HN user
Contact me: rjkaplan at cs dot stanford edu
This is the correct answer. A key reason Rebco is so much better than older alternatives is that it can be cooled to superconductivity using only liquid nitrogen (77K), as opposed to liquid helium, which is much harder to work with.
Longtime DVC user here - this is going to be so helpful. We use DVC for all of our model and data versioning, but what's been missing is the ability to cleanly integrate that into our CI workflow. Looks like that's solved now! The cml.yaml syntax also looks quite nice, very easy to follow. Looking forward to trying this out.
Great anecdote from Jeff Dean's twitter [1] about how they thought to try this experiment, given that doctors didn't know such a thing was possible:
"Funny story. We had a new team member joining, and @lhpeng suggested an orientation project for them of "Why don't you just try predicting age and gender from the images?" to get them familiar with our software setup, thinking age might be accurate within a couple of decades,and gender would be no better than chance. They went away and worked on this and came back with results that were much more accurate than expected, leading to more investigation about what else could be predicted."
This is not "AI that builds AI". The actual research behind AutoML is called NASNet (https://arxiv.org/pdf/1707.07012.pdf), and all it is simply: we found two good neural network layers (called NASNet normal cells / reduction cells in the paper) that work well on many different image datasets. It's a very cool research result. But it's not something that will replace AI researchers.
The AI Grant fellowship has been an awesome experience so far; I highly recommend applying if you're on the fence. The grant itself is useful (and this round includes $20k of Google cloud credits!) but equally valuable is the community that Nat has curated around the project. There are a lot of great folks working on AI from all sorts of places just a Slack message away. Now that Daniel Gross (YC's AI-focused partner) is involved as well, I'd expect that community to continue to grow.
This post doesn't even mention the easiest way to use deep learning without a lot of data: download a pretrained model and fine-tune the last few layers on your small dataset. In many domains (like image classification, the task in this blog post) fine-tuning works extremely well, because the pretrained model has learned generic features in the early layers that are useful for many datasets, not just the one trained on.
Even the best skin cancer classifier [1] was pretrained on ImageNet.
"In contrast to deep neural networks which require great effort in hyper-parameter tuning, gcForest is much easier to train."
Hyperparameter tuning is not as much of an issue with deep neural networks anymore. Thanks to BatchNorm and more robust optimization algorithms, most of the time you can simply use Adam with a default learning rate of 0.001 and do pretty well. Dropout is not even necessary with many models that use BatchNorm nowadays, so generally tuning there is not an issue either. Many layers of 3x3 conv with stride 1 is still magical.
Basically: deep NNs can work pretty well with little to no tuning these days. The defaults just work.
This paper is important for a few reasons:
- Labeled data is very expensive. Historically attempts to learn on synthetic data has failed because ConvNets are very good at detecting small visual artifacts in the synthetic data and using those for classification during training. At test time on real data, those artifacts aren't present so model fails. A technique that can beat state-of-the-art (admittedly on a very narrow Eye Gaze dataset, but still) by only training on labels from synthetic data and testing on real data is important.
- They present a useful new idea to improve GAN training: using a history of "fake" images, rather than only the latest fake images from the generator. Ask anyone who has tried to train a GAN: the training is really unstable, each network only cares about beating the latest version of its "opponent". They show good improvements by saving many previous fake outputs to make the generator more robust. This reminds me of Experience Replay from DeepMind for RL.
- It's a published paper from Apple! Great that they are starting to contribute back to the research community.
This paper builds off of DeepMind's previous work on differentiable computation: Neural Turing Machines. That paper generated a lot of enthusiasm when it came out in 2014, but not many researchers use NTMs today.
The feeling among researchers I've spoken to is not that NTMs aren't useful. DeepMind is simply operating on another level. Other researchers don't understand the intuitions behind the architecture well enough to make progress with it. But it seems like DeepMind, and specifically Alex Graves (first author on NTMs and now this), can.
Highly recommend this video on the Deep Visualization Toolbox to anyone interesting in understanding more about how convnets work through visualization:
I'm in the class. He really said this. The CS183C videos are posted online about a week after each class, so you can watch it for yourself soon.
Me too. Use the embed link instead: http://www.slideshare.net/slideshow/embed_code/40175706?rel=...
Absolutely, there are guest codes for everyone who shows up.
What's fantastic about this is that even though the class is at Stanford, everyone from around the world can follow along just as easily. Stanford is opening up more and more of its opportunities to the general public, which is awesome. Another one I'd recommend taking advantage of is Stanford's Hacking Hours:
https://www.facebook.com/groups/750201348380852/
This is a weekly on-campus meetup where people hack on their projects in a collaborative environment. It's open to the general public, and anyone nearby with code to write is welcome to come work on it at Stanford.
Interesting to note that you started this reply with "I think." Would it have carried more weight if you started with "perhaps" instead, or stripped it altogether? Maybe. Either way it's fun to think about.
For the curious: the reason that this property with 24 holds is because 24 = 2^3 * 3. For any prime number p:
p^2 - 1 = (p+1)(p-1)
And p+1 and p-1 must both be multiples of 2 because p is odd. Furthermore, one of p+1 or p-1 is also a multiple of 4 (because they are both multiples of 2 and only 2 apart). So, we can see where the 2^3 factor comes from in the magic number 24. The remaining factor, 3, comes from the fact that p is prime and not a multiple of 3, so either p+1 or p-1 must be a multiple of 3 (otherwise p-1, p, and p+1 would be three consecutive numbers, none of which are divisible by 3, which is impossible).
As a result, for any prime p > 3, (p+1)(p-1) is divisible by 24, so p^2 - 1 is also divisible by 24.
I'm interested. What's the solution?
My aunt used to work at a school that did not have this policy in place. The principle had a debilitating head injury that put her in a coma; she is now unfit to work for the rest of her life. It was tremendously sad for everyone involved, but it has also severely strained the school financially. Under Massachusetts law, the school is required to pay a large percentage of her original salary now that she is disabled. Had they taken out an LTD insurance policy this wouldn't have been a problem.
What is most interesting to me is how the test shows the value we enjoy of living in the information age.
Without instant access to rote information, knowing some of these facts was probably rather important a hundred years ago. Schools today can avoid burdening the student with memorizing so many facts and focus more on developing conceptual understanding, which – coupled with free, rapid access to factual information – is much more valuable.
For anyone trying to determine how they can help: make a phone call. Politicians care about phone calls infinitely more than tweets or Change.org signatures. If enough people call the DoR (617-887-MDOR) and their MA representative with complaints, it's very possible that something could change.
This reminds me very much of pg's recent essay about doing things that don't scale (http://paulgraham.com/ds.html). That essay's very direct title seems to come from the fact that technical founders, like the Stanford students mentioned in the article, can be biased to solving engineering problems before solving business problems. "Do things that don't scale" (as well as Steve's advice here) is a reminder to solve the actual problem, and to solve it in the easiest way, not the most fun, technically challenging way.
Because no new hardware is required from either the vender or the buyer. That's why it has the potential to take hold so quickly, and presumably why so many investors are interested.
Can anyone explain this a quirk line of the source code to me? (In the framebuster portion at the top of the file):
window.location.href = 'ht' + 'tp' + ':' + '//stewd'+ '.io/' + 'pong';
Why has he written 'ht' + 'tp' + ':' etc instead of just the URL as one string? My suspicion is something to do with preventing framebuster-busters from working, but I'm not sure. I've never seen that before.
I found the footnote particularly interesting:
"The Moon has no oxygen atmosphere, so how can something explode? Lunar meteors don't require oxygen or combustion to make themselves visible. They hit the ground with so much kinetic energy that even a pebble can make a crater several feet wide. The flash of light comes not from combustion but rather from the thermal glow of molten rock and hot vapors at the impact site."
I'm fortunate enough to have been included in that 5.7% today. Nonetheless, the statistic is highly misleading and the entire college process is becoming increasingly arbitrary. With so many talented candidates, it is often simply a crapshoot for all but the very best applicants.
Admission to an individual school in particular is extremely difficult to predict. I have friends who were just accepted to Harvard and rejected from Stanford, vice versa, accepted to Yale and neither Harvard nor Stanford, etc. It's similar to an earlier discussion today here on HN about the applicability of group statistics to an individual situation: even if you're an exceedingly qualified applicant, your essays just might not click with a particular school's screener. Or perhaps you're not fully qualified, but someone in the admissions committee really connects with your personal story. These kinds of things happen all of the time in college admissions; with such a competitive pool, getting into any particular school often comes down to chance.
As for the acceptance rate itself, it's as much a measure of how good a school is at marketing as it is of its competitiveness. Harvard (and many others) sends pamphlets by the thousands trying to bait almost-certainly-hopeless students into sending an application, just to drive down their acceptance percentage. Many schools also reject candidates that are "too good," ("Tufts syndrome") because an admissions office will gamble that such candidates will matriculate at a better school. They would prefer to get their acceptance percentage lower than accept a student who probably wouldn't matriculate.
Besides everything that has been already mentioned, there's another glaring reason Google would be unwise to do this:
Antitrust.
Google is already being considered for an antitrust investigation by the DoJ. This makes sense, given their dominance in search and the tremendous network of complementary products in their massive ecosystem. (Note that I am not asserting that it makes sense for the DoJ to take regulatory action against Google -- merely that it is reasonable for them to take a closer look).
With a carrier on their hands, in addition to a device manufacturer (Motorola), a dominant mobile OS, search, AdSense, Fiber, tons of spectrum, YouTube, etc., it is unlikely they will be able to avoid antitrust regulation.
This is fantastic. The UI is wonderful, site design is responsive, and from what I've found so far it has great content. Kudos for you for committing the time to building such a major project and finally shipping.
Initial feedback (major to minor):
- Please add a button to let me add songs in my feed to a queue rather than cutting off the current song.
- There's a bug where, for some reason, the modal popup that I used to sign up / add artists is still capturing mouse events even though it is now invisible. It's taking away my ability to click in the middle of the screen, which is very frustrating cause I can't use search. Not sure why this is happening. It's Z-index is 201 and opacity set to 0 via CSS, if that helps, and I logged in with Facebook.
- My screen is small (11" Air), and some parts of the site are cut off height-wise. For example, the left nav bar's "find friends" button.
- When I click on a new song, the artist in the top bar updates immediately, but the title takes a few seconds.
- All of the letters in the colored boxes next to song names are too far to the right by one pixel.
Again, congratulations on shipping! I'd be happy to provide more feedback if you're interested.
This is a great tool. However, of the two domains I've clicked on to buy so far, both have been purchased already by a new party since their expiration. It would be great if this list were filtered by what's still available.
This could start a trend towards moving away from Kickstarter. So far, the more famous projects that have pursued funding from a similar model without using Kickstarter itself (e.g. Lockitron, App.net) have done so out of necessity – they weren't allowed to use the Kickstarter platform.
But if people keep succeeding without being hosted on Kickstarter itself, that 5% fee might look more and more unattractive to people starting large projects. How much value does being on Kickstarter really add to your project, and how much is simply due to the brilliant fundraising model?
This is an issue for me as well. Ideally, I'd like to be able to command-click five or ten articles, then get through them one by one, rather than having to return to SkimLinks after I read each one.