2

Seems relevant to what you're working on. It starts with a randomly initialized, overparameterized neural net, but instead of gradient descent backpropagation, it learns by deleting connection edges.

The data types make sense to statisticians. Ordered factors are great when you need to fit, say, an ordered logit regression model.

A lot of programmers dislike R for the same reason they dislike PHP: weak typing.

If you can get over that, though, there's a lot to like about R. It's sort of like a hybrid between a Lisp and an array language.

Elixir runs on the Erlang VM, which is designed for distributed multi-process concurrency, right? Why would this be advantageous for interactive data analysis work, which is typically done on a single node? I don't quite understand the use case, hoping someone can explain.

That is also happening--lots of Hawaiians are being forced to move to the mainland because they can't afford housing in their home state. https://news.yahoo.com/native-hawaiians-flock-las-vegas-1357...

I lived in Korea in 2008-2010 and can confirm the ActiveX thing was still in widespread use then. It was a pain because every website effectively only worked in IE, meanwhile I had a Mac and couldn't log into any secure websites such as my local bank account from it.

I think part of the problem with that pitch is that the cost of an IT integration in this industry could pay for a lot of people in Asia to manage a forwarding operation in Excel. Ocean carriers don't have like, APIs. It's all custom integrations with EDI, SOAP, or CSV-over-FTP and you have to have a lot of tedious meetings with them to figure out how to map out and interpret the data they can send and receive.

Yes, a freight forwarder is sort of like a travel agent for freight--it's a middleperson role, in contrast to carriers, which are "asset-based" and operate vessels/planes/trucks.

Another significant cost factor for a freight forwarder is IT integrations with customers and carriers--many of which are very old-school and use EDI, SOAP, or even CSV-over-FTP to communicate shipment instructions and statuses.

I'm not sure to what extent Flexport implements these kinds of integrations with their partners, but there is definitely a significant hurdle to breaking even on investment in this kind of automation vs. just hiring clerks to process everything manually with e-mails and phone calls, especially in countries with lower labor costs.

Chilling effects are not unequivocally bad though. When the speech in question is advocating for stripping away the civil rights of or committing genocidal acts against certain groups of people, I'd much rather that speech be chilled.

data.table also uses OpenMP to parallelize operations, so it tends to be much faster

Why not? Honest question--I haven't used fastai myself but my understanding is it's just a high-level wrapper API for PyTorch. Any particular reason why you wouldn't want to use it outside of the course?

Yes, you can. There are algorithms for rounding up, rounding down, rounding to nearest, and banker's rounding, on the results of integer division. This is a solved problem.

I'm even seeing stories about U-Haul trucks all being rented out by tourists because rental cars are unavailable, causing a shortage of rental trucks for movers.

I think they're referring to the JIT compilation time. There are some ways to mitigate it, such as creating an image using PackageCompiler.jl, but it's definitely a noticeable issue IME. https://julialang.github.io/PackageCompiler.jl/dev/

Cool idea, I hope it grows. Would like to see more positions in Asian countries, they currently appear very limited.

I went from Pipenv to Poetry and then back to Pipenv after encountering an infinite loop in Poetry's solver on multiple separate occasions.

Yeah I think Rust's place in this ecosystem is a C/C++ replacement. All the Python and R data science packages call into native code to do the heavy lifting--the linear algebra, gradient descent etc. The next Tensorflow or PyTorch type framework could be implemented in Rust instead of C++. But it shouldn't matter to the end users--they would still use the Python or R bindings so they can have an interactive REPL environment to do their analysis in.

Julia is interesting because it's performant enough to implement these things directly--Julia's machine learning frameworks are written in pure Julia (with the exception of the calls to CUDA libraries for GPU) and can achieve near-native performance, so solving the "two language problem" is part of its value proposition. I do wonder how valuable it is for the whole stack to be written in one language, and whether that will blur the lines between the software engineers who implement data science packages and the data scientists who use them.

"Deep Learning in Clojure with Fewer Parentheses than Keras and Python"

Love it! :D What better way to define a neural network in code than an S-expression?

People go where the jobs are. They are going to live wherever employers "insist on" locating their offices.

20% down payment isn't a myth; if you don't put 20% down (and don't have some special situation like a VA loan) you have to pay for private mortgage insurance which can be hundreds per month, as you're buying the lender insurance against the possibility of you defaulting on your loan.

Are you talking about things like very small differences in the results of floating point calculations?

You lose a lot of the benefits, but you can still take advantage of time range partition elimination just as long as your data is still physically partitioned by the timestamp column. That's the most important thing when processing time series data--never read from disk any of the data that's outside the time range you actually need for your query.

Is this generally the case across all major C standard library implementations or are some of them much faster than others?

Still around the Seattle Area, but a cabin deep in the Cascade mountains where I could go hiking every day, or snowshoeing in the winter.

SQLite was started by Dwayne Richard Hipp from North Carolina. 20 years ago.

Yep, compressibility is another big win made possible by columnar layout, and it looks like DuckDB implements at least some form of column compression.

Yes, it's geared toward reporting / analytics / data science workloads where the most common read operation is calculating aggregate metrics like count/min/mean/max and histograms. Statistics stuff.

Ah, good to know. I saw it when someone I follow retweeted it yesterday.

Perhaps, but the point is, if you're working in DevRel and you think your role is to defend the product from criticism, and you do it by placing the onus on the developer to figure out how to use your product safely, you've totally lost the plot.

DuckDB is developed by the same database research group at CWI that developed MonetDB, so they have applied a lot of the lessons they learned from MonetDB.