Top all-time · #1498
Ferenc Huszár
Personal blog · inference.vcaimathscience
Ferenc Huszár writes inFERENCe, a personal blog at inference.vc about ai, math and science. Since 2014, 50 different Hacker News readers have linked it 64 times, most often “Deep Learning Is Easy – Learn Something Harder”. It is still writing: the last post was on 2026-02-25, and you can follow it by feed.
A blog about machine learning research, deep learning, causal inference, variational learning, by Ferenc Huszár.
Most-linked posts?
- Deep Learning Is Easy – Learn Something Harder6 people · 6 mentions · first linked 2016-01-25
- We May be Surprised Again: Why I take LLMs seriously4 people · 4 mentions · first linked 2023-03-23
- ML Beyond Curve Fitting: An Intro to Causal Inference and Do-Calculus3 people · 3 mentions · first linked 2018-05-24
- Causal Inference 3: Counterfactuals3 people · 3 mentions · first linked 2019-01-30
- Gaussian Distributions Are Soap Bubbles3 people · 3 mentions · first linked 2017-11-10
- GANs are broken in more than one way: the numerics of GANs3 people · 3 mentions · first linked 2017-10-05
- /2 people · 2 mentions · first linked 2020-02-08
- “Everything that works works because it's Bayesian” Why Deep Nets Generalize?2 people · 2 mentions · first linked 2017-05-25
- Notes on the Origin of Implicit Regularization in SGD2 people · 2 mentions · first linked 2025-05-08
- Eastern European Guide to Writing Reference Letters2 people · 2 mentions · first linked 2022-03-13
What Hacker News says?
My mentor was very very adamant about Bayesian network and hierarchical as being deep learning. He sees the latent layer in the hierarchical model as the hidden layer and the Bayesian just have a strict restrictions/assumptions to the network where as the deep learning is more dumb and less assuming. A few of my professor thinks that PGM, probability graphical model is a super set of deep learning/neural network. Th…
Comment linking /everything-that-works-works-because-its-bayesian-2/
on “AI winter is well on its way”Submitted Deep Learning Is Easy – Learn Something Harder as a story · 244 points
HN discussionFirst of all, let me say that what you all have put together is truly excellent. It covers the sort of things toy treatments and papers leave out or gloss over if you actually want to get something working in the real world. That said, I strongly disagree with your disagreement. There was a recent paper whose abstract I read that made me think of homotopies between convolutional networks. Unfortunately I lost the pa…
Comment linking /deep-learning-is-easy/
on “Practical Deep Learning for Coders”Submitted A Cookbook for Machine Learning: Vol 1 as a story · 193 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 ML Beyond Curve Fitting: An Intro to Causal Inference and Do-Calculus as a story · 184 points
HN discussion> do you think math could have ever come into being if it weren't to fill the human need of describing and categorizing the world? I'm a bit confused. What exactly is the counterfactual[0] here? If it is hyper-specific to categorizing and describing then I think yes, those creatures could still invent math. But my confusion is because I'm having a difficult time thinking where such things aren't also necessary conse…
Comment linking /causal-inference-3-counterfactuals/
on “The language brain matters more for programming than the math brain? (2020)”Submitted Gaussian Distributions Are Soap Bubbles as a story · 140 points
HN discussion> without telling me about the specific results that your argument rests on We've been discussing it the whole time. You even repeated it in the last comment. A model that is accurate does not need to be causal By causal I mean that the elements involved are directly related. We've seen several examples. The most complex one I've mentioned is the geocentric model. People made very accurate predictions with their mod…
Comment linking /causal-inference-2-illustrating-interventions-in-a-toy-example/
on “Persona vectors: Monitoring and controlling character traits in language models”Submitted Eastern European Guide to Writing Reference Letters as a story · 110 points
HN discussionIn vanilla policy gradient, one plays the game to the end and then bumps the probability of all actions taken by the agent up (if AlphaGo won) or down (if it lost). This is very slow because there ~150 moves in an expert game, and we do not know which moves caused decisive victory or loss - i.e. the problem of "long term credit assignment". Also, it is actually preferable to compute the policy gradient with respect…
Comment linking /alphago-zero-policy-improvement-and-vector-fields/
on “Building an AlphaZero AI using Python and Keras”Submitted ML Beyond Curve Fitting: An Intro to Causal Inference and Do-Calculus (2018) as a story · 105 points
HN discussionExploration versus Exploitation: Deciding When to Get Married - http://www.inference.vc/my_big_fat_data_driven_wedding_how_t...
Comment linking /my_big_fat_data_driven_wedding_how_to_decide_when_to_get_married/
on “How to Pick Your Life Partner (2014)”Submitted “Everything that works works because it's Bayesian” Why Deep Nets Generalize? as a story · 40 points
HN discussionA few remarks: In the "Doing bad digital humanities with color vectors", if you consider colors as 3D vectors, which they do, you'll see that summing enough uniformly sampled vectors always gives medium browns, because that's the color in the middle of the colorspace. Instead, you should model colors in a polar space and sum vectors in that space. This will prevent going inside the sphere and losing color saturation…
Comment linking /high-dimensional-gaussian-distributions-are-soap-bubble/
on “Understanding word vectors”Submitted ML beyond Curve Fitting: An Intro to Causal Inference and do-Calculus (2018) as a story · 20 points
HN discussion...because stochastic methods are implicit regularizers, leading to solutions that generalize better. Let's spell it out for those that don't know. https://www.inference.vc/notes-on-the-origin-of-implicit-reg...
Comment linking /notes-on-the-origin-of-implicit-regularization-in-stochastic-gradient-descent/
on “How linear regression works intuitively and how it leads to gradient descent”Submitted Deep Learning Is Easy – Learn Something Harder as a story · 14 points
HN discussionI think you are on to something. this post was on HN a few weeks ago about how deep learning is no longer interesting because it has become routine. My first thought when I read that was "deep learning just became interesting because now it's stable enough to build a business around". The end of one cycle is just the beginning of another one. Find out what that new cycle is, hop on, and strap in. http://www.inferenc…
Comment linking /deep-learning-is-easy/
on “After the Gold Rush”Submitted The secular Bayesian: Using belief distributions without believing as a story · 6 points
HN discussionAbsolutely no. Random search does not work in high dimensions because of the "curse of dimensionality" - the number of directions to search grows exponentially with dimension. Gradient descent avoids the problem because it knows the right direction. Evolutionary strategies are doing (natural) gradient descent using an estimate of the gradient. Here's a good article: http://www.inference.vc/evolutionary-strategies-em…
Comment linking /evolutionary-strategies-embarrassingly-parallelizable-optimization/
on “The quest to evolve neural networks through evolutionary algorithms”Submitted The Future of Software as a story · 4 points
HN discussionThis reminds me of Ferenc's reminder that 'deep learning is easy' - ' http://www.inference.vc/deep-learning-is-easy/
Comment linking /deep-learning-is-easy/
on “Confession of a so-called AI expert”Submitted We May be Surprised Again: Why I take LLMs seriously as a story · 4 points
HN discussionSearch for example on "deep learning". Results: + nice: open books (deeplearningbook.org) + not nice: online seminars (udemy, udacity, edx) + not nice: wikipedia copycats (investopedia) + not nice: SEO optimized business analytics sites recognizable by their top 10 this, top 20 that (kdnuggets) + not nice: commercial software (mathworks) + not nice: big companies hitchhiking on buzzwords (sas) What would be nice are…
Comment linking /
on “Amazon, Facebook and Google can all be beaten”Submitted Discrete Diffusion: Continuous-Time Markov Chains as a story · 3 points
HN discussionOr how the volume of an n-ball goes to 0[0,1] Or how gaussian balls are like soap bubbles[2] The latter of which being highly relevant to vector embeddings. Because if you aren't a uniform distribution, the density of your mass isn't uniform. MEANING if you linearly interpolate between two points in the space you are likely to get things that are not representative of your distribution. It happens because it is easy…
Comment linking /high-dimensional-gaussian-distributions-are-soap-bubble/
on “A visual exploration of vector embeddings”Submitted GANs are broken in more than one way: the numerics of GANs as a story · 3 points
HN discussionMeasuring causality is hard. Without randomized control trials, you pretty much need to assume some causal graph structure (plain data is not enough). See: "ML beyond Curve Fitting: An Intro to Causal Inference and do-Calculus" http://www.inference.vc/untitled/ When it comes to this SO vs HN article based on data one can look at correlations. Even if some delayed correlations, then still - it does not imply causatio…
Comment linking /untitled/
on “Popularity of technology on Stack Overflow and Hacker News: Causality Analysis”Submitted GANs Are Broken in More Than One Way: The Numerics of GANs as a story · 3 points
HN discussion
Linked by the same people?
- S
Sebastian Ruderruder.io - J
Julian Togeliustogelius.blogspot.com - LLil'Loglilianweng.github.io
- 大大トロ ・ Machine Learningblog.otoro.net
- S
State of the Smeritysmerity.com - P
Piekniewski's blogblog.piekniewski.info - A
Algorithmic Assertions - Craig Gidney's Computer Science Blogalgassert.com - FFrançois-René Rideaufare.livejournal.com