Unless you are strictly talking about predictive modeling, I would disagree with this. PCA just tries to represent N-dimensional observations in a k < N dimensional subspace (for a given k) such that it captures the most variation. This does not mean that any obtained component loadings refer to anything real or meaningful.
To clarify your point, we need to distinguish between pruning dimensions and projecting onto a k < N set of new orthogonal dimensions. The name of PCA does make it sound like you are selecting dimensions.
> The name of PCA does make it sound like you are selecting dimensions.
I certainly thought that!
Using somes terms from the previous comment, does this mean that the k < N subspace is not (necessarily) a subset of the N-space? Or is the subspace a subset of the data with a different coordinate system?
(Yes, I'm still trying to intuitively grasp these ideas.)
In my opinion (MSc math, PhD in applied statistics, currently postdoc at epidemiology department), the fundamentals are books like The Elements of Statistical Learning: https://hastie.su.domains/Papers/ESLII.pdf
I feel like the term data science is completely useless. Machine learning is an approach to "do AI" through statistics. Specifically, it is a branch of statistics where the sole focus is on prediction, compared to e.g. inference.
The topics in that book are good typical "study material" but; 1) we can easily make anyone fail an interview by asking them to derive algorithms or formulas from there, and 2) all these topics predate deep learning, and everything that comes with it.
I agree these are good "filtering" topics, among many others.
I find these definitions funny…. how statisticians vs computer scientists define machine learning differently. You get two different perspectives. Everyone wants to claim AI for themselves. I think the statisticians are pissed off that DNNs works so well.
Anyway least we forget: Neural networks came from cybernetics!
I never claimed that everyone wants to claim AI for themselves.
Artificial intelligence is ofcourse a scientific field on it's own right, even before machine learning was a thing. I'm just saying that AI scientists have used concepts from statistics to create an approach to AI called machine learning. I'm not saying that ML is a subset of statistics, mind you, but the statistical underpinnings of it definitely are. ML is not _just_ statistics too.
Moreover, why would statisticians be pissed about the efficacy of a model?
Firstly, many problems/questions that I work on are not concerned with prediction.
Secondly, even if I did, I would love to use DNNs. It's just that I never have a use for it considering I'm only looking at tabular data. Why bother with DNNs when, say, a random forest will do?
I hear the complaint that R is a terrible language mainly from professional programmers.
I'm sure it is poorly designed from a computer science perspective, but there is no programming language out there which makes it so easy to run any model under the sun as does R.
I think that’s not a feature of the language, but of its libraries. R is popular for similar reasons why it’s so hard to replace python.
In Python’s case, the language isn’t bad, but the implementation is horrible for performance. It sits on the back of lots of fast C/Fortran/whatever code, and polished the (relative) turd a lot, making the end result quite usable, but the “sitting on top of part” also makes it hard to improve the performance of the implementation.
For both R and python the end result is good, but they still are local optima in the space of what is possible.
I've had the opposite experience. My preferred programming style is functional. My current gig has been in Clojure, and I couldn't ask for a better product programming language. I find Python code very difficult to write in a functional way (even the official howto is littered with for loops [1]), but in R I can feel an underlying lispiness, and the pipe operator `%>%` from magittr is really nice: whereas in Clojure there is `->`, `->>` and `as->` for thread-first, thread-last, and named threading respectively, in R with `%>%` I get thread-first as the default and I can override it by placing `.` wherever I want in the downstream function calls.
I feel somewhat embarassed to talk about this, but does anyone else feel a sense of dread and depression when reading news like this? I can't seem to shake it and it's quite strongly affecting my worldview.
It's very jarring to read information from experts and scentific reports on where we're heading and at the same time everyone in the media and all people around me just ignore the problem completely except some comments now and then that reveal how greatly they underestimate the problem. If you start trying to talk about how severe our problems are becoming, you may be labeled a conspiracy theorist / "prepper" and your opinion thrown out the window.
I'm 24, and two years ago I think I had a full nervous breakdown over climate change. I couldn't stop thinking about it to the point that I stopped sleeping, lost 10 lbs, and almost dropped out of college.
If any train of thought leads you to this point, you have a medical problem, regardless of what you are thinking about.
Fortunately, I was able to get help and had fantastic support from my family and friends.
Today I'm off medicine and therapy. I still have bad times lasting 1 - 2 weeks, but never to the point that I can't live.
The important thing that I'm able to see now (even when my anxiety is back) is that everything that is important without climate change, is important with climate change. Even though the world is ending (I am not hopeful), I can still find love, learn new things, see the world, enjoy good food, risk my life, make money, enjoy art, and play sports. You can substitute into that list anything that brings you joy. You can also include fighting climate change in that list.
I used to feel that way. But now I'm old enough that I'm pretty confident that I'll die before shit gets too bad. I'd feel differently if I had kids, but I don't. So mostly I'm left with ironic amusement.
If I were younger, and/or had young family, maybe I'd be doing more. As it is, I spent maybe a decade working on climate change issues at an NGO. And my carbon footprint is relatively low. I don't commute, or travel long distances, I don't eat much meat, and my home is very well insulated. My main sin is running too many computers. But at least I've switched to SSDs.
I know that this is my personal view on life, but:
If you don't care at all about what happens after you're dead, what's the purpose of your life?
You'll be dead and it won't have mattered a bit that you're been here.
If you're strongly moved, you could commit your life to doing something about it. But it's also important to keep in mind that none of this likely matters long term, on the scale of millenia. Future historians will likely consider these times as far more tragic than the fall of Rome. But that's how it goes.
In my case, not exactly nihilism, but est/Landmark. Basically, there is no innate meaning or purpose, only whatever you say that there is. As a biologist, of course, I'd add that there's selection for reproductive fitness. But that's readily ignorable, if you're paying attention.
Maybe from a selfish and individualistic point of view (which is one of the reason things are going in the wrong direction), but if you have empathy and care about others people and life forms it is a very different matter.
I feel strange saying this, but really, it's not going to be that bad. Sure, human civilization may collapse. Billions will die. And we're pretty much locked into an historic extinction event.
But the human race won't get wiped out. And the global ecosystem will recover. It's not like Earth will flip into Venus mode. In a million years or two, this will all be just a blip.
Having fun is the best I've come up with. And yes, working on stuff that I care about. But that's part of what I consider having fun. Long term, I'll be dead, no matter what.
I don’t think it matters as much if what you do in actuality ends up making a significant difference. What matters is that you did what you could. From the perspectives of future societies, you and I will be blips anyways, so that’s irrelevant. But how can one be contempt with knowing that they didn’t try to make a difference, and that their mark on the world disappears when the energy has left their body and their corpse has been burned and buried, I do not understand, because if you don’t (try to) leave anything of value behind, your existence is practically for nothing, except for maybe even a net negative.
You can be pretty confident most people will die before shit gets too bad, also confident that most people will die (latest estimation I heard was over 75% before 2060).
But unless you're over 70-80 you will have to face the shit, it has already been happening for some time now and is accelerating.
If you zoomed us forward a century from today, with present day technology, and assume the worst case scenario for climate models we'd still be perfectly fine. The most impactful practical effects would be the changing of coastlines and more severe weather conditions. It's not the end of the world.
Now enter technology. Think about where we were a century ago technologically. Consider that in many ways technology today is accelerating even faster than it was then. Imagine now where we'll be in a century from now. Nobody would from 1918 would be able to guess what 2018 would look like, from a technological point of view, and I'm certain nobody in 2018 can even imagine what 2118 will look like from the same metric. I mean we are already today approaching the level of technology required to live on other planets which are completely and absolutely inhospitable. In the worst case scenarios of climate change, Earth would still be a utopia by comparison. And then enter in near future ideas like geo-engineering, atmospheric manipulation, and so on.
So I don't think it's smart to run a species level experiment on seeing what happens if we just keeping pumping out CO2, but the worst case scenario is both survivable and improbable. It could even end up changing us vastly for the better. The Black Death is something no one would have ever wanted, yet it paradoxically accelerated, if not sparked, the change that nearly everybody would have wanted of a transition away from feudal society to a more free society.
>If you zoomed us forward a century from today, with present day technology, and assume the worst case scenario for climate models we'd still be perfectly fine.
The average case looks like that. The worst case looks like the Permian-Triassic extinction.
Remember, first and foremost, the climate change isn't something that will happen, it is happening. It's not like one day everything is just dandy and the next day everything is underwater and everybody's freaking out realizing how serious an issue it was all along. It is an extremely gradual process. The characteristics, locations, and composition of food production may change but global level starvation is not a realistic concern. One of the more cynical papers of recent time proposed up to 500k deaths related to food production by 2050. Work that out in terms of net effect on the global population over 30 years. It's certainly not something that should be shrugged off, but at the same time e.g. rising obesity rates will cause magnitudes more deaths before then.
Over time, if our models hold up, we'll see the desirability of coastal areas begin to gradually decline. As sea levels gradually rise we'll probably respond with pumping systems and other technological solutions and if/when technological solutions start to lose ground we'll begin to see demand for coastal property start to decline but as there will always be people with different appetites for risk and prospecting, it'll constantly be a gradual process. We'll probably also see the rise of entrepreneurs taking advantage of the changing landscapes to offer new solutions. For instance check out the ideas for ideas being carried out by Dutch Docklands. [1] Some phenomenal stuff there. For others, we'll likely see emigrations more inland as I imagine we've already seen, at least to some degree, from places like Louisiana.
In the worst case, there may be lots of change. But change is not necessarily bad in the longrun, even if undesirable in the shortrun.
Nope. Just a strong motivation to get ready.
From where I stand in terms of worldview for my children, the 21st century looks like a race between geo-engineering and forced adaptations, inclusive of genetics, with the lowlands' populations having one chance at armed migration.
So maybe not as bad as the 20th...
The things that concern me about this are:
1. The amplifying impact that the global financial system will have on climate change. When banks feel agricultural loans are too risky, they will pull farmer credit lines, resulting in a halving of yields. This might already be happening in some places.
2. The organizations which have a mandate for resolving this problem are not fit-for-purpose. Big bloated and corrupt beaurocracies - breaking up in disarray: see the Green Climate Fund.
3. Active attempts to frustrate solutions. And I simply don’t understand why people would want to do so. Even if your livelihood is dependent on a petrol company, or similar, your families’ ultimate livelihood will be dependent on a stable climate.
For reference, I don’t think life on earth is threatened (that’s just hyperbole) but there is a high probability that the population will readjust down to 3-4 billion and this readjustment will be a rather painful down-sizing.
Here's my view on this: I've stopped caring. I don't have kids and I don't particularly care about the future of humanity. Humans will most likely ruin the planet and, at some point, due to a lack of foresight, kill themselves to extinction. After a few hundred million years, the Earth will be back in operation.
In my opinion, humans just don't do well with restriction. If politicians focus on restricting consumption of goods and energy, we'll just end up in another "general malaise"[1] like in the 1970s.
The only way this will ever get solved is through enormous amounts of clean energy (nuclear, which nobody wants) and recapturing carbon.
> After a few hundred million years, the Earth will be back in operation
Well... The sun is gradually getting hotter over geological timescales. Eventually it will experience Venus-style runaway warming, and this could happen as soon as 500 million years from now.
I feel like someone found an asteroid headed for Earth and everyone just shrugged and decided to pretend it wasn't there. People still make plans for retirement, etc. assuming the world will be the same as it is now, and it dumbfounds me. It's not just climate change deniers - it's people who agree with the science and even see that we're not doing enough, and that it may be accelerating. But apparently it's too abstract to consider it in your own life. I kind of wish it were that way for me; it would beat living in dread and fear.
I've read about this psychological theory somewhere - that there is so much danger around human that human brain has to ignore everything that doesn't impact him directly, othervise everyone would be paralized by fear all the time. Without this mechanism, one would be unable to leave his apartment - there are hundreds of potentialy dangerous individuals you'll meet every day, moving in things that would kill you instantly if someone just made a small mistake. And if you decide to use this thing, you depend on tens of people you've never met that they did their job correctly, othervise you may be killed even though you haven't made any mistake. Public transportation is even worse in this regard.
It would also explain how anyone is able to do things that are unhealthy - smoke, drink alcohol, eat sugar. Even though we all know that it's really unhealthy, it doesn't affect us directly, so it's hard to take it as serious threat.
I'm not sure if this theory is true, but it keeps explaining a lot of human behaviour to me.
Personally, I worry about a lot. Global warming, not so much. It will have an impact on coasta cities but that aside, it really isn't that big a deal. We have much bigger risks to worry about (both environmental and otherwise).
Once the permafrost starts to melt, it will cause a positive feedback loop causing all trapped methane to rapidly be released, doubling the amount of greenhouse gases in our atmosphere. It's literally a ticking bomb. I worry about it.
Nice nitpick. I am of course referring to the fact that the ongoing melting of the permafrost will soon reach a point where methane will start to be released on a terrifying scale.
then again, this is not a process that can be stopped once started and there more than enough methane to have dramatic consequences.
then again this is one reinforcing feedback loop among several others.
Are you familiar with the concept of planet albedo, and how it is changing on earth due to melting polar cap?
Ever heard of ocean's anoxic events and their consequences ?
The reason I'm citing this is not because I think we shouldn't worry, but because if the methane is released (and this depends on the release speed as well) we might have a period with an atmosphere with high concentrations of methane, then a sudden reduction.
Runaway greenhouse gas concentrations and higher temperature might trigger (the reason being we know some things, but we don't know everything) a runaway CO2 capture process. It sounds SciFi, and it probably is, and it seems Nitrogen is a limiting factor. But I wouldn't say it is impossible. Or it could be possible with human help.
Then at the end of this period we could end up with less CO2 in the atmosphere, with oil running out and the clathrates emptied.
Most of the worlds largest cities are near the coasts. And weather systems hundreds of km from coastal areas are still affected by things like ocean currents that affect the temperature. E.g. weather in large parts of North Western Europe is substantially warmer than you'd expect because of this.
So on one hand expect large scale migrations and/or massively expensive programs to mitigate the problems. On the other hand expect far more extreme weather far inland.
The idea it will only affect coastal cities is a dangerous one.
>Most of the worlds largest cities are near the coasts.
There are some cities (New Orleans, Venice, Boston to an extent, to name a few), which are highly vulnerable to sea level rise, but most coastal cities are at high enough elevation that a sea level rise of a few meters wouldn't be the end of the world - it would require better levees, dikes, etc.
Contrast with such risks as increased capability of biological warfare, the persistent risk of nuclear warfare, or, ecologically, declining natural forestation and vegetation due to heavy use of land for cultivated agriculture.
As for extreme weather, it's not like the Earth naturally has less extreme weather at colder temperatures. Warmer temperatures will cause changes in weather patterns, and some places are getting more extreme weather than before, but there is no reason it will be net worse for the whole world.
Not that simple. It's not just about managing a few meters, but managing a few meters increase in worst case. E.g for London a few meters is a minor problem in normal situations, but a few meters in worst case situations have potentially dramatic impact on costs to upgrade the Thames Barrier, and will have dramatic impact in terms of how large areas downstream of London will get flooded if the Thames Barrier needs to be closed.
A lot of cities face issues with follow on effects like that.
It won't be 'the end of the world' if its just a few meters, but it is still enough that cities like London face multi billion flood defence upgrades, and towns housing hundreds of thousands will face increased flood risks as a consequence of anything done to upgrade the defences.
And you're severely underestimating the direct damage. Many of the highest population density areas in the world are at risk,and many of the areas at worst risk are also among the poorest, like Bangladesh.
As for extreme weather, yes indications are that we're heading for more overall extreme weather, not just changes.
Try not living in the past and regurgitating media nonsense. Global warming is what global media uses to brainwash people away from the problem. Climate change is the real thing and real deal.
Inform yourself and learn that one of the current issue we face is a continental increase of temperature and shift inseason that would make agriculture impossible in about 50 years.
There are reinforcing feedback loops that will push the change past a threshold where cascading effects would make life on earth surface near impossible.
Life on earth is a complex interdependent system and the larger animals are the first to go, guess what's next now that elephants, whales, tigers, and pretty much every animal larger than humans is on the brink of extinction.
It won't make life on earth's surface impossible. At worst everything largish (plus some other stuff) will die off and open up the environment to new life (maybe a new cephalopod world order?). Remember, all of this CO2 was once in the atmosphere during the carboniferous and other periods. Nature doesn't give a damn about us and will just roll with the punches we're throwing and crush us like it has other times when life totally fucked the environment up.
Based on my experience over the past few decades, it's going to just keep getting worse. And there's not much that we can do about that. Except to maybe mitigate the extremes. And even that's iffy, because we don't know what'll happen with the permafrost.
