theCricketer

Piero is now at Stanford :). Piero is definitely worth listening to. He's been in the weeds, knows the details and yet keeps the bigger picture in mind. Also one of the nicest people around.

The videos from the winter 2017 offering are freely available on YouTube.

https://www.youtube.com/playlist?list=PLqdrfNEc5QnuV9RwUAhoJ...

The assignments are at the class webpage too.

I took this class and can vouch for it. They update the class every year to go over recent research - not an easy task in such a fast moving field. For example, this offering covers the Transformer architecture which has recently been used to obtain state of the art results across a wide range of NLP tasks.

Since we're on the topic of tutorials to understand neural nets and modern deep learning, I will throw in Michael Nielsen's excellently written free online "book" on neural nets. It's really a set of 6 long posts that gets you from 0 to understanding all of the fundamentals with almost no prerequisite math needed.

Using clear and easy to understand language, Michael explains neural nets, the backprop algorithm, challenges in training these models, some commonly used modern building blocks and more:

http://neuralnetworksanddeeplearning.com/

This book opened my eyes to the power of textbooks written in such easy to understand, clear style. Bet it took repeated revisions, incorporating feedback from others and hours of work but such writing is a huge value add to the world.

I'm not too knowledgeable on how these deals work, but figured someone on HN would know:

A quick Google search shows that Twilio's market cap is currently $7.4 billion. Does this $2 billion "all-stock" transaction mean that they are giving away over a quarter of the company to pay for this acquisition? Or how else should I read this?

Auto-encoders are overplayed, mostly because they're a pretty easy intro ML project.

I think you mean "normal" autoencoders, like denoising autoencoders or the identity autoencoder that are used for feature learning. Note that variational autoencoders are not really autoencoders in that sense. They are called “autoencoders” only because the final training objective that derives from the probabilistic setup does have an encoder and a decoder, and resembles a traditional autoencoder.

Traditional autoencoders are the common intro projects used for representation learning and to bootstrap other networks, not variational autoencoders.

The insight that made it possible for me to grasp VAEs was digging into the probabilistic setup that leads to this formulation. The neural networks are "just" a powerful function approximators applied on top of this probabilistic framework.

"Without a warning, cause or right of reply the board fired me".. There is more from Tim Kentley's twitter: https://twitter.com/TimKentleyKlay

He just posted a bunch of messages he got from his team at Zoox. Seems like a lot of employees really liked him and his leadership.

Elon Musk has constantly underestimated the difficulty of autonomous driving.

This video (https://youtu.be/wsixsRI-Sz4?t=1h18m28s) shows Elon Musk, two years ago, saying the following:

"I basically consider autonomous driving to be a solved problem".

"A Model S and Model X can drive with greater safety than a person, already. Right now."

"We are less than two years away from complete autonomy".

Might be worth noting that this is old [1]. I think either in 2016 or 2017, the same document was posted in PDF form.

[1] http://martin.zinkevich.org/rules_of_ml/rules_of_ml.pdf

For people that know more about web security than I: Is there a reason it isn't good practice to hash the password client side so that the backends only ever see the hashed password and there is no chance for such mistakes?

I've mentioned this on HN before but I think its still relevant to people interested in learning ML who feel they are behind on the math. If, like me, you can't sit thru lots of pages of mathematics text and instead prefer that a human explains it to you via videos that you can replay, here is a list of courses that take you from basic algebra and pre-calculus math all the way to the concepts you need to understand the principles behind the most advanced ML algorithms. All explained by very energetic people who are experts in their fields, and starting from the very basics.

This covers calculus, linear algebra, probability, statistics, convex optimization and a math for ML course thrown in for the HN audience:

(The first two are "MOOCs" recorded in the 1970s! probably the first ever recorded MOOC, even before the internet, and the lecturer is absolute gold)

Calculus Revisited: Single Variable Calculus | MIT https://ocw.mit.edu/resources/res-18-006-calculus-revisited-....

Calculus Revisited: Multivariable Calculus | MIT https://ocw.mit.edu/resources/res-18-007-calculus-revisited-....

Complex Variables, Differential Equations, and Linear Algebra | MIT https://ocw.mit.edu/resources/res-18-008-calculus-revisited-....

Linear Algebra | MIT - https://www.youtube.com/watch?v=ZK3O402wf1c&list=PLE7DDD9101....

Introduction to Linear Dynamical Systems |Stanford https://see.stanford.edu/Course/EE263

Probability | Harvard https://www.youtube.com/playlist?list=PL2SOU6wwxB0uwwH80KTQ6....

Intermediate Statistics | CMU https://www.youtube.com/playlist?list=PLcW8xNfZoh7eI7KSWneVW....

Convex Optimization I | Stanford https://see.stanford.edu/Course/EE364A

Math Background for ML | CMU https://www.youtube.com/playlist?list=PL7y-1rk2cCsA339crwXMW....

While I am not a fan of Facebook myself for the all reasons recently talked about and the addicting nature of it, yes the self updating address book part is nice. Here's what I did: I got a Chrome extention[1] that unfollows all my friends. I've also unliked every Facebook Page that I used to like. This means my feed is now totally empty. When I login from anywhere there is nothing in my feed. Also I rarely ever get notifications, except when someone adds me as a friend. This way the feed is gone, there's nothing to be addicted to (feed was my main addiction), its very hard to get back because it involves re-following 100s of people which my lazy mind won't do and Facebook doesn't make any money off me (no feed, no ads). So I still get to use it as a address book.

Plus Codes is a "system is based on dividing the geographical surface of the Earth into tiny ‘tiled areas’, attributing a unique code to each of them. "

The original blog post: https://india.googleblog.com/2018/03/searching-for-indian-ad...

For people interested in a bit more technical background, this paper by the authors featured in the NYT article has some good introductory background:

https://research.google.com/pubs/pub45826.html

Note how Uber doesn't even mention plans for an Android app in 2008. Fast forward 10 years, 80% of smartphone shipments are Android phones.

An interview with Maryam from 2008 when she was a Research Fellow at the Clay Mathematical Institute -> this is inspiring: http://www.claymath.org/library/annual_report/ar2008/08Inter...

I don't think they want it to be a different service from Search (like Google Flights is). It's more that when you search for something like "jobs near me" you get these job results, just like the custom restaurant search/flight search from google.com

If you find it easier to keep at it and learn from lecture videos instead of from textbooks, here's a math curriculum of lecture videos I've curated. This covers calculus, linear algebra, probability, statistics, convex optimization and a math for ML course thrown in for the HN audience:

Calculus Revisited: Single Variable Calculus | MIT https://ocw.mit.edu/resources/res-18-006-calculus-revisited-...

Calculus Revisited: Multivariable Calculus | MIT https://ocw.mit.edu/resources/res-18-007-calculus-revisited-...

Complex Variables, Differential Equations, and Linear Algebra | MIT https://ocw.mit.edu/resources/res-18-008-calculus-revisited-...

Linear Algebra | MIT - https://www.youtube.com/watch?v=ZK3O402wf1c&list=PLE7DDD9101...

Introduction to Linear Dynamical Systems |Stanford https://see.stanford.edu/Course/EE263

Probability | Harvard https://www.youtube.com/playlist?list=PL2SOU6wwxB0uwwH80KTQ6...

Intermediate Statistics | CMU https://www.youtube.com/playlist?list=PLcW8xNfZoh7eI7KSWneVW...

Convex Optimization I | Stanford https://see.stanford.edu/Course/EE364A

Math Background for ML | CMU https://www.youtube.com/playlist?list=PL7y-1rk2cCsA339crwXMW...

The recommendations weren't leaked. It was made publicly available by Uber. It is even on their website:

https://newsroom.uber.com/covington-recommendations/

The author might be suggesting that you could have surge priced compensation for the drivers to incentivize them to move to the demand but also amortize that cost for the consumer.

Could you elaborate on Android's weaknesses when it comes to privacy and contrast this with Apple's approach? Asking genuinely to update my understanding.

Question for any Google Cloud folks hanging out here: Is it possible to use Cloud TPUs without using TensorFlow? Is there a more low level library/API to run instructions on TPUs so other frameworks can work with TPUs?

Out of curiousity and so we can all concretely imagine how Sidewalk Labs plans to transform cities, can someone explain what are the speicific changes and associated benefits in this new kind of city? It mentions, autonomous transit, ride sharing, cheaper housing (how?). Can someone paint a fuller picture of what will be new and how it will be beneficial.

I'm not skeptical at all, in fact I'm very hopeful. I just want to be able to picture the new kind of city more concretely.

Related: Here is an interesting piece from Benedict Evans, where he argues that there are currently many reasons voice technology isn't ready to be the next platform.

http://ben-evans.com/benedictevans/2017/2/22/voice-and-the-u...

An excerpt:

"There's a set of contradictions here, I think. Voice UIs look, conceptually, like much more unrestricted and general purpose interfaces than a smartphone, but they're actually narrower and more single-purpose. They look like less friction than pulling out your phone, unlocking it, loading an app and so on, and they are - but only if you've shifted your mental model. They look like the future beyond smartphones, but in their (necessarily) closed, locked-down nature they also look a lot like feature phones or carrier decks. And they're a platform, but one that might get worse the bigger the developer ecosystem"

Sorry I should have made it more clear - the links are at the end of my comment.

MIT recorded a set of Calculus video courses back in 1970s that they have since made publicly available. It is taught by a lecturer named Herbert Gross. His style of lecturing is clear, he states why things are defined the way they are and derives everything from first principles. There is an unusual mix of rigor and focus on building understanding - where everything comes from. It also taught me that math is about reasoning logically and rigorously and we shouldn't always rely on intuition (at least while doing math). Deriving almost all the basic calculus results that were drilled into me from the basic concept of a limit, deltas and epsilons was really refreshing.

Compared to more recent OCW calculus videos, I found this to be better in terms of respecting the learner's intellect, presenting the whole proof rigorously and teaching the student to think a certain way.

Calculus Revisited: Single Variable Calculus | MIT OpenCourseWare - https://ocw.mit.edu/resources/res-18-006-calculus-revisited-...

Complex Variables, Differential Equations, and Linear Algebra - https://ocw.mit.edu/resources/res-18-008-calculus-revisited-...

Calculus Revisited: Multivariable Calculus | MIT OpenCourseWare - https://ocw.mit.edu/resources/res-18-007-calculus-revisited-...

Nice.

I use Dropbox Paper to take math notes, and it works great with math formulas and Latex support. You get sharable Paper documents, downloadable as MS Word docs and Dropbox paper overall feels light, fast and minimalistic.

The only thing I'd improve about Paper's Latex support is to add some autocomplete features that make it faster to type out Latex and a couple of minor bugs, like the incomplete Latex disappearing when you switch tabs.

Nice, great first step.

It would be great to start this experiment and eventually come to an API spec that all airports could implement. There is a lot more incentive to build great apps if they can be used at all (or many) of the hubs.

Chris Leary, a compiler engineer at Google gave a talk about XLA at the recent Tensorflow Dev Summit:

https://www.youtube.com/watch?v=kAOanJczHA0