Top all-time · #786
Machine Learning (Theory)
Personal blog · hunch.netaiscience
Machine Learning (Theory) is a personal blog at hunch.net about ai and science. Since 2007, 81 different Hacker News readers have linked it 101 times, most often “AlphaGo is not the solution to AI”. It is still writing: the last post was on 2026-09-21, and you can follow it by feed.
Sometimes recommended · 10 of 40 sampled comments recommend it?
Why HN links it
HN readers link John Langford's Machine Learning (Theory) blog for practical and theoretical machine learning, bandit and exploration ideas, and commentary on ML research practice.
Summarised by Claude from a sample of 40 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Most-linked posts?
- AlphaGo is not the solution to AI6 people · 6 mentions · first linked 2016-03-14
- Vowpal Wabbit: fast, scalable, useful learning algorithm6 people · 6 mentions · first linked 2008-09-06
- Clever methods of overfitting5 people · 5 mentions · first linked 2012-01-04
- John Langford & much of Yahoo Research NY Going to MSR5 people · 5 mentions · first linked 2007-11-13
- All Models of Machine Learning Have Flaws (2007)4 people · 4 mentions · first linked 2010-06-28
- Scaling Up Machine Learning, the Tutorial, KDD 20113 people · 3 mentions · first linked 2011-08-22
- John Langford on the end of YahooResearch3 people · 3 mentions · first linked 2012-05-03
- Metacademy: a package manager for knowledge3 people · 3 mentions · first linked 2014-02-16
- Programming Languages for Machine Learning2 people · 3 mentions · first linked 2007-10-23
- Hadoop AllReduce and Terascale Learning2 people · 2 mentions · first linked 2011-12-03
What Hacker News says?
John Langford's page is here: http://hunch.net/~jl/
Comment linking /~jl/
on “20 lines of code that beat A/B testing every time”Submitted When the bubble bursts as a story · 114 points
HN discussionThis is the most important critique of A/B testing. It far outweighs the traditional hoopla about simultaneous inference and Bonferonni corrections. Epsilon greedy does well on k-armed bandit problems, but in most applications you likely can do significantly better by customizing the strategy to individual users. That's a contextual bandit and there are simple strategies that to pretty well here too. For instance: h…
Comment linking /~exploration_learning/main.pdf
on “20 lines of code that beat A/B testing every time”Submitted A tutorial on active learning (2009) [pdf] as a story · 85 points
HN discussionI love reading these two for Python/Recommender Systems: 1. Ben Frederickson: https://www.benfrederickson.com/blog/ 2. Erik Bernhardsson: https://erikbern.com/ And for pure ML: 1. Ferenc Huszár: https://www.inference.vc/ 2. John Langford: https://hunch.net (although he hasn't written a technical post in a while)
Comment linking /
on “Ask HN: What are good solo developer blogs that you enjoy reading?”Submitted Programming Languages for Machine Learning as a story · 67 points
HN discussionI'm not sure if it's a deep learning framework. Perhaps they are in some ways complementary http://hunch.net/?p=2875548 . Who knows, maybe CNTK is one big reduction.
Comment linking /?p=2875548
on “Microsoft releases CNTK, its open source deep learning toolkit, on GitHub”Submitted All Models of Machine Learning Have Flaws (2007) as a story · 57 points
HN discussionMissing Vowpal Wabbit: http://hunch.net/~vw/
Comment linking /~vw/
on “Microsoft releases CNTK, its open source deep learning toolkit, on GitHub”Submitted Vowpal Wabbit: fast, scalable, useful learning algorithm as a story · 15 points
HN discussionI can totally appreciate the value of having well defined and well understood definitions and notations that might require lots of background to understand. This allows for precise and concise discourse. But there have been many published bounds in Machine Learning Theory that turned out to be completely vacuous. That is, they prove a bound on generalization error that is always greater than 1. http://hunch.net/?p=1…
Comment linking /?p=181
on “Machine learning is easier than it looks”Submitted Three levels of addressing the Netflix Prize as a story · 15 points
HN discussionTo use the test set explicitly for evaluation is a deadly sin. When found out, you'd face serious damage to your reputation (Like Baidu did a few years ago). [1] What the decreasing performance results on the remade CIFAR-10 test set shows, is probably more akin to a subtle form of overfitting (Due to these datasets being around for a long time, the good results get published, and the bad results discarded, leading…
Comment linking /?p=22
on “Myths in Machine Learning Research”Submitted Clever methods of overfitting as a story · 14 points
HN discussionPreface: AlphaGo is an amazing achievement and does show an interesting advancement in the field. Yet ... it really doesn't mean almost anything that people are predicting it to mean. Slashdot went so far as to say that "We know now that we don't need any big new breakthroughs to get to true AI". The field of ML/AI is in a fight where people want more science fiction than scientific reality. Science fiction is sexy,…
Comment linking /?p=3692542
on “Yann LeCun's comment on AlphaGo and true AI”Submitted All Models of Machine Learning have Flaws as a story · 12 points
HN discussionUnknown is relative, but here are some of the rss feeds I'm subscribed too: http://nosql.mypopescu.com/ http://lambda-the-ultimate.org/ http://eli.thegreenplace.net/ http://www.stefankrause.net/wp/ http://mechanical-sympathy.blogspot.com/ http://www.cowtowncoder.com/blog/blog.html http://highlyscalable.wordpress.com/ http://hunch.net/ http://perspectives.mvdirona.com/ http://www.scottaaronson.com/blog/ http://www.so…
Comment linking /
on “Ask HN: What unknown technical blogs or sites do you read?”Submitted Critical issues in digital contract tracing as a story · 9 points
HN discussionA few others: Machine Learning (Theory), http://hunch.net/ , is a good place to learn about new machine-learning papers. It doesn't really do research exposition on the blog, but it posts about recent ML conferences, highlighting some of the papers the author finds interesting. Embedded in Academia, http://blog.regehr.org/ , is about 50% personal stuff, but 50% posts on John Regehr's work on C-compiler fuzzing, with…
Comment linking /
on “A set of top Computer Science blogs”Submitted Effective Research Funding as a story · 8 points
HN discussionI'll ask him whether he considers himself mostly a pencil pushing fraud when I see him in December. I'll also see what he thinks of the claim that VW's usefulness is "not due to any theory". I'll think you'll be surprised given he's written a blog post on this topic titled "Use of Learning Theory" http://hunch.net/?p=496 at his blog which, by the way, is called "Machine Learning (Theory)". It's easy to dismiss what…
Comment linking /?p=496
on “An Intuitive Guide to Linear Algebra”- ot · 2016-03-14
Submitted AlphaGo is not the solution to AI as a story · 7 points
HN discussion Co-creator of Metacademy here. Colorado and I are intending Metacademy to be a “package manager for knowledge,” where you can easily find the particular thing you want to learn about (e.g. deep neural nets) without having to track down all the prerequisites (e.g. gradient descent) yourself. Basically, we’ve annotated a dependency graph for the core concepts in the field, and it uses the graph to produce step-by-step…
Comment linking /?p=2714
on “Metacademy”Submitted Programming Languages for Machine Learning Implementations as a story · 6 points
HN discussionDo note that the more the gate is kept, the stronger the incentive for cheating. Moreover, peer review is overrated. An experiment by the NIPS conference in 2014 found that ~60% of submissions are in the "gray middle" - not obviously great or obviously crap. In more detail: they split the PC in two, and had 10% of papers (166) reviewed by both halves. Each half had to accept ~37 papers. The halves disagreed on 21 ac…
Comment linking /?p=467864
on “Faked Beta-Amyloid Data. What Does It Mean?”Submitted Scaling Up Machine Learning, the Tutorial, KDD 2011 as a story · 6 points
HN discussionMetacademy ( http://www.metacademy.org/ ) is exactly what you described. The website gives a concept map and each concept has links to free educational resources. Here's the concept map for SVD: http://www.metacademy.org/graphs/concepts/singular_value_dec... You can read a bit about the ideas behind Metacademy here: http://hunch.net/?p=2714
Comment linking /?p=2714
on “Ask HN: Idea Sunday”Submitted A machine learning researcher's view on the Bing/Google debacle as a story · 5 points
HN discussionBest of luck. Do you have a hypothesis? More experienced ML folks don't seem to be on board [1] [2]. I haven't seen anyone posit how to build a machine that can set its own goals. I'm sure you'll build and learn something cool nonetheless Another organization that tried to build AI was called Thinking Machines [3]. They were good at parallel programming. A few of them would later form "Ab Initio Software" [4]. Their…
Comment linking /?p=3692542
on “My path to OpenAI”Submitted What is the most effective policy response to the new coronavirus pandemic? as a story · 4 points
HN discussion
