Rapidics Accelerates my Machine Learning Learning

rapidics-header2For the past week or so, I’ve been digging into the topic of machine learning. It’s something I’ve been interested in for a long time. I’ve done some reading on the subject, and collected links to informational resources and open source tools for years, but I long ago reached the point of diminishing returns, where I either needed to start actually experimenting with it on my own, or otherwise needed a specific reason to learn more.

Thanks to a chance meeting at a coffee shop, I now have the latter. Mark Seligman of Rapidics has helped me understand more about the application of machine mearning techniques. What I find particularly interesting is that Mark and his company are in the business of providing machine learning infrastructure. They are doing some of the heavy lifting to help make machine learning easier and more useful for others to use by taking a generally applicable algorithm called Random Forest and creating a solid, fast multicore implementation called Arborist that can work with the popular, open source, R statistical computing package. By doing so, they’ve achieved major speedups over the standard R implementation, and efficiency and scaling advantages over many of the coarse-grained approaches to speeding up R on parallel hardware.

What’s particularly interesting to me is that they’ve sped things up enough that it could fundamentally change the way people use Random Forest for machine learning, while at the same time making it useful to people who haven’t even heard of machine learning today. That makes the subject triply interesting to me, because I’m learning about machine learning and getting to think about infrastructure and user experience.

So, thanks to Mark, and his partner Mark, for the education!

WWDC and MOOCs

I have been overwhelmed by the spread of what some have been calling MOOCs (massive, open, online classrooms). I remain enthusiastic about the potential for an internet-catalized revolution in education, but I think the real missing ingredient isn’t the content, or the scalable assessment tools, which is what these commercial efforts have been focusing on, what’s missing is culture and community.

I think there are interesting parallels to Apple’s WWDC even, which has been selling out in record time:

Many more people want to attend WWDC than the conference can accommodate. There has been no shortage of interesting suggestions for how to fix this. Broadly speaking, WWDC has not changed in decades. Apple and its developer ecosystem, on the other hand, are radically different than they were just five years ago. Something has to give.

via Hypercritical: The Lottery.

Apple has tried to remedy the situation somewhat by putting all the materials online, but it seems that many recognize that this isn’t sufficient because it doesn’t provide the cultural and community benefits of attending WWDC. Right now the practical size of WWDC is partially due to a limit on the available venues in San Francisco, but fundamentally it is the number of attendees to Apple’s engineers, what in education is called the student-faculty ratio. Lower is generally better. Apple, on the other hand, probably doesn’t want to get too big, and definitely doesn’t want to grow to fast because there are limits to the rate at which they can train qualified engineers they can train up in “the Apple Way”

2013 Apple WWDC Reactions

I’ve been skimming over the coverage of the keynote presentation at Apple’s WWDC and I thought I’d post some of my reactions:

  • New “Mavericks” branding of OS X releases is interesting. I didn’t know the significance of the name until I read that they’ll be naming releases after inspiring places in California. Mavericks is a surfing spot on Half Moon Bay, CA. I find this to be an interesting way to reinforce the ongoing “Designed by Apple in California” message, particularly since they are planning to start manufacturing one or more products in the US again.
  • 575,000,000 accounts with credit card info on file. That is huge.
  • $10 B paid to app developers, $5B in the past year:  Nice, clearly there is still a lot of opportunity for Apple developers.
  • Demo of AnkiDrive, Bluetooth remote control cars by an independent developer: Taken together with the last detail, the message I take is that “indie” hardware makers are foolish to ignore the Apple ecosystem.
  • Finder Tabs & Tags sound good.
  • Better multi-monitor support sounds great (and overdue). Might be enough to get me using my 24″ monitor again.
  • Compressed Memory:  Interesting to see old ideas make a return. 15 years ago, RAM Doubler did the same thing, and more recently, a compressed VM backing store made a reappearance in Linux recently. I’m curious about whether this will breathe some more life into my wife’s old MacBook Pro which is limited to 6GB of memory, not enough with multitab web browsing.
  • Timer Coalescing & App Nap for backgrounded Safari Tabs. Anything to improve battery life is great, and Safari is an excellent target. Most background CPU usage on my machine tends to be web pages.
  • iCloud Keychain I wonder if this will work between Chrome on the desktop (which uses the OSX keychain) and Mobile Safari
  • Maps on OS X Nice, how about bringing it to the web too, please?
  • Overall it looks like Mavericks is another big step in making our computing environment seamless across Mac, iPhones and iPads. This is something that is harder for others to compete with. Samsung doesn’t have a real desktop presence. Google is trying, but they can’t reap the benefits of tight hardware integration. Microsoft could do it, but its marketshare in phones and tablets is smaller than Apple’s marketshare in the traditional PC market.
  • General release of Mavericks is this fall. I’m looking forward to it.
  • MacBook Air sounds like a nice upgrade, and the bump up to 802.11ac networking sounds good too.
  • Mac Pro a cylinder?!?!?!
  • External expansion?!?!? via Thunderbolt 2 (six of them!). I don’t think I like that if I were the intended customer. Whether or not I’d want to stuff extra cards into the case, the ability to stuff a bunch of disks inside is nice.  On the other hand, this should give a boost to the market for Thunderbolt peripherals.
  • No Intel Xeon Phi: Oh well. I thought Apple might offer a MacPro with ridiculous performance on multithreaded x86 apps. They did, however, stuff in two dual-OpenCL capable AMD FirePro GPUs in addition to a 12 core Intel CPU.
  • Tiny compared to the old MacPro, 1/8th the volume.
  • Designed by Apple in California. Assembled in the USA.
  • 300 million iCloud accounts: An impressive number, if only iCloud it sucked less.
  • iCloud version of iWork allows editing in the cloud. Not bad!
  • 600 million iOS devices sold.  That seems like a lot.
  • 93% of iOS users on on latest OS.  Very good, and a datapoint one could use in estimating what % of those 600 million devices sold are still in use.
  • iOS 7… Ok, so the notifications panel looks Windows Phone-like to me…
  • Control Center looks nice and useful. I still don’t see why it is so essential to have quick access to WiFi and Bluetooth settings, but I might feel differently if I was getting on planes multiple times per week.
  • Adaptive background app scheduling and push notifications to update apps in background (I initially misunderstood this later features as App store updates). Background processing comes to all iOS apps!
  • Automatic background app-store updates. Continuous deployment comes to iOS apps. I wonder if developers will be able to do staged deployments…
  • Mobile Safari.  I like the new tab switching!
  • AirDrop for sharing with other nearby iOS device users. Nice. It is a bit surprising it has taken them so long to enable this. Yes, some android phones have been able to do this sort of thing for a while, but this is a situation where the uniformity of the iOS ecosystem works in Apple’s favor.
  • Not supported on the iPhone 4s. Bah. The implication is that this is a hardware limitation. If that isn’t true, it seems like a stupid limitation to impose since the value of AirDrop increases exponentially with the number of people who have devices that support it.
  • Photos app. Nice to see Apple making improvements in this area, even if it includes some of the ideas we came up with for Wideangle. They seem to be missing an important use case though that Wideangle has also forsaken so-far. I’d tell you more, but, well, I can’t.
  • PhotoStreams become multi-user. It is about damn time, really. Question is, whether they can take some ground from Facebook…
  • Siri can control bluetooth, brightness. Can she trigger airplane mode, or would that be suicidal (Siri needs an internet connection)?
  • Bing search results in Siri! If I’m understanding this correctly, Apple has added web search results to the range of options it weighs in Siri, rather than making it a fall-through of last-resort like it is today. Oh, and they aren’t using Google…
  • iOS integration with automobiles: Looks nice, particularly the maps feature. Another reason why Apple and Google divorced over maps. What if I could just install a iPad mini in my dash though…
  • Location specific app recommendations. Cool idea.
  • iRadio, free, with ads, or without ads as part of iTunes Match. Nice.
  • Remote device lock for lost/stolen phones. On the one hand, I hate that my iOS devices are centrally controlled to such an extent. On the other hand, this has to be a major deterrent to device thieves.
  • Audio-only FaceTime: A further step toward commoditizing the mobile carriers. This is a feature where market share matters.
  • 1500 new APIs, including iBeacons for Bluetooth LE location.  Hmm. I’d like to hear more, but I guess that is what the rest of WWDC is for…
  • Final release this fall…
  • And that’s it. No killer new product category. The whining pundits will have something to keep themselves busy for a while. I think though that Apple understands their business better than most of the commentators…

Comcast expands its mobile reach.

Comcast is now turning the WiFi routers it rents to cable Internet customers into neighborhood hotspots.

Comcast is transforming its customers’ home modems into public Wi-Fi hotspots by adding a second signal to each device. In addition to a customer’s home Wi-Fi connection, Xfinity wireless gateways (which include the cable modem and wireless router) will by default broadcast a separate signal that other Comcast subscribers can log in to with a Comcast username and password.

via Comcast turns your Xfinity modem into public Wi-Fi hotspot | Ars Technica.

I am reminded again of low-cost mobile carrier Republic Wireless, which offloads mobile VoIP and data traffic to WiFi whenever possible. One of the things that intrigues me about Republic Wireless is that it’s parent company, Bandwidth.com has a line of business providing VoIP termination to cable ISPs.

On Apple’s Hiring of Kevin Lynch, Former Adobe CTO

Yesterday the news came out that Apple had hired Kevin Lynch away from Adobe, where he served as CTO. The hire hasn’t been without controversy.

Over on Daring Fireball, John Gruber reacted to the news with contempt and disbelief, pointing out that while at Adobe, Lynch had displayed questionable judgement in his championing of Flash at the expense of Apple and iOS:

Lynch wasn’t just an employee pushing the company line. As CTO, he was the guy who defined the company line — and his line had Adobe still pushing for Flash on mobile devices over three years after the iPhone shipped.

Gruber concludes that Lynch is a “bozo.” He makes a strong case for pinning the label on lynch, but he fails to consider alternative explanations for the hire.

On Apple Insider, Daniel Eran Dilger has a different take on the hire. He points out that Lynch has a long and successful track record with digital media creation tools. He came to Adobe when they hired Macromedia, where he was their top technical and product exec. He was instrumental in the creation of Dreamweaver, and the Mac version of FrameMaker.

It is also worth noting that despite the fact that while the success of the iPhone and iOS caught Adobe, not to mention the rest of the tech-industry, flat-footed, Flash had a damn good run up until that point, and since then, Adobe has done a reasonable job establishing itself on iOS with end-user-apps, and tools for content creators, even without Flash.

I’ll suggest an alternative take: Apple hired Lynch as part of an ongoing effort to improve their tools for creating content and apps for iOS devices. Time will tell whether or not he was a good hire. Certainly other tech execs have fallen from grace, only to redeem themselves. Take Google’s Eric Schmidt, who had a great run at Google after getting beaten badly by Microsoft while leading Novell, or Steve Jobs, who was run out of Apple by a guy he himself hired and had a middling run with NeXT before returning to Apple and leading it to its current preeminence.