Coursening is not as elegant as differential privacy. But using coursening is not a "privacy emergency", it is a very-slightly-less-accurate census. And no one knows what the actual economic impacts of this level of accuracy difference are.
I would love to see the more elegant techniques used, and also have the intuition that this is bad policy, but I don't see the "emergency" here. Labeling it as such just histrionic to me.
Well, it's not just that they'll publish the same data with minimal protection. Which IMO is pretty disastrous. The likely consequence is that the Census Bureau just won't publish data, because confidentiality can't be sufficiently protected w/o noise infusion. That's a lot of data products (and detailed info in those products) that Census Bureau just won't be able publish anymore. For example, I can't really see race and ethnicity data being published at granular geographies anymore--and that's required for redistricting and whatever is left of the voting rights act.
Oh god yes, I've been trying to make a LLM Assisted Magic the Gathering card scanner... its been a hell of a time trying to get it to just OCR card names well....
Yep, Its pretty damn good compared to classic OCR and even more lightweight ones as well that I can run locally. the cards just vary too much over time.
> Edit: Changed this from email because email validation is a can of worms as an example
Email honestly seems much more straightforward than dates... Sweden had a Feb 30 in 1712, and there's all sorts of date ranges that never existed in most countries (e.g. the American colonies skipped September 3-13 in 1752).
It’s a ISO-standard to use Gregorian dates even for dates predating its invention. If you need to support anything else (I never had to in my Eurocentric work so far), you’ll need to model calendars, similar to how temporal did for JavaScript: https://tc39.es/proposal-temporal/docs/calendars.html
Why not have a conversation instead of downvoting. What did I say is wrong?
Your second paragraph is implying that the half of Americans who voted for Trump are "bad Americans". That seems to be sowing the division that your first paragraph warns against (even if it is a reason to dislike Trump).
I don't think either democrats or republicans can claim the moral high ground about sowing division.
It seems to me as though you're reading a lot in to that second paragraph. Are you disputing the basic facts outlined, about "masked agents roaming the streets kidnapping people in broad daylight"? Because that is, in fact, a thing that is happening in cities all over the country right now, and simply pointing out that it is happening is not a partisan act.
This is good, in means we have checks and balances. I don't have the details of this particular case, but does it mean every action ICE takes is illegal?
If someone breaks the law ICE or otherwise, there should be enforcement and justice.
I posted a substantial reply to this comment but immediately deleted it. It's impossible honestly to take issue here without crossing into culture war territory.
First of all, half of Americans didn't vote for Trump, at best a fourth. Look up voter turnout and of those that voted. And yes, not voting is legitimate when you believe both parties don't represent you. This idea that half of Americans voted for Trump makes no sense.
Not only that but most people don't approve of his immigration policy.
But... I think repeated letters are just easier to type than any other string, and since X looks like the classic "marks the spot" logo, it's what people jump to.
I always thought it was from Java, but that's probably a personal bias; I am sure it was used long before Java was a thing. I did find this though (archived from 1999):
> Some hackers liken ‘XXX’ to the notional heavy-porn movie rating.
This seems plausible given the older culture ("this is metaphorically dirty, and therefore like porn", insert puerile snickering) and I can recall old jokes about "searching for" these markings. But I think it's also just about it visually standing out - the X character filling the terminal display cell with sharp lines.
I used to work on what your paper calls "unsupervised transport", that is machine translation between two languages without alignment data. You note that this field has existed since ~2016 and you provide a number of references, but you only dedicate ~4 lines of text to this branch of research. There's no comparison about why your technique is different to this prior work or why the prior algorithms can't be applied to the output of modern LLMs.
I can imagine there is lots of room for improvements around implicit regularization in the algorithms. Specifically, these algorithms were designed with word2vec output in mind (typically 300 dimensional vectors with 200000 observations), but your problem has higher dimensional vectors with fewer observations and so would likely require different hyperparameter tuning. IIRC, there's no explicit regularization in these methods, but hyperparameters like stepsize/stepcount can implicitly add L2 regularization, which you probably need for your application.
---
PS.
I *strongly dislike* your name of vec2vec. You aren't the first/only algorithm for taking vectors as input and getting vectors as output, and you have no right to claim such a general title.
---
PPS.
I believe there is a minor typo with footnote 1. The note is "Our code is available on GitHub." but it is attached to the sentence "In practice, it is unrealistic to expect that such a database be available."
Hey, I appreciate the perspective. We definitely should cite both those papers, and will do so in the next version of our draft. There are a lot of papers in this area, and they're all a few years old now, so you might understand how we missed two of them.
We tested all of the methods in the Python Optimal Transport package (https://pythonot.github.io/) and reported the max in most of our tables. So some of this is covered. A lot of these methods also require a seed dictionary, which we don't have in our case. That said, you're welcome to take any number of these tools and plug them into our codebase; the results would definitely be interesting, although we can expect the adversarial methods still work best, as they do in the problem settings you mention.
As for the name – the paper you recommend is called 'vecmap' which seems equally general, doesn't it? Google shows me there are others who have developed their own 'vec2vec'. There is a lot of repetition in AI these days, so collisions happen.
> We tested all of the methods in the Python Optimal Transport package (https://pythonot.github.io/) and reported the max in most of our tables.
Sorry if I'm being obtuse, but I don't see any mention of the POT package in your paper or of what specific algorithms you used from it to compare against. My best guess is that you used the linear map similar to the example at <https://pythonot.github.io/auto_examples/domain-adaptation/p...>. The methods I mentioned are also linear, but contain a number of additional tricks that result in much better performance than a standard L2 loss, and so I would expect those methods to outperform your OT baseline.
> As for the name – the paper you recommend is called 'vecmap' which seems equally general, doesn't it? Google shows me there are others who have developed their own 'vec2vec'. There is a lot of repetition in AI these days, so collisions happen.
But both of those papers are about generic vector alignment, so the generality of the name makes sense. Your contribution here seems specifically about the LLM use case, and so a name that implies the LLM use case would be preferable.
I do agree though that in general naming is hard and I don't have a better name to suggest. I also agree that there's lots of related papers, and you can't cite/discuss them all reasonably.
And I don't mean to be overly critical... the application to LLMs is definitely cool. I wouldn't have read the paper and written up my critiques if I didn't overall like it :)
Naming things is hard. Noting the two alternative approaches that you referenced are called "vecmap" and "alignment" which "aren't the first/only algorithm for ... and you have no right to claim such a general title" could easily apply there as well.
Except those papers are 8ish years old; they actually were among the first 2-3 algs for this task; and they studied the fully general vector space alignment problem. But I agree that naming things is hard and don't have a better name.
Imagine having more than a passing understanding of philosophy, and then reading much of any major computer science papers. By this "No right to claim" logic, I'd have you all on trial.
The problem solved in this paper is strictly harder than alignment. Alignment works with multiple, unmatched representations of the same inputs (e.g, different embeddings of the same words). The goal is to match them up.
The goal here is harder: given an embedding of an unknown text in one space, generate a vector in another space that's close to the embedding of the same text -- but, unlike in the word alignment problem, the texts are not known in advance.
Neither unsupervised transport, nor optimal alignment can solve this problem. Their input sets must be embeddings of the same texts. The input sets here are embeddings of different texts.
FWIW, this is all explained in the paper, including even the abstract. The comparisons with optimal assignment explicitly note that this is an idealized pseudo-baseline, and in reality OA cannot used for embedding translation (as opposed to matching, alignment, correspondence, etc.)
It seems like you have some misconceptions about Strassen's alg:
1. It is a standard example of the divide and conquer approach to algorithm design, not the dynamic programming approach. (I'm not even sure how you'd squint at it to convert it into a dynamic programming problem.)
2. Strassen's does not require complex valued matrices. Everything can be done in the real numbers.
I think the OP was pointing out that the reason Strasssen's algorithm works is that it somehow uncovered a kind of repeated work that's not evident in a simple divide and conquer approach. It's by the clever definition of the various submatrices that this "overlapping" work can be avoided.
In other words, the power of Strasssens algorithm comes from a strategy that's similar to / reminiscent of dynamic programming.
I would love to see the more elegant techniques used, and also have the intuition that this is bad policy, but I don't see the "emergency" here. Labeling it as such just histrionic to me.