Nick Haynes

Machine learning, statistics, software engineering. Occasionally, I do other things, too.

[Nick Haynes @NickDHaynes]

@bbstats op is not a good take but

[Nick Haynes @NickDHaynes]

@bbstats Proud of you, son.

[Nick Haynes @NickDHaynes]

@bbstats new plot who dis

[Nick Haynes @NickDHaynes]

@tublitzed Torties are the best!

[Nick Haynes @NickDHaynes]

Listening to the most recent @strongerbysci fireside chat that talks about data analysis errors in Excel, and open up my computer this morning to find... 🤦

@GregNuckols
https://t.co/bXNYFoG56P

[Josh Tobin @josh_tobin_]

What makes production ML hard?
- Cleaning, labeling, and augmenting data
- Troubleshooting training and ensuring reproducibility
- Deploying models and monitoring their real-world impact

To help, we're excited to announce our online production ML course:
https://t.co/CWWDjLC6Mh

[Nick Haynes @NickDHaynes]

@AdamBSmith +ve for SMB: shorter sales cycle = you learn what works (with messaging, pricing, product, etc) and what doesn't faster

[hardmaru @hardmaru]

Many people start attacking a problem by deploying the most sophisticated method possible with the belief that it will lead to the best results. I believe it might be better to chisel away at the complexity and find the simplest possible combination that expresses the key ideas.

[Nick Haynes @NickDHaynes]

@stanfordnlp Whoa whoa, chill. I agree with their approach - it's a great paper!

[Nick Haynes @NickDHaynes]

Important to remember that BLEU (or ROGUE, or other quantitative metrics) don't tell the whole story - two texts w/ same metrics can have very different perceived levels of quality to human readers. https://t.co/NKu2qNct0H

[Stanford NLP Group @stanfordnlp]

We’d still put this in the hyping AI bucket and don’t really believe the translations are as good as a careful human’s. Nevertheless, this paper does examine and discuss measuring human parity much more carefully than most #dlearn papers that claim super-human performance…. https://t.co/HI0DEGdwoo

[NYC Media Lab @nycmedialab]

Teams + Fans + Tech panel moderated by @YahooSports Michael Beshar looks at how new tech enables teams to build fan affinity & drive broader consumption of content @jaynebwise @nakanofrank #SMT18

[Denny Britz @dennybritz]

Maybe progress in Vision/NLP/etc is not due to neural nets, but rather due to the influx of smart researchers, lots of compute, and industry pushes.

Focusing on a different paradigm (graphical models, evolution, ..) could’ve made the same amount, but different type, of progress.

[Nick Haynes @NickDHaynes]

The fundamental irony of Twitter: the short format allows your audience to exist, but bc the audience exists, you want a longer format.

[Nick Haynes @NickDHaynes]

And it's hard to be concise with something that feels incredibly important to you.

[Nick Haynes @NickDHaynes]

So when you have something (seemingly) important to say, you want to say it to the largest possible audience.

[Nick Haynes @NickDHaynes]

The reason tweetstorms exist is that, for the vast majority of people, Twitter is the largest audience available to you.

[Nick Haynes @NickDHaynes]

GO PACK GO https://t.co/LMhg4d3z2A

[Nick Haynes @NickDHaynes]

"as a general principle of learning, it 'somehow smells right.'" Well, I'm convinced.

https://t.co/a6sDJBIjYg

[François Chollet @fchollet]

The belief that stacking layers deeper leads to better models is just a myth pushed by the big DL lobby to sell more tensors