cosmic_ape

Regarding 2:

>> Laziness and cowardice are the reasons why so great a proportion of men ... It is so easy to be immature. If I have a book to serve as my understanding, a pastor to serve as my conscience, a physician to determine my diet for me, and so on, I need not exert myself at all. I need not think, if only I can pay: others will readily undertake the irksome work for me.

This is beautiful in its simplicity and still has a piece of truth. But a more elaborate modern take would be probably that we also have finite resources. I do believe for instance that the earth surface is spherical, although I cann't immediately come up with an argument to prove this. I have not personally traveled around the world in a boat or launched satellites.

It seems that there is fundamentally no choice but to trust some people or traditions of thought on some things and then somehow to decide when to start doubting. Which is why fake news is a problem, for instance.

At its heart, yes. But that is also one of the basic tenets of probability theory in general, since the very beginnings. Called the law of large numbers. Monte Carlo usually refers to situations where "trying something a bunch of times" is not a straight forward procedure, because some "somethings" are more interesting than others and there is also a question of how large the "bunch" of times should be to see the true picture.

You are probably right that in principle one could eventually come up with a full catalog of features of a reservation. There would be about, say, 100 of those.

I seriously doubt that they will proceed to define and collect them, since those are probably 10% or less of all reservations, but lets say they would.

Then still, the conversation you make to make the reservation is a process in which you make the decision.

Say, there is a place inside at 20:00 or a place in the garden at 20:30. Are you going to let Google choose between the two options for you?

Do you imagine there would be an api in which you specify to the assistant, before it makes a call, your preferences in that much granularity?

They have listed the problems with such systems in the first paragraph. They claim to overcome these by restricting to very specific domains. But specific domains are usually still wide. A table for two, but in the garden or inside? Inside on which floor? Sometimes it doesn't matter, sometimes it does, and these things are specific and different for every business.

Really skeptical about this. And if this does become a thing, it will dumb down the interaction.

>>Anything graphical or interactive

but ipython is already well integrated with matplotlib in the --pylab mode, and is pretty interactive. Thats how I use it.

re:saving state - its technically a nightmare, I agree. I though it might be possible at OS level - just dumping the whole process. There still would be issues of what to do with open files, network connections, etc, but somehow it seems that in many use cases that would be enough.

IPython's interactive shell is way more convenient than using the default python command line when trying things out. But I just don't get why one would need the whole Jupiter thing on top of that, except for maybe making presentations.

One thing that I think is missing from IPython though is the ability to save a given state of the interpreter, with all the variables in it. So that one could preform a time consuming data loading/parsing once, and restart from that point if some variables get messed up. Jupiter cann't do that either, afaik.

No problem :) re transparent walls I'd say they might have been suitable in late 17th century or so...

But seriously. To begin, the whole construct of comparing theoretical discovery to "seeing new things" is very old. Take Newton's famous "If I have seen further it is by standing on the shoulders of Giants." That actually is late 17th, although wikipedia says it could be even from the 12th, [1]. The windows probably have been used metaphorically even before that, but to be on the safe side, google ngrams finds them before 1900, [2]. Don't see much numbers about actual transparent walls, but a "glass ceiling" was used metaphorically as early as 1848, [3]. Thats remarkable on its own, although thats a different metaphor. Literal uses are even earlier.

Anyway, the point is cliche alert went off right at that first stanza.

There's a deeper problem, though. And that is that "proving" a theorem is not the thing that enables one to see. Proving is typically a rather mundane process, once one knows what one wants to prove. Take Pythagorean theorem even. Once one can formulate it, its not really hard to prove. But coming up with the notion that that is a thing that could be right, thats more difficult, thats the moment of "seeing". Its a bit of a simplification, but its mostly the discovery of new formulations that drives mathematics, not proofs. And so, the opening "I proved a theorem and then whooaaa..." is quite ridiculous.

[1] https://en.wikipedia.org/wiki/Standing_on_the_shoulders_of_g...

[2] https://books.google.com/ngrams/graph?content=window+on+the+...

[3] https://books.google.com/ngrams/graph?content=glass+ceiling&...

Just saw your comment, what you're saying is interesting to know, thanks. I thought it was clear that I was expressing my own opinion. Let me tell you just this: I do not think that Rita Dove, Pulitzer notwithstanding, have ever, literally, proved any new theorem. What I think happens here, is a typical journalist, or professional writer, trying to write about science, or just to use science language, whichever the intention was. Also, windows and transparent walls as artistic device in a 21st century poem? C'mon...

I don't think the lines you refer to would be considered a good poem. It's a poor imitation.

Cann't right now think of anything I know by a proper poet in English, but consider for example Leonard Cohen's

The Maestro says it's Mozart

But it sounds like bubble gum

When you're waiting

For the miracle, for the miracle to come

How much did he tell you in just those 4 lines? More importantly, have ever noticed the particular phenomenon described here? Now you have it in words, and you'll be able to notice it next time. I don't mean you specifically here, but just to convey the point.

That's poetry to me, not so much about "emotion", just noticing and expressing somewhat subtle facets of life. Because at the time this was written this was not a particularly new or subtle insight, that wasn't "true" poetry. Still a good song though.

ok, we can define it like that. but its useful to distinguish between these different ways supply and demand can act then, under your definition. One is concerned with just the number of available jobs, and concerns the "efficient market" situation where employers simply can not increase wages and stay profitable. This is the more common "supply and demand" notion, afaik. The other is artificial - workers can unionize, and what the employers can do is described in the article.

>>Except power itself is determined by supply and demand.

not necessarily. consider what would have happened if the workers were unionized, for example. You might still have had the same number of jobs, but a different distribution of profits to employees/owners.

The term research became extremely diluted now. It attempts to put very different things under one roof, and so is not very useful.

Near the end of that blog post you find:

>>Reading the book, I felt that most of my research is not science, barely engineering and absolutely not mathematics. But I still think I do valuable and interesting research, so I set out to explain what I am doing.

Of course, both the blog author and the book author are right in a way. But probably the useful thing to do here would be to come up with new names for these different kind of activities, rather than destroy the language by trying to just place one more thing into "research in general" category.

The study to which the blog links is about a survey, and it finds that among those who chose to respond to the survey, about 40% have moderate-to-severe depression, and 40% have moderate-to-severe anxiety, according to a certain clinically validated scale. Also, among those who reported their work-life balance as "healthy", the percentages where about 23%, while for those reporting "unhealthy", 55%.

I take issue with the notion of "work-life balance". Metaphors do matter. This one, contraposes "work" and "life". They are now mutually exclusive, and "life" has a well known meaning, and living your life is what you certainly want to do.

And so, if you take the metaphor for granted, then when you're in the lab, or next to a blackboard, during your best years, during the most sunny hours of the day, you are not living your life. It is very easy to feel depressed in these circumstances. There is this famous Russian fable, about a girl who invented a monster, and now she is afraid of it.

My point is, a platitude when spelled out like that, that this kind of work is a part of life, not a mere staircase to some dubious place where you have "made it". I'd conjecture that a large fraction of these depressions are inner conflicts resulting from buying into the narrative that there must be some end-game to this "work", while the only thing that does not require an end-game is "life", which allegedly consists of cuddling with your loved ones, raising children, watching TV and climbing the Kilimanjaro.

Nothing wrong with caring about those, they are essential, as well as financial stability. But if you're in research due to your curiosity, its part of your life. Its an also an opportunity. And, although society wouldn't generally perceive it as sexy as Kilimanjaro, its still way better.

The only caveat is that, as with everything in your life, it requires taking responsibility. If the attitude is "I've been in the system for all these years, did all they asked me to do, and what have I got for it?", that's a problem.

Besides the data, its interesting how the actual targeting was performed.

Does facebook provide an option to show a particular given ad to a particular given user? Or is it possible to select a group of people with a given set of likes? How fine-grained is facebook's audience selection mechanism for ads?

Or was the targeting performed by creating fake groups, befriending people?

It would help if the datasets were categorized by data type. Timeseries, multilabel, etc...

Can someone list the advantages of using this instead of a regular bank+credit card?

1) Monitoring spending in real time or on a weekly basis is useful, and this seems to make it easier. But just a tool for downloading, viewing and analyzing the data from a regular credit card site would suffice, and can be self made.

2) It creates competition for the regular banks, so they potentially would stop being hogs about some of the fees.

But what else?

I understand what you mean. But the personality properties like agreeableness, extroversion, openness, should reveal themselves through behavior over time. Otherwise, what's the point in them? And, say, hi-tech work seems to provide enough opportunities, say over a year, to reveal all of them. This is what I meant by "contact".

Besides, have you not heard the "we do not spend enough time together" idea from a spouse? Not to glorify it, nothing good about it, but it seems to be the reality for more people than it should be.

Have a feeling that if one used something deeper than just lasso regression, much better results could have been obtained. This is very similar to a recommender system, or a text classification problem, and there much better than lasso can be done. The data is sort of open, in principle, btw.

But nevermind what can be predicted from the likes. They say that coworkers have r=0.27, while spouse has r=0.58. This looks like a serious problem with the test more than anything else.

This may be interpreted as "cowokers don't know you (and don't care...)" kind of stuff. But coworkers often have a comparable amount of contact with a person vs spouse. Alternatively, one could also interpret this as there is no "the one true personality", but different personalities in different environments (at least when we define personality as a result of a test). And so coworkers are not wrong, they just see a different picture.

And there might be other possibilities.

In fact, one should try to predict the coworker's (and spuse's, ect...) scores of a person from person's likes, in addition to trying to predict person's own assessment.

Could you point to some reasons why you'd think that there is anything deep and profound about deep learning? As far as the evidence I know goes, its very shallow as a scientific concept. All there is, is the idea of using convolutions for image feature extraction, which predates "deep learning" by decades. And there is little evidence that the nets do anything more than memorization, which is hardly profound too.

Funny that in the abstract it aims specifically at electrical engineers, since electrical engineers do machine learning for about six decades now, under different names. Since the invention of the Wiener filter, the Kalman filter, etc...

There is also an overview of tools for "complex systems", by the same author: https://arxiv.org/abs/nlin/0307015

Second that. Some of these notebooks are excellent literature reviews, highlighting really neat connections between results.

Yes, but I'm not interested in works where researchers are trying to validate a hypothesis like "stereotype threat discourages girls from pursuing math-intensive STEM coursework and careers". I'm asking about what these adolescents are actually saying. Raw answers, perhaps in two paragraphs, to a simple question of how they think they would choose.

edit: and then of course, such dataset should be anonymized and made public, so anyone could see what we are talking about.

People talk about research involving indicators such as "ability" and social factors. But have anyone asked the boys and girls before college, or before starting a graduate degree why and how they make their choices? It seems to me that doing it on scale, like 1000 participants, in a given fixed country, would clear the picture.

In a way, they are voting with their choices, and all we need is a focus group.

I agree that there are many natural groups that should be given much more attention when teaching group theory.

But the virtue of abstractness to me is in the notion of isomorphism, which is surely critical to how we understand groups. Then you can find same (that is, isomorphic) groups with different representations - like sitting as subgroups in two different symmetric groups.

And to define when two groups are isomorphic is more or less the same as to define what a group is abstractly.

There is an assumption that most work environments are not like factories anymore. I wonder how true is that, even in "modern" fields such as corporate software development or maybe accounting. It seems these are still overwhelmingly often series of short 2-3 hour tasks that depend on immediate input from others - for task definition, or details needed for implementation.

Also, as other comments mention, work is used for socializing. I know many people for whom work today is one of the major forms of social interaction. As a society, we just do not sit on benches in front of the houses, talking to neighbors as they pass by, anymore. That maybe for better of for worse, but it would be good to construct some alternatives, before we remove the synchronous workplace.

Can you give examples where the paper is wrong? In large things, not nitpicking. I'm just genuinely curious.

there is a link to a white paper on the github. they show there is a x10 or more decrease in train time on a couple of datasets.

>>Actually, I still have no idea why there is a Markov chain. What is the structure of the chain? Where does it come from and why can't we just sample the parameter without using a chain?

If the parameter can be sampled from the posterior directly, then no chain is usually needed, just sample it. However, most often you can not directly sample from the posterior. Note that this is not necessarily because "posterior is hard to analytically compute". Most times you know the analytical expression for the posterior distribution (up to the normalization constant) but this does not mean you can generate samples from it easily.

So, instead, you construct a Markov chain, from which you can sample easily, and which has the property that in the limit its distribution coincides with your posterior. Then you sample the chain for a long time an hope for the best.

Any chain for which you can show that its stationary distribution is the posterior works in principle. But some chains are better than others. There are also some generic ways to construct such chains, most famous are the Gibbs and the Metropolis-Hastings samplers.

Finally, do not sample blogs by "wikipedia researchers". Sample books like "Bayesian Data Analysis" by Gelman et al. They will not lie to you.

I know that in some journals it is good some of the time. But yeah, didn't mean to imply thats always good. Maybe we should introduce a process of reviewing the reviews, and some incentives accordingly. Not sure how to do that.