FAIR, DeepMind, Brain, MSR and many others all employ large numbers of people working on pure research with no immediate business application, and with at least as much freedom as their academic counterparts.
honestly, then y'all are mostly responsible. If you are going to have a younger student do a guest blog post, you should make sure that the project makes sense technically, or at least is not silly. What if the student tried to train a neural net to build a perpetual motion machine and reported their progress on that? Would you have posted it? That project would be about the same level of silliness as what is reported here. I don't blame them at all for doing a silly thing, but I surely blame you for making it your guest blog post.
I think you are missing the parent's point. A CNN is a more complicated model, not a simpler one- it is better to try simple linear classifiers on bags of word or bags of bigrams or trigrams before breaking out the more complicated neural models. Note that you can do this easily with VW or FastText.
the aim of my post is not to make a prediction based on a standard alogrithm.
my goal is to show that DL techniques make better predictions. What I wrote was a first draft of the algorithm, and I've not made my point yet, unfortunately. I'm working now on a better solution.
Again, you're missing the point. If your goal is to show that DL techniques make better predictions, you need to begin by answering the question "better than what?"
The exact phrasing matters too, I guess. There are lots of ways to find a good packing, and even good bounds on how far your good packing is from the global optimizer. So a reasonable question to your interviewer would have been "does it need to be the exact global optimizer, or can it be a good approximation ?". If they wanted a good approximation it might not be a bad question, and the question may have been designed to get you to ask for the distinction? Or maybe they were just dicks, who knows.
Note that the fact that this can be easily accomplished in VW doesn't really take away from the message the authors are trying to make; namely, simple models done carefully are nearly as (or more) effective in these sorts of problems as fancy deep models, but much cheaper to train and test.
This is implied through research on reductions in machine learning.
That simple models can solve complex tasks.
For example, you can do multiclass classification, cost sensitive (importance weighted) multiclass and binary, quantile regression, structured prediction (as is done with HMMs, CRFs, MEMMs, structured SVMs etc.), just using a binary classifier.
So, if your implementation of that binary classifier is efficient and performant, you'll be (given that your reduction is consistent) efficient and performant on any of the above tasks.
What the authors of the paper above did, is that they rediscovered some old tricks, removed the theory of reductions, and that's that - without referencing vowpal wabbit that does way more useful tricks. I'm not sure why, because VW team consistently references Leon Bottou (out of all others) that is member of FAIR, and has been using implementation tricks for decades.
Their log(k) implementation is probably less performant than the one-against-some consistent reduction in VW due to the latter having better theoretical bounds on performance.
I lived on henry street for a while, and I can't say that I know anything about the back end of the distribution network, but I can say this article is missing a lot of things about the front end. The produce is cheaper than a standard western store, but the quality is much more varied. Moreover, the market (in the economic sense) seems really efficient. If you see something cheap in Manhattan chinatown in every store, it might be a really good deal, and just that thing is in season or a lot of it was just delivered. But if something is cheap in just one store, it is probably in bad shape. Stores are willing to sell produce that simply wouldn't be sold in ordinary western stores; and do things I think wouldn't be allowed elsewhere. For example, I have seen one place where if they have baskets of strawberries, and some of the strawberries are getting moldy, they will by hand separate out the non moldy ones, toss the moldy ones, and repackage the baskets. Moreover, the stores don't bother to keep clean at all, you can smell them from a block away. If you like food shopping, and you pay attention, its great, but if you want to just get your food, it is a lot of work. I think most americans do not want this kind of tradeoff.
They definitely missed the point about ugly fruit and vegetables. From speaking to some of the vendors of cheaper fruits and veggies at farmer's markets, some of them (but not all) buy the second dibs (what the supermarkets and their other obligations don't buy) as well as more irregular fruit and produce.
Now, of course there is a growing ugly fruit and vegetable movement[1]. It's becoming more mainstream and even Whole Foods is buying in[2]. There is an effort in Boston to provide good affordable ugly fruit and vegetables to underserved communities.
> some of them (but not all) buy the second dibs (what the supermarkets and their other obligations don't buy) as well as more irregular fruit and produce.
On top of that, the wholesalers know that these buyers will never pay a premium price, so they may get better deals on 1st tier items than mainstream supermarkets when the wholesaler really wants to sell.
> For example, I have seen one place where if they have baskets of strawberries, and some of the strawberries are getting moldy, they will by hand separate out the non moldy ones, toss the moldy ones, and repackage the baskets.
The financial crisis, explained with strawberries.
Well, to approximate a synthetic CDO squared, you'd first put all the moldy strawberries in a new basket, label it "Strawberries", send it to the rating agency where it would be rated AAA American strawberries, and then sell firms their choice of the least, middling, or most risky slices of the basket of "AAA American strawberries."
There are lots and lots of strawberries of various different ages in a big pile. Some will be fine for the week and others are hours away from exploding in furry mould.
Someone splits them into the newest and oldest strawberries, and attaches a higher price to the former.
Someone then buys the older ones, and splits them again into the freshest and oldest. And so on, and so on until each group of strawberries is a set of near identical ages.
Finally, a customer comes along looking for strawberries to supply Wimbledon. It's in two days!
They find the basket of the very cheapest strawberries, and when asked how long they'll last they're told "Well, some will go off, but it's not like they'll all go off at the same time! As long as you buy them assuming maybe 5% will have gone off you'll be fine". One day later and 70% of them go mouldy and the rich folks at Wimbledon are shocked at the sudden shortage of strawberries.
> For example, I have seen one place where if they have baskets of strawberries, and some of the strawberries are getting moldy, they will by hand separate out the non moldy ones, toss the moldy ones, and repackage the baskets.
Having worked produce in both regular chain markets, and in "gourmet natural foods" markets, I can assure you that this is a common practice.
Makes sense to me. I get moldy strawberries so often at the local supermarket that I absolutely open the container and rumage around before I buy. If they bothered to remove the moldy ones for me, I'd appreciate it.
What's the alternative? Should they be throwing away the entire container?
At some point in the ripening, handling the berries damages them enough that it's not worth it, but otherwise the practice is fine. At the gourmet market I worked at, pretty much all the berries were taken from the original flats they arrived in and repackaged (removing the bad berries along the way).
Pro tip for delicate berries: when you get them home, remove them from the package they came in and store them in a broad container lined with paper towels, something big enough so they can be spread out with a minimum of stacking. They'll last much longer that way.
Mould is not like dirt that can just be rubbed off the product. If a package of strawberries has some mouldy berries in it, then that mould has already spread to its neighbours without being visible. All they're doing is repackaging the berries that don't have any visible mould on them, and that mould will grow very quickly.
If you buy a supposedly "mould-free" bundle that was repackaged from a mouldy bundle, the entire batch is likely to rot within a day or two of purchase.
They all have mold on them. It's only the ones that have mold colonies large enough to be spotted by eye that are sorted out. Does that correspond to how much mold is dangerous? Who can say?
PS: Side note, either USDA AMS hasn't updated their site in awhile, or some "most recent" frozen berry standards date back to the 50s/60s. Which, I guess makes sense; I imagine the process hasn't changed that much.
Picking the good strawberries and repackaging not only should be allowed, it should be mandatory. How are we gone so far away from (agricultural) reality that such a simple and necessary task is /forbidden/ ?
And this describes grocery shopping in China, not just Chinatown. I look forward to the day I can just go to a western supermarket and not have to worry about getting jipped on veggies...save so much work even if they cost a lot more money.
FYI "jipped" is actually spelled "gypped", which is a slur against Gypsies. I only raise this because I used to say it a lot, and was unaware of its etymology and connotations.
And "Gypsies", describing the Romani people, is sometimes considered derogatory because of its negative connotations, and because it is reflective of the mistaken belief that they were itinerant Egyptians.
Sorry, I had no idea where this word came from (or was spelled!). I should have just said "ripped off". In the states, the word has completely lost its original context (and more so in China).
>"I encounter a lot of people who tell me that they never knew the word 'gypped' had anything to do with gypsies, or that it's offensive — especially when the word is heard not read," says University of Texas at Austin professor Ian Hancock, who was born in Britain to Romani parents. "My response to them is, That's okay. You didn't know but now you do. So stop using it. It may mean nothing to you, but when we hear it, it still hurts."
Perhaps if you live in an area where the word "gypsy" isn't well-known - but it would be in most of Europe at least. And given the history of abuse and oppression (eg: forced sterilization, even before the nazi concentration camps) it's not really a word that should be forgotten - in the sense that forgetting history is a great way to later repeat the same mistakes, and commit similar abuse.
I don't really see how "jipped" doesn't end up pronounced as "gypped" - so the connotation remains. FWIW the Moby Thesaurus lists a long litany of synonyms along the lines of "con man" and "cheat" for the term "gypper".
Slav and slave have different pronunciation, not just different spelling (and actually, 'slav' in the meaning of slavic people was also spelled 'slave' - not sure if that was with a silent 'e' on the end or not). It's also as I understand it a name given to the slavic people by themselves: "Slav \Slav\ (sl[aum]v or sl[a^]v), n.; pl. {Slavs}. [A word originally meaning, intelligible, and used to contrast the people so called with foreigners who spoke languages unintelligible to the Slavs; akin to OSlav. slovo a word, slava fame, Skr. [,c]ru to hear. Cf. {Loud}.] (Ethnol.) One of a race of people occupying a large part of Eastern and Northern Europe, including the Russians, Bulgarians, Roumanians, Servo-Croats, Slovenes, Poles, Czechs, Wends or Sorbs, Slovaks, etc. [Written also {Slave}, and {Sclav}.] [1913 Webster]"
I'm reminded of a quote about what if the German army named their machines of war after the people they tried to exterminate, like the Apache helicopter etc.
I'm not saying we should "prohibit" words, I'm just saying I don't quite see how one could make much of a case for "gypping" being a word that isn't steeped in history of abuse of and discrimination against the Roma people. As recent as the 70s and 80s (and even to this day) for most of Europe. Eg, to quote song "Magnificent Seven" by the Clash: "What Do We Have For Entertainment? / Cops Kickin' Gypsies On The Pavement". (I never interpreted the song any other way than as a critique of systematic abuse by the police, and callousness of mass media - not as a slur by the Clash against "gypsies").
Just because "nigger" is derived from "negro" and means "black" doesn't automatically mean we can choose to pretend black people hasn't been and still are victim of racism, and decide that "nigger is now ok to use".
Because the American speakers who use it don't have a history of discriminating and abusing Roma people? There are barely any Roma people in America to abuse. Certainly in much of Europe Roma people have been treated very badly.
It's certainly true that many portrayals of black people in continental Europe often look very odd or offensive to American eyes.
I'm not sure that makes much sense. By that logic I should be fine using "indian" rather than Native American or First Nation, because it's been a thousand years since Norwegian Vikings first killed Native Americans? (Which makes doubly no sense, as there were a number of Norwegians in the early settlements of the west, and I'm sure there are a few tales of horrors to be told about some of their exploits).
To clarify, I didn't mean that the discrimination/abuse of the Roma happened a long time ago. I meant that most discrimination/abuse of the Roma has happened in Europe, not in the US. Anti-Roma racism in Europe is obvious, ugly, and contemporary. In Europe, Roma people are a coherent social group of millions of people; in the US, without mass immigration, they are a scattered minority so small that the only thing most Americans "know" about them comes from "The Hunchback of Notre Dame". With nothing to refer to, the word has been bleached of reference to Roma people in American use.
Or dig deep into English vocabulary and select a more appropriate synonym without the unwanted association: hustled, conned, swindled, scammed, deceived, ripped off, tricked, cheated, flimflammed, bamboozled, bilked, duped, hornswoggled, rooked, fleeced, burned, chiseled, snowed, fast talked, screwed, shafted, bled, suckered, stiffed, milked, gouged, overcharged... there are probably more. Some of them are shaded according to the intent of the other party and how you felt about it.
But I'm not sure any of these apply to the practice under consideration, which is selling substandard produce to people who haven't properly calibrated their expectations.
The fruit in China, at least in the big cities, is pristine. Apples are individually wrapped, everything is generally impeccable. You can get cheaper beat-up fruit in some food stalls at a cheaper price if you want.
This definitely happens outside of Chinatown as well. In fact there's a place at the border of Los Altos and Cupertino (right next to a Trader Joe's in fact) that deals with produce that is very close to expiration that was clearly sent over from other supermarkets. They also sell green that aren't "baby greens" -- i.e. they're grown a bit too mature and are now bitter and tougher than what the typical grocery store patron would like.
Obviously the nutritional value is the same, so the place makes for a great source for smoothie ingredients and frozen fruits to be consumed at a later date.
It's a very impolite but common way that Americans differentiate the two groups of people that are called Indians when the context isn't clear. Native Americans (feathers) and people from India (dots). Feathers refers to Native headdress and dot refers to a bindi, the decoration commonly worn by Hindu women.
I understand that the dot/feather thing is considered crass, but I will say it's a bit overboard to be offended at it. The idea isn't that all India Indians have dots or whatever, it's just synechdoche, where one aspect of the broader culture stands in for the whole. Like if you said, "He's from America - suit not cowboy boots." (Referring to urban vs rural America) it would hardly be offensive as a stereotype.
"Yankee" or simply "Yank" is a synechdoche that the British use for people from the US.
Tell a 6th generation Georgian whose family farm was destroyed by Sherman's March to the Sea that he's a "Yankee" and see what response you'll get.
And your 'suit not cowboy boots' comparison doesn't even make sense. In the parts of the US where cowboy boots are popular in rural areas, they are also popular in urban areas. Visit Dallas and see that dress cowboy boots go just fine with a suit. Governors and legislators, President Bush and innumerable wedding party members, have all paired cowboy boots and suits.
Lets start with nobody said they were offended, just pointed out the GP was using a racial stereotype. Calling a racial stereotype a racial stereotype when asked for clarification isn't the same as "I'm offended," that is jumping to unfounded conclusions.
Saying "Indian (dots)" doesn't make sense out of context unless you are American or very familiar with American stereotypes. You also need the "feathers" part to be able to compare the two to give it some context for those who are familiar with American stereotypes but haven't been asked "dots or feathers?" before. Saying it in an international form is confusing and unclear thus a bad use of language no matter if its offensive or not. Additionally using racially charged language unnecessarily detracts from the overall message the poster was trying to convey and their post is now hidden behind [flagged].
>Visit Dallas and see that dress cowboy boots go just fine with a suit.
I just went to a wedding a few weeks ago where the entire wedding party (including the bride) were formally dressed and wore cowboy boots and I don't even live in the south or a rural area!
> And your 'suit not cowboy boots' comparison doesn't even make sense. In the parts of the US where cowboy boots are popular in rural areas, they are also popular in urban areas. Visit Dallas and see that dress cowboy boots go just fine with a suit. Governors and legislators, President Bush and innumerable wedding party members, have all paired cowboy boots and suits.
Sounds like it makes perfect sense to me, given that my whole point was that synechdoche is about using a salient feature, not necessarily a universal one. That's what I was trying to say about the dots/feathers thing - cowboy boots are associated with (mostly southern) rural areas - they're even called cowboy boots, even if they are used other places and even if most people in the south or in rural areas don't wear them most of the time. Suits are associated with big metropolitan companies, even though people in rural areas still wear suits.
> cowboy boots are associated with (mostly southern) rural areas
Your expression seems to become less meaningful as you clarify it.
Cowboy boots are part of Western wear. The specific style (as differing from general equestrian boots) started during the cattle drive era of Texas, Oklahoma, and Kansas in the late 1800s.
It spread, certainly. Cowboy boots are a much more salient aspect of Montana, New Mexico, and Wyoming - decidedly non-Southern states - than Virginia, Florida, or North Carolina. What cotton farmer, tobacco farmer, or citrus grower works in cowboy boots? The traditional Florida cowboy wasn't even called a "cowboy", which is the western term, but a cowman or cracker.
I'll say this again, they are indeed called cowboy boots, which means they come from the area where there are cowboys. That's the West, not the South. (The two areas overlap in Texas and Oklahoma.)
So 1) are cowboy boots a salient feature of rural life in the US? No. They are not common to rural life in Michigan, New York, Maine, and other non-plains states.
3) Is the south the area where cowboy boots are the most salient feature? No. That would be the West. In the South you buy cowboy boots at places which sell Western wear. You don't buy cowboy boots at a places which sells "Southern clothing."
On the other hand, in the West, where cowboy boots are a salient feature, you are also likely to find people in the cities wearing dress cowboy boots with a suit.
So "cowboy boots" means someone from the West, and "suit" means someone dressed formally, "suit and cowboy boots" means someone dressed formally in the Western clothing tradition.
While "suit not cowboy boots" tells someone in the Western US to dress in the standard US/European business tradition.
But "cowboy boots not suit" makes little sense as a disjunctive. It certainly doesn't characterize someone from the rural US, or characterize someone from the rural South in the US, with someone from the city.
I'd like to chime in for future readers of this thread, a passive observation. In 2016 Seattle tech, I hear the "dots not feathers" formulation (or rarely the reverse), at least monthly, always in professional settings and nearly always used by a person in describing him or herself. Subjectively it would ring very odd and impolite to hear it from someone not describing his own ethnicity.
Clarification: Americans confuse Indians between those from Bharat/India and Native Americans. Dots and feathers is a common way to differentiate between both. I am from India.
I am from the US. "Dots and feathers" is not a common way in the US. It feels pejorative as well.
If it's ambiguous, I say "Indian (from India)" or "Indian (subcontinent)" vs. "Indian (Native American)". Or I'll just say "from India" or "Native American", and avoid that ambiguity.
Plus, I think many in the US don't know about the bindi, so referring to "dot" in that context may not help as much as you, someone from India, may think.
>I am from the US. "Dots and feathers" is not a common way in the US.
I am from the US, I have heard people say things like "I picked up some Indian artwork this weekend, dots not feathers." I've also been asked "dots or feathers?" when I was talking about something Indian. I've also been asked in gestures, pointing to the forehead to represent a bindi then saying "or" and covering and uncovering the mouth while making a noise to do that mock Indian battle cry (see the scene in Peter Pan if you are unfamiliar with it)(1).
Its not super common (for obvious reasons) but I've heard to several times from several different people so its common enough in at least some parts of the United States. Being casually racist in informal conversation is still a "thing" in the United States.
That being said, its impolite racially charged language thus very inappropriate in most cases and should be avoided especially in formal communication.
(1) I am ashamed to admit that I did that while playing "cowboys and Indians" as a child before I knew better.
A Google Book search finds only a 2002 Weekly World News opinion piece by "Ed Anger" and a 2009 book "Stealing Ganymede: A Novel".
There are ~150 hits in DDG for "Indian" plus any of the four combination of order and choice of {"and", "or", or "not"}. There are 94 hits in Google for "dots or feathers" and 120 "dots and feathers". This includes "polka dots and feathers."
Almost every single one referring to people uses that phrase in a derogatory way.
None of the 6 phrases exist in Google's n-grams.
That makes it decidedly not common, at least in the written word.
> Being casually racist in informal conversation ...
Correct, I never, ever, ever said it was appropriate, it is decidedly not. If you look up a couple posts I highlighted the problems with using it on HN (https://news.ycombinator.com/item?id=11985794). I don't expect that it would be in written word to often either. Point is its common enough that I've heard it several times over many years from several different people. Perhaps you don't talk to too many casually racist people or it is regional.
I still don't think it's common, though I agree it's not rare. I provided counter-evidence that my observations are not solely biased by the people I know nor the regions I've lived in. To flip your question around, perhaps you are in the one region of the US where people do that?
The other part of my response was because I didn't like the wriggle room in your paragraph describing how being 'casually racist in informal conversation is still a "thing" in the United States' and that such uses are 'very inappropriate in most cases and should be avoided especially in formal communication.'
HN is full of informal conversation. Is HN casual enough that it's one of the few places where this "thing" is appropriate? I don't think that's what you meant. I wanted to make a more firm statement which removes that possible interpretation.
I wouldn't claim that the phrase is PC, and it is indeed lazy in written form, but people need to get over being so easily offended on behalf of others.
Yes, it seems that phrase is more common than I originally assumed. That said, movies use derogatory terms, even Oscar winning movies.
People need to get over the word "PC". It's a nearly meaningless term because it comes with the pejorative assumption that the reasoning behind it is always excessive.
Do you have any word for "politically correct but not excessively so, so I think I'll change my language"? Or is every such suggestion inherently based on offense and therefore easily dismissed?
Because my statement above was to show pictures of Puebloans in Santa Fe from now and from 100 years ago, and argue that "feather" makes about as much sense as calling a 6th generation true son of Georgia a Yankee.
Are you going to go to the South and start calling everyone Yankees? Or would that not be politically correct?
You do realize that these things were already being studied in the 90's, right? The Grefenstette paper mentioned in this article makes use of an architecture from Das et al. in 1992. More generally, people have been talking about functional programming and relationships with AI since there was functional programming. However nice Colah's blog post is, he most certainly did not come up with any of these ideas first.
any visualization of these algorithms in 2 dimensions (with cubic feature expansion!) is completely misleading if you intend to work on any real problem with many dimensions. Also, for those asking for execution times, these would be horribly misleading as well.
Just because wired says something doesn't make it true. Whatever other accomplishments he may have, Le was not a contributor to word2vec, and he did not originate the idea of mapping words to vectors.
I think the poster got mixed up. Quoc Le was the first author on "Distributed Representations of Sentences and Documents" aka paragraph2vec, so he has been involved in the x2vec scene. Just not the word2vec. And I would argue that word2vec did not originate the idea either, just popularized it and showed the power of such an approach (plus an awesome demo!).
The original paper on word embeddings from 2001 from Bengio et. al. "A Neural Probabilistic Language Model"[1] is the first I am aware of.
