madrafi

You can find aggregate liquidation data for Bitcoin futures and perpetual swaps in bybt.com

2013 called it wants its comments back..

well because quant trading isn't about import xgboost, you need a sustainable infra to handle api failovers, bad data... not even going to mention risk management which is 50% of what quant trading is about. the data provided is anonymized but would probably be a mix of laggard measurements (moving averages, rsi...) and maybe some flow data... quant trading isn't really about finding "secret stuff" most profitable strats you can deploy can be based on stat-arb, basis trading or even just delta-neutral funding farming and such

1.I usually try to write about things that both interest me and wish I have found when I started learning about subject X, the process is mostly doing a lot of reading and working out the ideas for example the most recent thing I published was a primer series about zkProofs.I did about 7 months of studying (largely sparse about ~1400 hours). I then start by writing multiple drafts, updating as I go, I have terrible writing skills but I try to accomodate as I go. I also tend to use pen and paper to develop any math that I need to explain.

2. I started following the advice of Andrew Trask, blogged about HE when I was studying the topic first, then a few posts about Deep Learning.I wouldn't stick a difficulty level to it,to be fair writing largely depends on the person,I think, for me it was hard because I was in a hurry to post and get feedback and trying to take my time so as not to do stupid things.

3. Motivation isn't something I feel like I can control but when doing the writing, not the research, I try to do it every morning.

4. The thing I struggle most with is mental illness (Inattentive ADHD, Depression, Anxiety) hardest part is time management and overcoming the fear of humiliation, my fix for the first was to make the writing week a routine, first thing after breakfast I spend a few hours writing and editing the latter I just faced it and published (great insofar).

Although I am not proud of my writing it gives me a tangible evaluation of what I learned in the past X weeks/months, like an exam. I also write things I don't publish such as sort essays, diaries or ramblings these I just put in notebooks.

A few more thoughts, I would say that getting your blogging process setup ready before you even start thinking about subjects helps along the way.This saves time later, I jumped between a few platforms but now I publish on write.as because it has markdown + latex without all the hassle.

I'd also recommend not to think of it as a duty but a creative adventure, something you do when you feel like. I tend to only write about niche topics or things I've been learning and related to tech (for example I've never blogged about physics or mathematics even though I keep studying both subjects).

Best of luck in your endeavors !

- Edit (some posts I've written) :

- https://radicalrafi.github.io/posts/more-homomorphic-encrypt...

- https://radicalrafi.github.io/posts/homomorphic-encryption/

- https://radicalrafi.github.io/posts/secret-sharing/

- https://write.as/knowledgeprover/zero-knowledge-proof-system...

I meant to skip over PCPs because it would need a separate posts and make the whole longer, I also avoided explicitly mentioning PCPs or SNARKs constructions from PCPs[1] for the same reason. Yes, concluding about the transparency vs proof-size trade-off is a mistake of my part. Thanks for taking the time ! will push an update on that bit soon.

[1] https://eprint.iacr.org/2012/718.pdf

Author here, I did gloss over PCPs certainly, and agree that SNARKs also use PCPs and I might have made the terrible conclusion about proof sizes. Will push an update soon ! thanks for taking the time.

I recently took time to dive into compilers and programming languages by reading Thorsten Bell's excellent books, and this is what came out !

This was quite a treat !

Last year a group of companies and institutions created a working standard. https://homomorphicencryption.org/standard/

Not really ! Activation functions are usually non-linear. What happens is we turn our numbers into their counterparts in another group (abstract group) and work there (at a high-level).

Would like to point that the work done by the curve25519 team is solid, Henry is also behind the ristretto RFC. The reason Facebook used the BFT algorithm is for pure regulatory purposes (they needed a Blockchain therefore a solid consensus algorithm with failure tolerance). The cryptographic constructions used are quite solid unlike OP claims.

imagine arguing PGP security with tptacek

I use the same exact method.

ADHD isn't just overdiagnosed, it's both https://slatestarcodex.com/2014/09/17/joint-over-and-underdi...

Africa and the Middle East are 10-sigmas.

This touches close to home :')

I thought Visa sponsorship was the issue for third-world countries,I and many of my friends (local code meetup) share the same story ; apply to 150+ companies and not a single response. We are white from North-Africa. I stopped applying abroad because these kind of "diversity" policies scare me because they're mostly ethnic based I believe it should be idea based otherwise you end up with a group that all think alike.

- Startup invents chat widget - Chrome extension to block widget I love the internet

Just follow MIT Open Courseware for curriculum as for how to learn either the lectures or textbooks, lectures are good for initial understanding but then you'll need a textbook to go over the definitions,theorems and very important EXERCICES . you might want to take it easy don't try to put everything in some sort of timeline like (I am going to do all undergraduate in 1 year) that's really stupid as may be advertised on the internet. Your real goal ,as you're taking this endeavor ,is to understand it by making it your own, get a feel for it.I cannot describe it but you'll know . Best of luck in your quests .

Finally someone got that pun lol

- Rationality From AI to Zombies (huge collection of essays about (rationality,intelligence,quantum physics,bayesian probability,philosophy...) that can be read on lesswrong.com)

-Intuition Pumps and other Thinking Tools

-Sapiens

-Edward.O Wilson Letters to a young scientist

-Cédric Villani Birth of a Theorem

-Emanuel Derman Models Behaving Badly

-Letters From A Stoic by Seneca

-Mathematics it's contents methods 3 Volumes (Aleksandrov et al.)

-Nick Bostrom Superinteligence

-The Moral Animal by Robert Wright

I used mypy to build a tiny toy deep learning library it was nice to use.

I use Go for systems programming and networking such as a toy P2P app or a blockchain implementation.I also used it for cryptography https://github.com/radicalrafi/gomorph

It depends on the use cases but I'd say hope on the Go train. I only use Python for things such as quick impementations or numerical stuff and machine learning Python is still unbeatable in these two areas .

Learning Go isn't that hard if you know Python it's pretty easy most of your time will be spent awing at the Standard Library which has everything you'll need.

Polished my Golang skills and Python skills. Studied the Deep Learning Book Theory and toy examples with Keras and Torch. Some good ol C by implementing linear solvers. Understood the inner workings of bitcoin and started working on a better version of it. Most of all the month I spent pouring on Homomorphic Encryption literature was very useful. I also went trough some cryptography books (math from Koblitz book and Serious Cryptography). 2018 was less of a tech year for me as I spent more time doing mathematics (love it) and reading various books about (cognitive science, evolutionary theory and math textbooks)

For 2019 since I'll be graduating I think I'll spend more time on mathematics, reinforcement Learning and OCaml which I was postponing this whole year. And of course Rust.

The Tour of C++ and a project or two

- Neromancer - 1984 - The Bitcoin Standard - The Conquest of Happiness (Bertrand Russel) - The Book of Why

There's also the unlikely possibility of discovering a truly homomorphic encryption scheme with no constraints on operations

It was posted on HN 14 Days ago Thanks

Zero Knowledge Proofs are better in this scenario.

Unclear tbh, the only thing I could tell from numerai is that the data is time series. The evaluation of the predictions isn't public afaik so you can't tell.