9508
showing only twitter_favs [see all]
Twitter
Here's a message I sent to my students at with some emergency AQ tips. Take care of yourselves, Bay Area…
from twitter_favs
2 days ago
The Remarkable Brain Waves of High Level Meditators
The Remarkable Brain Waves of High Level Meditators [video]
from twitter_favs
2 days ago
Twitter
I was just asked to share my table about all the different initialisation syntaxes and what they do, from the "Init…
from twitter_favs
3 days ago
Introducing Chartify: Easier chart creation in Python for data scientists
Chartify from Spotify: “Easier chart creation in Python for data scientists”
from twitter_favs
5 days ago
Twitter
Accelerate your TensorFlow models with XLA. TensorFlow 1.12 w/XLA is 1.5x faster than TensorFlow 1.11 on resnet50 t…
from twitter_favs
6 days ago
Twitter
We just open-sourced a suite of ODE solvers in PyTorch:

Everything happens on the GPU and i…
from twitter_favs
6 days ago
Twitter
American essayist and scholar Nassim Nicholas Taleb (), whose work focuses on problems of randomness, proba…
from twitter_favs
7 days ago
Twitter
Learning always happens in context. There are so many topics I didn’t understand at first and gave up on. Then I ca…
from twitter_favs
8 days ago
Twitter
If you want some life inspiration, here is Yuri Orlov's resumé:
from twitter_favs
8 days ago
Twitter
How do people organize their reading? Many terrific replies:
from twitter_favs
9 days ago
HTTP/3 | daniel.haxx.se
HTTP-over-QUIC becomes HTTP/3. No more TCP in the future of the web?
from twitter_favs
9 days ago
Twitter
【Scaling Machine Learning at Uber with Michelangelo】对 Uber 这三年来在 machine learning 方面的进展(公司内部组织、流程、技术平台)进行了深度总结。Uber…
from twitter_favs
9 days ago
Twenty things I wish I’d known when I started my PhD
Great tips for beginning grad students - Twenty things I wish I’d known when I started my PhD
from twitter_favs
9 days ago
Twitter
Yup, it just happened! After 2 yrs, I wrote "Model evaluation & algo. sele. in machine learning Part IV" -- since i…
from twitter_favs
10 days ago
Twitter
Human history, in one chart

Almost all the gains in human well-being in history happened since the Industrial Revo…
from twitter_favs
10 days ago
Twitter
Slides for some of my recent talks:

- IHÉS "Gomax" workshop, 2018-10-15:
title: Graph Embedding, content unders…
from twitter_favs
10 days ago
Twitter
I uploaded the slides of my Keynote “The Paradox of Choice - How to thrive in an industry with too many po…
from twitter_favs
10 days ago
Twitter
what does a lagrange multiplier look like? you won't find (m)any illustrations in books. nevertheless, it's a very…
from twitter_favs
10 days ago
Twitter
Dicts now preserve insertion order. So, are they the same as collections.OrderedDict? Nope:

>>> a = { 'x': 1, 'y'…
from twitter_favs
11 days ago
Twitter
Measuring the Effects of Data Parallelism on Neural Network Training: Large empirical study investigating the trade…
from twitter_favs
11 days ago
(500) https://home.sandiego.edu/~shulman/papers/rabbithole.pdf
The weird and wonderful world of constructive mathematics (2017) [pdf]
from twitter_favs
12 days ago
Twitter
tip: If your data is in a dictionary, perhaps from a JSON file, then str.format_map() is your new best fri…
from twitter_favs
12 days ago
The Illustrated TLS 1.3 Connection: Every Byte Explained
Rewritten for TLS 1.3: The New Illustrated TLS, Every Byte Explained
from twitter_favs
13 days ago
Twitter
通过在濒死的病人头部插入灵敏的电极,得以第一次记录下了脑死亡的详细过程。随着心脏停跳血压降低,脑神经细胞会自动进入几分钟的休眠以节省能量,直至电位差过大时会突然爆发出一场猛烈的去极化风暴,释放出的大量钾离子最终会毒死无力自救的大脑…
from twitter_favs
13 days ago
Twitter
A while back listened to a Scott Meyers talk on optimizing C++ code for cache efficiency. Applied some of the techn…
from twitter_favs
14 days ago
Twitter
About 5 years ago, I was incensed by reading ‘s suggestion that one can be ‘“over-informed”, that newspape…
from twitter_favs
14 days ago
Twitter
I've been working through the Pierce "Types and Programming Languages" book. It's not easy reading, but in my high…
from twitter_favs
14 days ago
Twitter
Tesla’s engineers believe that neural networks can write code to automate driving, replacing code written by humans…
from twitter_favs
15 days ago
Twitter
If you have a boring car drive coming up, you could listen to interviewing me on a podcast about refactoring…
from twitter_favs
15 days ago
Twitter
DeepTileBars: Visualizing Term Distribution for Neural Information Retrieval.
from twitter_favs
15 days ago
Twitter
You shouldn’t start with the Spotify model. Spotify didn’t start with the Spotify model.

You shouldn’t start with…
from twitter_favs
15 days ago
Twitter
The Open Images Dataset V4: Unified image classification, object detection, and visual relationship detection at sc…
from twitter_favs
15 days ago
Persistence vs. consistent | Seth's Blog
"Persistence is sort of annoying. Consistency, on the other hand, is the happy twin brother of persistence."
from twitter_favs
16 days ago
Twitter
Recently I’ve been playing with production RL deployments and FB’s Horizon hits the nail on the head with its appro…
from twitter_favs
17 days ago
Twitter
“在糖果和可卡因之间,它更接近于可卡因,”安德森认为打造出这些产品的技术专家和观察科技革命的作者都很天真。
“我们以为能控制它,”安德森说。“但它已经超越了我们的控制能力。它直接被传送到了正在发育的大脑的愉悦中枢。这超越了我们作为…
from twitter_favs
20 days ago
Untitled (https://www.youtube.com/watch?v=-s8kNtc5lX0)
Pacific++ 2018: Kate Gregory "Simplicity: not just for beginners"


from twitter_favs
21 days ago
Twitter
In "Failing Loudly", we tackle the problem of detecting distribution shift in realistic high-dim datasets. In low-s…
from twitter_favs
22 days ago
Untitled (https://www.youtube.com/watch?feature=player_embedded&v=yy8jQgmhbAU)
My CppCon 2018 talk – “OOP Is Dead, Long Live Data-oriented Design”
from twitter_favs
25 days ago
Twitter
In addition, if you have the right mental model for something, it is generally easy to work out the algorithmic det…
from twitter_favs
25 days ago
Twitter
My go-to example of this is PCA. If you know how to diagonalize a 5x5 matrix by hand, then you "know the math" behi…
from twitter_favs
25 days ago
Twitter
Many people in engineering believe that to understand something, it is necessary and sufficient to have a low-level…
from twitter_favs
25 days ago
Twitter
The same is true of backprop in deep learning -- knowing how to code up backprop by hand gives you no useful knowle…
from twitter_favs
25 days ago
Twitter
This is almost universally true: to understand something, you need the *right* mental models, that capture what *ac…
from twitter_favs
25 days ago
Twitter
Neighbourhood Consensus Networks. More deep learning for image-to-image dense correspondence.…
from twitter_favs
26 days ago
Twitter
Nice initiative to increase reporting of null results and decrease the publication bias from only reporting positiv…
from twitter_favs
27 days ago
Twitter
"Simplicity is the unavoidable price which we must pay for reliability." - Tony Hoare
from twitter_favs
27 days ago
Twitter
【第1274期】:三个人的工程团队。我们是如何写一本教科书的。The Law of Leaky Abstractions。Fall In Love With The Problem, Not The Solution。Stack…
from twitter_favs
27 days ago
Twitter
You repeatedly ask yourself what is the significance of what you are doing. I.e. how is this going to cha…
from twitter_favs
4 weeks ago
Twitter
At Google, we've been getting a better understanding of issues of bias & fairness in machine learning models as we'…
from twitter_favs
4 weeks ago
Twitter
当保健品卖的可发酵的可溶性纤维(美其名曰“益生元”,例如菊糖之类的果聚糖)可导致肝癌。不溶性纤维(例如纤维素)不会。托莱多大学研究人员本来是要研究菊糖(国内也叫菊粉)能减肥,他们让小鼠吃富含菊糖的食物,6个月后,40%小鼠的确减肥…
from twitter_favs
4 weeks ago
Twitter
Approximate Fisher Information Matrix to Characterise the Training of Deep Neural Networks. Want to automagically d…
from twitter_favs
4 weeks ago
Twitter
The discriminator often knows something about the data distribution that the generator didn't manage to capture. By…
from twitter_favs
5 weeks ago
Twitter
I've spent most of 2018 training models that could barely fit 1-4 samples/GPU.
But SGD usually needs more than few…
from twitter_favs
5 weeks ago
Twitter
ML results are definitely not hardware agnostic. You should expect to need to retune hyperparameters at…
from twitter_favs
5 weeks ago
Twitter
Scaling up ML yields interesting/terrifying regression testing / reproducibility challenges. According to Yangqing…
from twitter_favs
5 weeks ago
CppCon - YouTube
These are really high quality talks. Recommended:
from twitter_favs
5 weeks ago
The 1-hour workday | Science
and writing:

And no, I don't "walk the walk" on either of these, though I aspire to...

2/2
from twitter_favs
5 weeks ago
« earlier      
academic ad airline algorithm amazon api app apple architecture automation aws backup baidu bash benchmark bigdata biography bitcoin book business c c++ c++11 car career chrome clojure cmake cmd code compiler conference config course cpp crawler creativity credit-card cv d3.js data data-science data-structures database debug deep-learning deliberate-practice design distributed dns docker driving dropbox earphones economics editor education emacs email encode encryption engineering facebook finance focus font functional-programming game gan geometry gfw git github golang google gpu graph gtd guide habits hackernews hadoop hardware haskell hbase health ide image_processing intellij interview invest ios iptables ipython iterm java javascript jeffdean jetbrains julia jvm kaggle kent_beck keyboard knowledge latex learning leiningen library linux lisp logging lua mac machine-learning management math mathematica matplotlib matrix medicine memory method microsoft mosh mwc neovim netflix network neural-network nlp node note notebook numpy nvidia omnifocus openmp optimization org-mode osx papers parallel performance phd physics planning plugin privacy probability procrastination productivity profiling programming projects proxy psychology python pytorch quadrotor quora r reading recommender reference reinforcement-learning reproducibility research resources rnn rsync ruby rumor rust sbt scala scalability science security server service setup shell shopping shortcuts sleep software spark ssh stackoverflow startup statistics studyhacks sublime-text surge sync sysadmin system tdd tensorflow terminal test thinking time tmux tools trading travel tutorial twitter ubuntu unix unread v2ex vagrant version_control vim virtualization vision visualization vpn web wiki windows workflow writing zhihu zotero zsh

Copy this bookmark:



description:


tags: