Hacker News Reader

strebleron Airbnb Raises Over $100M as It Touts Strong Growth

> This year, Airbnb expects an operating loss of about $150 million

strebleron How Early Exit Disease Stunts the Growth of Midwest Startup Communities

Perhaps if the investors in these Midwest startups enabled founders to take some funds off the table from a round, it could help to alleviate this problem.

If founders have all the stress of low salary + not much savings + lots of work to do + maybe starting a family, of course a $10M exit starts to look good.

But if they're achieving good results, why not give the founders a small taste of the billion dollar carrot?

strebleron Deep Neural Decision Forests [pdf]

Oh that's right, GoogLeNet did edge out VGG in the pure classification task. Localization is essential to us, so VGG was the clear winner for us :)

strebleron Deep Neural Decision Forests [pdf]

Why do they only compare their method against GoogLeNet? VGG was the winner in 2014, most papers seem to use VGG and my team has observed it to always gets higher accuracy than GoogLeNet.

strebleron Canada's R&D tax credit program hurts R&D in Canada

Agreed, this post is nonsense. The author should consider doing research before forming & posting such opinions.

Executive compensation is not included in SR&ED. Even if you're a CEO who codes, only the time you spend coding can be claimed. The salary claimable is capped at $100K (I believe) for any major shareholders (or family members).

It is also well established that whoever owns the IP gets the SR&ED credits and foreign companies are not eligible for SR&ED.

strebleron Theranos didn’t work with the huge drug company it supposedly made money from

3. Evernote

4. Fab

strebleron China Flexes Tech Muscles Before a State Visit

"...executives from top American tech companies including Apple, Facebook, IBM, Google and Uber, have been invited..."

They may well attend, but at this point the article only indicates that Google/Facebook have been invited.

strebleron Classical music generation with recurrent neural networks

What's wrong with 15MB? A Googlenet model is only ~50MB while being able to recognize objects in real world images with good accuracy.

strebleron Yahoo Acquires Polyvore

http://recode.net/2015/08/03/yahoo-paid-around-200-million-f...

Apparently it was 200M, not 60M. I have to stick to my original hypothesis, Polyvore is more valuable than some realize.

strebleron Yahoo Acquires Polyvore

I didn't see the $60M number, it's certainly cheaper than I'd expect. I agree, I wish I knew the full story!

strebleron Yahoo Acquires Polyvore

My point was related to them being incorrectly labeled as "least known" - site visits is directly relevant to that. As for the impetus behind the acquisition, I'm not one to speculate.

strebleron Yahoo Acquires Polyvore

You may not be their target demographic, but Polyvore is certainly far from that. At 20M uniques per month, 2.4M new sets per month, $22M raised, 150 employees, highest AOV of all social media sites (ahead of Facebook, Pinterest and Twitter), Polyvore is literally the highest converting social platform of them all. They're a pillar of the fashion community.

strebleron How Google Translate squeezes deep learning onto a phone

Considering the very significant accuracy gains deep learning has achieved over previous approaches (and across a number of fields), it's certainly not a simple fad. Having worked in computer vision for a good 8+ years, deep learning is basically amazing.

Deep learning is a form of representation/feature learning.

strebleron What the NYT Got Wrong About Nail Salons

Here is the same ad posted somewhere else:

http://www.meimin.us/bencandy.php?fid=27&id=265577&city_id=3

I wonder if the $10 isn't per hour, rather than per day.

strebleron Google Ideas

It seems to be a collection of things to help fight against evil. I couldn't navigate the site (firefox), so I just looked at the source. What I see is:

-a reverse proxy -service to scrape various business records, helping journalists analyzing records -anti-phishing tool (browser plugin?) that verifies you're not putting your (Google) password into a wrong site.

strebleron Zero Downtime with HAProxy

Apologies for the delay. We do use HAProxy.

To have 2 servers with the same IP, we use VRRP (Virtual Router Redundancy Protocol) and keepalived. HAProxy is setup on 2 separate servers/instances, and using VRRP/keepalive they both share an IP address (which HAProxy binds to). The servers also have their own unique IP address(es) (on top of the shared one), so the shared address doesn't really "belong" to any machine.

If one server goes down, VRRP gives the IP to the other server and that HAProxy takes over.

strebleron Zero Downtime with HAProxy

Exactly, with these "high uptime" requirements, isn't it usual to have 2+ HAProxy instances running on different servers bound to the same IP? At least that's how I do it. When you update the config, first update the backup HAProxy, then update the main one. Any requests that take place during the main proxy restart would go to the backup.

strebleron Ask HN: How do you keep track of everything?

http://www.mytinytodo.net/

strebleron AI-Powered Meeting Scheduler X.ai Raises $9.2M Series A

A treacherous path you tread, seeking 100% automation...hmmmm

strebleron Gmail blocked in China

Have you tried obfsproxy? You can make a tunnel that looks like web/skype/etc traffic, it's worked well for us in China when we've needed it.

strebleron A New, Deep-Learning Take on Image Recognition

My thoughts exactly - not to mention the fact that [1] was based on the earlier Spatial Pyramid work of Grauman & Darrell.

The language in the article is a bit zealous - Microsoft did well, but didn't "win" any challenge in Imagenet. VGG and Google seem to be the ones people are really watching / emulating.

strebleron A Comparison of Image Hashing Libraries

Hmmm...400K rows seems a bit small for this, but you might be able to build a (FLANN) based vocabulary tree that turns phash into a more stable signature (that makes use of database indexes). Then your SQL query would be more like: "where (images.phashsig=inputsig)". Phash might need a tweak, to output 64 floats instead of bits. But it would be more robust to the "random bits that flip" problem.

It's similar to what an engine like TinEye does, but instead of using a bunch of SIFT/SURF/etc features to do numerous visual word lookups, you'd just put in (a modified?) phash and get one word out.

strebleron Burn Baby Burn

I don't think any of them are idiots. But here's a small sample of companies, off the top of my head, who have had serious problems in terms of monetization / revenue and needed to use investors / host company's money to run:

Pinterest. Groupon. Reddit. Snapchat. Instagram. Youtube (for many years). Twitter took a long time to make money. Most news sites have a hard time, which is why WSJ is paywalled (as mentioned in the article). On the "smaller valuation side", we could say FriendFeed, Zite, Hot Potato, Beluga, GroupMe, TweetDeck, Dodgeball.

It's much harder to than you may realize to monetize.

strebleron Early Days on Street View

Ahh yes, the ladybug, we had a few of those back in 2005-6. One of my colleagues strapped one to the top of a car with a GPS and drove around downtown. We wrote a java applet (and eventually had a flash version) to navigate the capture sequences, eventually integrated with Google Maps. Someone also strapped it to a robot and did our school's campus (or at least as far as the battery would go). Good times.

strebleron Q&A With Mt. Gox’s Karpelès: What Went Wrong?

The police investigation is mentioned twice in the interview, i.e. "The police are investigating the case so I won’t be able to say much."

There are various new articles about the Tokyo Metropolitan Police and Mt.gox "working together", but it sounds like you are right that the police are not investigating Karpeles.

strebleron Visualizing Algorithms

I have found this useful for helping visualize what's happening inside a neural net: http://cs.stanford.edu/people/karpathy/convnetjs/

strebleron U.S. Is No. 1, China Is So Yesterday

Exactly what I'm talking about! Their one child policy has already been eased: http://online.wsj.com/news/articles/SB1000142405270230328990...

Either a) the author knew this and didn't think such a major policy shift was worth addressing or b) was ignorant.

This is what I mean - the article was poorly written / researched. China has many problems, the article doesn't have any substance. It's divisive, and inaccurate in several key "facts".

strebleron U.S. Is No. 1, China Is So Yesterday

This article is poorly written and researched, full of grandiose language & hyperbole with little to no justification. For example:

"China’s demography is a disaster ... by 2050, China will have lost one-third of its working-age population"

Grandiose hyperbole substantiated by the fact that in 35 years, people will be older!

"China’s politics are wrong"

"jobs will come home to the U.S...(due to) plentiful cheap energy from fracking"

No need to pander.

strebleron Poll: Where do you host your production environments?

Does anyone know of any hosting services that have GPU instances, other than Amazon?

strebleron Baidu Hires Andrew Ng

Baidu's deep learning based image search is actually quite impressive: http://shitu.baidu.com/