The F1 score favors classifiers that have similar precision and recall. This is not always what you want: in some contexts you mostly care about precision, and in other contexts you really care about recall. For example, if you trained a classifier to detect videos that are safe for kids, you would probably prefer a classifier that rejects many good videos (low recall) but keeps only safe ones (high precision), rather than a classifier that has a much higher recall but lets a few really bad videos show up in your product (in such cases, you may even want to add a human pipeline to check the classifier’s video selection). On the other hand, suppose you train a classifier to detect shoplifters in surveillance images: it is probably fine if your classifier only has 30% precision as long as it has 99% recall. Sure, the security guards will get a few false alerts, but almost all shoplifters will get caught. Similarly, medical diagnosis usually requires a high recall to avoid missing anything important. False positives can be ruled out by follow-up medical tests. Unfortunately, you can’t have it both ways: increasing precision reduces recall, and vice versa. This is called the precision/recall trade-off.

When it claims an image represents a 5, it is correct only 83.7% of the time. Moreover, it only detects 65.1% of the 5s.

why accuracy is generally not the preferred performance measure for classifiers, especially when you are dealing with skewed datasets (i.e., when some classes are much more frequent than others). A much better way to evaluate the performance of a classifier is to look at the confusion matrix (CM)

you need to put in place a monitoring system (with or without human raters to evaluate the live model), as well as all the relevant processes to define what to do in case of failures and how to prepare for them. Unfortunately, this can be a lot of work. In fact, it is often much more work than building and training a model.

if you train a RandomForestRegressor and measure the RMSE on the training set, you will find roughly 17,551: that’s much lower, meaning that there’s still quite a lot of overfitting going on

We know there’s an overfitting problem because the training error is low (actually zero) while the validation error is high.

the test set generated using stratified sampling has income category proportions almost identical to those in the full dataset, whereas the test set generated using purely random sampling is skewed.

the data has been scaled and capped at 15 (actually, 15.0001) for higher median incomes, and at 0.5 (actually, 0.4999) for lower median incomes

this is a multiple regression problem, since the system will use multiple features to make a prediction (the district’s population, the median income, etc.). It is also a univariate regression problem, since we are only trying to predict a single value for each district.

I like collecting examples of these

Map out problems using logic trees.

Bad Blood: Secrets and Lies in a Silicon Valley Startup by John Carreyrou

Murata also makes us see how the family members who find her love of the store's rituals strange are themselves trapped within a set of rules - dress this way, don't talk like that, get married and have kids.

fulfillment, nor does she spend too much time contemplating what it might mean to find transcendence in such work.

employee-store romance

misogynist who gets fired from the Smile Mart because he refuses to carry out any of his tasks.

a kind of grim post-capitalist reverie

anti-Bartleby

she is, at least, choosing a different kind of conformity than the rest of society, which insists that she marry and pursue a conventional career path.

beards

Clearly she longs for an authoritarian hand on her shoulder; she wants to know, at all times, what to do.

I don't have a very good answer for this

Our focus becomes fatigued after sustained use and needs a rest period to recover, just like a bodybuilder resting between sets at the gym.

Ask for help in our chat:

Teaching what you know to others is a great way to solidify what you have learned and can often reveal holes in your knowledge that you wouldn’t have identified otherwise.

The strongest assets you can have as a student are a desire to build, a problem-solving mind, and persistence in the face of setbacks.

To navigate through the man pages, you may use ↑ and ↓ to move line-by-line, or try B and Spacebar to skip up and down by a full page. To search for a character or word in the man pages, use / followed by the character or word you are searching for. Sometimes a search will result in multiple hits. If so, you can move between hits using N (for moving forward) and Shift+N (for moving backward). To quit the man pages, press Q.

An individual’s life can continue with an inertia that will lead them on to the next year or decade.

but it is distributed across the U.S. banking sector.

At least one bank which failed last week was not a Problem Bank three weeks ago. Reader, that should not ever happen.

(

Constraints on supply have made houses into scarce assets

log, which computes the natural logarithm

many avian influenza experts told me they don’t think the networks are functioning well enough given the threat level.

It’s like hanging out with a group of basketball players and listening to them talk about basketball. You might enjoy it, and feel as though you’re learning about basketball. But unless you actually spend a lot of time playing, you’re never going to learn to play basketball.

To date, recipients report relatively low levels of tension related to the transfers: 5% report arguments within their communities and 1% report violence or crime.

(a) manipulation of the list of eligible recipients and (b) diversion of transfers sent to eligible recipients

Do transfers create conflict or tension between recipients and non-recipients, or within households?

If you aren’t sure whether a message from or piece of information about GiveDirectly is valid, you can ask us for verification by emailing info@givedirectly.org.

We also find that GiveDirectly tends to make voters happy.

We send transfers using electronic payments services, such as M-Pesa (operated by Safaricom) in Kenya, MTN mobile money in Uganda, and MTN in Rwanda.

Households need at least a SIM card to participate, and we give SIM cards to households that do not already have one

”

one-time large grant programs, we typically send each recipient household roughly US$1,000

$120

this is not a 9-5 job

code-phrase for ‘please attune with me’ in intimate partnerships

To execute a program, run it in the terminal with the python command. For example, in command-line Unix:

How to move around in code is an essential skill

New York

For certain populations, such as the elderly and immunocompromised, the levels of protection may be decreased following both vaccination and infection.