MORE PARAMETERS DON’T AWAYS LEAD TO MORE ACCURACY
(thread)

Italy assigned to each region a COVID risk level based on a complex model with 21 parameters.

Many experts claimed that the more parameters, the more precise the model.

Is it true, though?

(1/6)
2/ In most real-world cases, more parameters lead to worse predictions.

See the example below (from my gum.co/heuristic) ImageImageImage
3/ If we order the 21 parameters by descending predictive power, and apply it to the data used to build the model, we discover that:
- The first parameter alone predicts the outcome of many regions.
- The last parameter, at most, helps predicting the outcome of a single region.
4/ With an informal reversal, we could say that the first parameter is validated using many data points, whereas the last parameter is validated using a single data point - which might be noisy!

(For detailed explanation, see the screenshots above.)
5/ Hence, no, it’s not a given that more parameters in a predictive model increase its accuracy.

Often, they decrease it.
6/ Humans and animals use heuristics and rules of thumb not (just) because they are fast, but (also) because they have very few parameters and are thus very resistant to noise.

Whereas complex models are prone to overfitting to data which is not representative of the future.

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Luca Dellanna

Luca Dellanna Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @DellAnnaLuca

17 Nov
ON STUDENT DEBT
(thread)

1/ The worst possible outcome is forgiving student debt without changing the structural problems that led to this situation.

This would only make tuitions higher & reduce access to higher education, as described below.



[1/9]
2/ As a general principle, solving the symptom of a problem without solving the underlying causes is a terrible idea. It removes the urgency to solve the problem once and for all, making it grow larger.
3/ One question worth asking is, why is student debt not a problem in Europe?

The naive answer: government subsidies to students.

The real answer: contained costs.

The total operational costs for my ex-university in 2018 were "just" $7.8k / student / year.
Read 11 tweets
12 Nov
THE FUTURE OF NEWSLETTERS
(and announcing my new newsletter 🎉)

Too much noise, too little content we actually use.

Take your favorite newsletter. Can you remember the contents of the edition-before-the-last-one?

Me neither. But I have a solution

(thread, 1/N)
2/ Today, I launch the RoamLetter.

A newsletter whose content directly integrates into your note-taking system.

A newsletter whose editions AUTOMATICALLY link with each other *and with your notes* 👏

A newsletter with built-in spaced repetition.

A timeless newsletter.
3/ If you are a Roam user, you'll enjoy how the topics of one edition of my newsletter automatically link with your body of knowledge, and the other way around.

If you are not a Roam user, no worries. You can still use all the other features of my newsletter (pics below)
Read 13 tweets
4 Nov
ROAM BOOKS ARE THE FUTURE OF EBOOKS
(thread)

3 days ago, I published the first @RoamResearch book ever (gum.co/ergodicity).

Once you try this format, you won't come back.

Here are 5 of its features (thread, 1/6)
2/ The first feature of Roam books, or rBooks for short, is that they allow for non-linear exploration.

Their pages are like Wikipedia pages – they contain links for optional deeper exploration.

Moreover, at the bottom of each page, there is an automatic list of references.
3/ Roam books are still books – they still have a table of content that guides you through your journey of discovery.

However, you can stray off the path or take shortcuts as you please.
Read 12 tweets
2 Nov
MASKS WORK EVEN IF THE VIRUS IS SMALLER THAN THE HOLES IN THE FABRIC

As a first effect, the virus bounces on the fibers. That's enough to decrease the distance at which it "jumps out". It might even get some of the particles to stuck to the fibers.

(thread, 1/N)
2/ Second, N95 masks also have electrostatic charges that capture particles even if they are smaller than the holes in the fabric.
3/ Third, masks would work even if the virus passed fully through.

They reduce the distance at which it travels after, say, a sneeze. It's as if they introduced additional distance between people.

Moreover, they work both on the way out and on the way in.
Read 7 tweets
1 Nov
More than 3 millions of Italians violated home quarantine (of those whose close contact tested positive or were waiting for test results),

according to the Italian Scientific Committee advising the government on COVID response.

(source: open.online/2020/11/01/tra…)
Part of the problem is the slow testing.

6 million Italians should be in home quarantine, only a fraction of them because they tested positive. Most of them are there because tests are few and slow.
Another part of the problem is that enforcing the quarantine & working through the tests backlog is cheaper, easier, and more effective when there are few cases than when there are a lot,

But one would also need the will, capability, and wisdom to do so when it seems less urgent
Read 6 tweets
27 Oct
A REVIEW OF BAD ARGUMENTS AGAINST FACE MASKS

1/ They don't fully protect you from virus in the air.

Yes, but their point is to prevent as much virus as possible from getting in the air in the first place.

(thread)
2/ There is this Randomized Control Trial I see posted on Twitter which allegedly shows that face masks do not work.

That trial shows that face masks do not protect doctors from patients *that do not wear face masks*.

Such studies do not demonstrate that face masks don't work.
3/ Paradoxically, the more you believe that someone wearing face masks is only partially protected, the more you should want everyone to wear them, so that there is less virus in the air.

That's how real herd immunity works.
Read 19 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!

Follow Us on Twitter!