david_ar

I have a similar system, but using pandoc to output markdown

https://github.com/davidar/peruse

See https://bugzilla.mozilla.org/show_bug.cgi?id=630181

My understanding is that the main problem is incompatibilities between the CSS and TeX layout models. That's not to say that there isn't still plenty of room for improvement, however.

http://uds-datalab.github.io/PDBF/ also does this

https://matrix.org and https://vector.im

Note that rocket chat can federate [1] over matrix [2], and I believe mattermost can as well.

[1] https://matrix.org/docs/projects/as/rocket-chat-federation.h... [2] https://matrix.org

Like Liquid Haskell?

http://www.haskellforall.com/2015/12/compile-time-memory-saf...

It wouldn't be the first mental illness possibly caused by an overactive immune system

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2919277/

The problem is not so much about whether predictions are biased. It's a problem of whether the algorithm is reinforcing pre-existing biases. In other words, it's about taking short-term losses in return for long-term gains.

Suppose a certain group of people have experienced some kind of historical discrimination, and as a result people in that group are starting from a disadvantaged position (wealth, etc) compared to the rest of the population. It may well be an unbiased prediction that such people will be a greater credit risk, etc. However, by making decisions solely on the basis of this, although the bank saves some money in the short-term, it reinforces the cycle of discrimination.

If your goal is to improve society in the long-term, making decisions based on these short-term predictors is not a desirable thing to do.

> After 1971 he also took amphetamines, despite the concern of his friends, one of whom (Ron Graham) bet him $500 that he could not stop taking the drug for a month.[18] Erdős won the bet, but complained that during his abstinence, mathematics had been set back by a month: "Before, when I looked at a piece of blank paper my mind was filled with ideas. Now all I see is a blank piece of paper." After he won the bet, he promptly resumed his amphetamine use.

So there's that

I think the reason this is effective is not so much that you've given yourself permission to fail, but rather realising that failure isn't as big a catastrophe as you initially felt it to be.

https://en.wikipedia.org/wiki/Decatastrophizing

That page isn't a complete listing, but you're right, only ~100 of the books seem to be properly open access, unfortunately. I guess Springer can decide to revoke access to the rest at any point in the future.

Edit: looks like they've revoked access already in fact https://news.ycombinator.com/item?id=10811854

From http://www.springeropen.com/books

> SpringerOpen books are published under the Creative Commons Non-Commercial (CC BY-NC) license, so they can be reused and redistributed for non-commercial purposes as long as the original author is attributed.

Yes, (most) linguists seem to be unique in treating this as dogma. It's not the only possible explanation consistent with current evidence.

Edit: by "innate", people mean that it's something you're born with, rather than learned through experience.

> What would a satisfactory "why" even look like exactly?

Before deep learning happened, neural networks used to be popular for regression and interpolation problems. For a long time, our understanding of how they worked wasn't much better than our current understanding of how deep learning works.

In 1994, Radford Neal showed [1] that weight-decay neural networks were in fact an approximation to Bayesian inference on a Gaussian process prior over the space of possible functions. Amongst other benefits, this allowed better approximations to be utilised, and essentially (along with the advent of SVMs) marked the end the first neural network era.

Something like that is what I would consider a satisfactory "why"

[1] http://www.cs.toronto.edu/~radford/pin.abstract.html

IPFS doesn't run natively in the browser yet

https://github.com/ipfs/ipfs.js

That's true, but Bayesian methods were handicapped by a lack of computational resources for quite a long time. Modern Bayesian methods, although fundamentally the same, are in practice quite different today compared to a century ago. Frequentist statistics gained popularity because it was more tractable to apply to complex problems until computational technology caught up.

Edit: one can argue that frequentist methods are generally approximations to Bayesian methods, much like Newton vs Einstein

Also read chapter 37 "Bayesian Inference and Sampling Theory" of http://www.inference.phy.cam.ac.uk/itila/book.html

It would be interesting to see an analysis of the models inferred by this method, to see what kinds of approximations they're learning.

Obligatory pedantic comment: "base-2 decimal" is an oxymoron

https://en.wikipedia.org/wiki/Repeating_decimal#Extension_to...

Replace the word "theory" with "abstraction" or "interface" then. All the article is saying is that there's a bunch of common things (Unix pipes, function composition, procedure sequencing) that all follow a few simple rules (a "contract" if you like). This means you can reason about them together without having to think about the individual "implementation details".

Indeed. http://nixos.org/nix/

You'll soon be able to store arbitrary data structures on IPFS [1], and we're discussing how to perform merging in a decentralised manner [2]. I'm actually quite interested in having a persistent data storage system (like Irmin) backed by IPFS in the future.

[1] https://github.com/ipfs/go-ipld [2] https://github.com/ipfs/notes/issues/40

I'd argue Haskell has the best support for this: http://safehaskell.scs.stanford.edu

TeX itself is quite fast, it's loading all the extra packages that slows things down.

I'm currently working on a project [1] that aims to tie together a number of existing tools to make "TeX on the Web" as painless as possible. It's still in the (very) early stages, so there's still a number of features yet to be implemented [2]. We're also working with the CorTeX project to bring this to existing collections of scientific publications like arXiv

[1] https://davidar.io/TeX.js/ [2] https://github.com/davidar/TeX.js/issues

Is there a particular reason why it's not supported?

http://hyperboot.org/ solves some of these issues

IPFS also allows IA to not just be a backup, but to help distribute the original content itself. There's no longer a distinction between origin hosts and backups.

Yeah, this is a common source of confusion, and the ipfs.pics tagline doesn't really help the situation. Personally I'd change it to "where you can ensure your pictures will be available forever".

You shouldn't read "permanent" as "eternal", but as "not temporary" --- in the same sense as "permanent address" or "permanent employment". HTTP is temporary in the sense that you're relying on a single server to actively continue serving a file for it to remain available. IPFS is permanent in the sense that, so long as the file exists somewhere in the network, it will remain available to everyone.

Edit: If you want to ensure your content remains available, you still need to host it somewhere (your own server, convincing/paying someone else to, etc). IPFS isn't so much about storage as it is about distribution.