vonnik

80 ms·

691.
by vonnik 10y ago
This is a great post as far as it goes. While Abhishek mentions Keras for neural networks (and Keras is a great Python library), he doesn't really go into the cases where deep neural networks are more useful than other algorithms, and
692.
by vonnik 10y ago
I agree with dbfclark that it's case by case. At Skymind ( https://skymind.io/ ), we support an open-source Java library ( http://deeplearning4j.org/ ). * So obviously Java, Scala, Clojure, and lower-lev
693.
by vonnik 10y ago
TensorFlow has kind of won the Python deep-learning community, although frameworks like Keras[0] make it easier to use. While Python programmers make up the majority of deep-learning practitioners, they don't have much penetration in e
694.
by vonnik 10y ago
Here's an intro to ConvNets in Java: http://deeplearning4j.org/convolutionalnets.html Karpathy's stuff is also great: https://cs231n.github.io/
695.
by vonnik 10y ago
Can we correct the headline to say "Turkish government"?
696.
by vonnik 10y ago
I'm curious to see whether clients actually want to move very large datasets through these APIs, or whether that's too costly. It strikes me that cloud services work best for data generated and managed in the same cloud...
697.
by vonnik 10y ago
Good to know, thanks! I'm a bit skeptical of the words "their own, unique algorithms", given the quality and pace of research in machine learning and deep learning, but that may just be my open-source bias...
698.
by vonnik 10y ago
I agree with you. When I say services, I'm actually talking about the consulting work that accompanies enterprise software sales, where you end up acting as an outsourced engineering team for your clients.
699.
by vonnik 10y ago
> Open source is wind in your back. So true. Open source lets small engineering teams punch above their weight. Thousands of extra eye balls on the code for QA make it so much better. And the project-user feedback loop is fast.
700.
by vonnik 10y ago
All open-source companies close-source some code, or else they struggle, because services don't scale. Red Hat's security layer is not open-source. Cloudera is an explicitly open-core company. So is my startup, Skymind.[1][2] Ther
701.
by vonnik 10y ago
The data may show a correlation, but it would show the same correlation to everyone, which means that the advantage of investing in Ivy Leaguers/4.0 GPA students is priced out. Anecdotally, I've seen a lot of excellent academics a
702.
by vonnik 10y ago
Streamlining the pitch process benefits everyone, because both investors and startups seeking funding have a lot of pitch sessions to wade through. Fund-raising is a long matching process, and the more you can make it like speed dating for
703.
by vonnik 10y ago
It would be great to know which algorithms are being used. Hard to know how to evaluate AI claims without that...
704.
by vonnik 10y ago
In a sense, replacing or augmenting labor is a false dichotomy. We can’t actually choose one or the other. ("Capital-Labor substitution" is a thing, in economics; it's structurally unavoidable.) Most nations will choose both,
705.
by vonnik 10y ago
Right, but the question is: Whose cloud and what kind of cloud? Are we talking private cloud, virtual private cloud? Who manages it? Even saying "as long as the compute is in the same data center" is a huge assumption. I think it&
706.
by vonnik 10y ago
The history of machine learning startups is littered with companies that thought a hosted web service was a good idea. The problem with this model is that big data, by definition, is costly to move. So if a managed service is not generating
707.
by vonnik 10y ago
He told people to spend more time with their parents. That's pretty good advice about how to relate to other human beings.
708.
by vonnik 10y ago
This was a long presentation. Yann gets to the issue in the title of the post about one third of the way through. The first third should probably be called "What's right with deep learning, or how DL works"... For each proble
709.
by vonnik 10y ago
Here's a collection of free online resources, including Ian Goodfellow and Yoshua Bengio's excellent book: http://deeplearning4j.org/deeplearningforbeginners.html
710.
Show HN: Deeplearning4j 4.0: Cudnn Wrapper, Multi-GPUs, Spark(deeplearning4j.org)
1 points by vonnik 10y ago|0 comments
711.
by vonnik 10y ago
totally agree! digging in a little more, he talked about the cycle of: cue -> craving -> action -> reward and the whole question is, how do you find rewards that instill craving which will lead you to an action that changes your li
712.
by vonnik 10y ago
I've spoken with the TF team and they've been clear that open-sourcing the lib was part of strategy to make Google Cloud more appealing. It was conceived as a lure, among other things.
713.
by vonnik 10y ago
We wrap Cudnn, which among the top performers.
714.
by vonnik 10y ago
Read The Power of Habit by Charles Duhigg. He has some good suggestions. One of the best ideas in there is that certain small habit changes create behavioral cascades. For people trying to lose weight, just writing down what they eat at eac
715.
by vonnik 10y ago
We already built what Amazon was looking for: An extensible deep-learning framework that works on distributed CPUs and GPUs heterogeneously, integrating with Spark as an access layer to orchestrate multiple host threads. https://
716.
by vonnik 10y ago
Whatever they are, they're either similar to algorithms that some human has used previously, or the employ mathematical ideas that others would have named and understood. So you could explain it by saying it's like X but with Y mo
717.
by vonnik 10y ago
I agree with you about redlining. I hadn't really thought about that... A recent scandal involving a predictive analytics firm called Northpointe shows how stupidly people use algorithms, and how much they are abused when there is no t
718.
by vonnik 10y ago
Stephen makes a lot of good points. And the list of questions he says people should ask is also useful. I have a few to add: * What algorithm(s) are you using? Nobody should be able to brag publicly about their AI without simply naming the
719.
by vonnik 10y ago
What's even more impressive, in a way, is that Bloomberg managed to write this story. Because Bloomberg is one of the most rigorously fact-checking news organizations around. Speaking as a former reporter there, it's hard to even
720.
by vonnik 10y ago
Strange does not begin to cover it. Santeria, hitmen, embezzlement and animal sacrifice in a huge, doomed Caribbean bank...

More ›