Spain's socialist government is about to undertake its largest-ever migrant regularization program, granting legal status and work permits to roughly 500,000 undocumented workers. As Europe hardens its borders, Madrid is betting on inclusion over exclusion.
Spain’s migrant regularization marks a rare progressive turn in Europe
Spain is about to undertake its largest-ever migrant regularization program, granting legal status and work permits to roughly 500,000 undocumented workers. Driven by grassroots campaigns and supporte...
canadiandimension.com
March 30, 2026 at 7:25 PM
Spain's socialist government is about to undertake its largest-ever migrant regularization program, granting legal status and work permits to roughly 500,000 undocumented workers. As Europe hardens its borders, Madrid is betting on inclusion over exclusion.
machine learning is the study of numerical errors and things which turn out to be L2 regularization wearing a different hat
March 20, 2026 at 8:50 PM
machine learning is the study of numerical errors and things which turn out to be L2 regularization wearing a different hat
the papers "This Thing We Thought Wasn't L2 Regularization Is Actually L2 Regularization" and "Here Is A New Alternative To Gradient Methods Which Doesn't Work" will still be being written on the day i die
July 15, 2025 at 5:49 AM
the papers "This Thing We Thought Wasn't L2 Regularization Is Actually L2 Regularization" and "Here Is A New Alternative To Gradient Methods Which Doesn't Work" will still be being written on the day i die
a lot of machine learning research is about discovering which parts of mathematics are actually L2 regularization and which parts of mathematics are actually Adam
January 4, 2025 at 6:58 AM
a lot of machine learning research is about discovering which parts of mathematics are actually L2 regularization and which parts of mathematics are actually Adam
Between the recent undocumented immigration regularization and being a Europe-leader on LGBT rights, Spanish Empire 2.0?
June 22, 2026 at 1:26 PM
Between the recent undocumented immigration regularization and being a Europe-leader on LGBT rights, Spanish Empire 2.0?
like, i am not currently privy to frontier lab discussions on this stuff the way i was last year, but i think that basic issues like "mesa-optimization falls directly out of the math of sequential training on non-IID data with weight-pinning regularization" needs attention
bsky.app/profile/theo...
bsky.app/profile/theo...
when a model is trained, it is trained on data X, subject to mathematical conditions {Y, Z, [...]}. trivially, a model trained on a particular corpus and then fine-tuned on subsequent data subject to a KL-div penalty from the reference distribution can have a basin of attraction predicted by neither
September 25, 2026 at 2:43 PM
like, i am not currently privy to frontier lab discussions on this stuff the way i was last year, but i think that basic issues like "mesa-optimization falls directly out of the math of sequential training on non-IID data with weight-pinning regularization" needs attention
bsky.app/profile/theo...
bsky.app/profile/theo...
regularization is when the optimizer fucks up and has to back out of what it mistakenly thought was an optimum and the more times it has to do that the more regularization occurs
September 3, 2025 at 2:18 AM
regularization is when the optimizer fucks up and has to back out of what it mistakenly thought was an optimum and the more times it has to do that the more regularization occurs
meshoptimizer v0.25 is out! Featuring new simplification function that optimizes positions and attributes for appearance, experimental permissive mode to simplify faceted regions with selective seam preservation, regularization option for improved tessellation quality and deformation, multiple ...
August 20, 2025 at 3:56 PM
meshoptimizer v0.25 is out! Featuring new simplification function that optimizes positions and attributes for appearance, experimental permissive mode to simplify faceted regions with selective seam preservation, regularization option for improved tessellation quality and deformation, multiple ...
NeurIPS friends! Come hang out at 4:30p to hear about my student Chris Scarvelis’ work “Nuclear Norm Regularization for Deep Learning.” Chris proved a clever/subtle formula making Jacobian nuclear regularization practical for networks, with application to denoising and manifold regularization.
December 13, 2024 at 5:13 PM
NeurIPS friends! Come hang out at 4:30p to hear about my student Chris Scarvelis’ work “Nuclear Norm Regularization for Deep Learning.” Chris proved a clever/subtle formula making Jacobian nuclear regularization practical for networks, with application to denoising and manifold regularization.
Oh great, Shriram is going to talk about VC dimension, regularization, and clustering, right?
OMG where to start. There's so much learning theory that I think is relevant…
November 27, 2024 at 5:02 AM
Oh great, Shriram is going to talk about VC dimension, regularization, and clustering, right?
*Relaxed Equivariance via Multitask Learning*
by @tkrusch.bsky.social @mmbronstein.bsky.social
They propose a regularization approach for exploiting symmetries over data (penalizing variable predictions over augmented data).
arxiv.org/abs/2410.17878
by @tkrusch.bsky.social @mmbronstein.bsky.social
They propose a regularization approach for exploiting symmetries over data (penalizing variable predictions over augmented data).
arxiv.org/abs/2410.17878
December 13, 2024 at 2:28 PM
*Relaxed Equivariance via Multitask Learning*
by @tkrusch.bsky.social @mmbronstein.bsky.social
They propose a regularization approach for exploiting symmetries over data (penalizing variable predictions over augmented data).
arxiv.org/abs/2410.17878
by @tkrusch.bsky.social @mmbronstein.bsky.social
They propose a regularization approach for exploiting symmetries over data (penalizing variable predictions over augmented data).
arxiv.org/abs/2410.17878
listen here mister, sometimes it's L1 regularization
March 20, 2026 at 9:09 PM
listen here mister, sometimes it's L1 regularization
This month, Spain is set to launch a mass regularization program for migrants that could give legal status to hundreds of thousands of undocumented people living in the country.
Spanish Immigration Is Different
Madrid is preparing for one of the most ambitious regularization schemes in European history.
foreignpolicy.com
April 13, 2026 at 11:30 PM
This month, Spain is set to launch a mass regularization program for migrants that could give legal status to hundreds of thousands of undocumented people living in the country.
In Spain's Almeria province, Europe's most protected horticultural region, the government's mass regularization of undocumented migrants is testing whether long-criticized conditions under the region's plastic greenhouses can improve.
May 27, 2026 at 1:13 AM
In Spain's Almeria province, Europe's most protected horticultural region, the government's mass regularization of undocumented migrants is testing whether long-criticized conditions under the region's plastic greenhouses can improve.
Rectified LpJEPA
A JEPA architecture that learns sparse, non-negative, informative representations through principled distributional regularization.
A JEPA architecture that learns sparse, non-negative, informative representations through principled distributional regularization.
February 4, 2026 at 12:40 PM
Rectified LpJEPA
A JEPA architecture that learns sparse, non-negative, informative representations through principled distributional regularization.
A JEPA architecture that learns sparse, non-negative, informative representations through principled distributional regularization.
Thou shalt not overfit to the training data
But more importantly
Thou shalt not overestimate the
ability of regularization approaches to prevent overfitting issues
But more importantly
Thou shalt not overestimate the
ability of regularization approaches to prevent overfitting issues
February 19, 2025 at 5:13 PM
Thou shalt not overfit to the training data
But more importantly
Thou shalt not overestimate the
ability of regularization approaches to prevent overfitting issues
But more importantly
Thou shalt not overestimate the
ability of regularization approaches to prevent overfitting issues
it’s just missing an explanation of what “random” means and how that “random” machinery creates smoothing and regularization
Need to explain (or understand) linear mixed effects regressions, random intercepts, and random slopes? Look no further than "A Visual Introduction to Hierarchical Models" by Michael Freeman, 2017. It's a banger!
A Visual Introduction to Hierarchical Models
A visual explanation of multi-level modeling
mfviz.com
March 27, 2025 at 11:31 PM
it’s just missing an explanation of what “random” means and how that “random” machinery creates smoothing and regularization
3. However, this leads to centering punishing ICE rather than helping immigrants (through regularization of status, increasing visa quotas, etc.). There are viral posts here about the need to prosecute ICE members after Trump leaves office. There aren't any about giving immigrants green cards.
August 27, 2025 at 4:47 PM
3. However, this leads to centering punishing ICE rather than helping immigrants (through regularization of status, increasing visa quotas, etc.). There are viral posts here about the need to prosecute ICE members after Trump leaves office. There aren't any about giving immigrants green cards.
“Data Science: Building Machine Learning Models (HarvardX)”: Machine learning algorithms, principal component analysis, regularization by building a movie recommendation system, training data, predictive relationships. Enroll today: edx.sjv.io/GbGeGB AI Academy😀
June 9, 2026 at 7:45 PM
“Data Science: Building Machine Learning Models (HarvardX)”: Machine learning algorithms, principal component analysis, regularization by building a movie recommendation system, training data, predictive relationships. Enroll today: edx.sjv.io/GbGeGB AI Academy😀
4️⃣ Regularization (L1, low-rank)
Both types of structural regularization reduce degeneracy across all levels. Regularization nudges networks toward more consistent, shared solutions.
Both types of structural regularization reduce degeneracy across all levels. Regularization nudges networks toward more consistent, shared solutions.
November 24, 2025 at 4:43 PM
4️⃣ Regularization (L1, low-rank)
Both types of structural regularization reduce degeneracy across all levels. Regularization nudges networks toward more consistent, shared solutions.
Both types of structural regularization reduce degeneracy across all levels. Regularization nudges networks toward more consistent, shared solutions.
🧠 Regularization, Action, and Attractors in the Dynamical “Bayesian” Brain
direct.mit.edu/jocn/article...
(still uncorrected proofs, but they should post the corrected one soon--also OA is forthcoming, for now PDF at brainandexperience.org/pdf/10.1162-...)
direct.mit.edu/jocn/article...
(still uncorrected proofs, but they should post the corrected one soon--also OA is forthcoming, for now PDF at brainandexperience.org/pdf/10.1162-...)
Regularization, Action, and Attractors in the Dynamical “Bayesian” Brain
Abstract. The idea that the brain is a probabilistic (Bayesian) inference machine, continuously trying to figure out the hidden causes of its inputs, has become very influential in cognitive (neuro)sc...
direct.mit.edu
October 22, 2025 at 8:59 AM
🧠 Regularization, Action, and Attractors in the Dynamical “Bayesian” Brain
direct.mit.edu/jocn/article...
(still uncorrected proofs, but they should post the corrected one soon--also OA is forthcoming, for now PDF at brainandexperience.org/pdf/10.1162-...)
direct.mit.edu/jocn/article...
(still uncorrected proofs, but they should post the corrected one soon--also OA is forthcoming, for now PDF at brainandexperience.org/pdf/10.1162-...)
Regularization is hot since more than half a century.
Name one idea that’s still hot 10 years on. Kernel methods? Gone. SVMs? Out. GMMs? Nothing. diffusion models 10 years from now? Probably a quaint -and inefficient- idea
November 28, 2024 at 7:07 PM
Regularization is hot since more than half a century.
New paper with @leonlufkin.bsky.social and @eringrant.bsky.social!
Why do we see localized receptive fields so often, even in models without sparisity regularization?
We present a theory in the minimal setting from @aingrosso.bsky.social and @sebgoldt.bsky.social
Why do we see localized receptive fields so often, even in models without sparisity regularization?
We present a theory in the minimal setting from @aingrosso.bsky.social and @sebgoldt.bsky.social
December 13, 2024 at 10:49 AM
New paper with @leonlufkin.bsky.social and @eringrant.bsky.social!
Why do we see localized receptive fields so often, even in models without sparisity regularization?
We present a theory in the minimal setting from @aingrosso.bsky.social and @sebgoldt.bsky.social
Why do we see localized receptive fields so often, even in models without sparisity regularization?
We present a theory in the minimal setting from @aingrosso.bsky.social and @sebgoldt.bsky.social
I'm fully against any further regularization of English's beautiful strong verbs. One thing I'll be an open and diehard prescriptivist over. If its the past participle you WILL say "molten" not "melted", "dreamt" not "dreamed", "burnt" not "burned
January 16, 2026 at 6:44 AM
I'm fully against any further regularization of English's beautiful strong verbs. One thing I'll be an open and diehard prescriptivist over. If its the past participle you WILL say "molten" not "melted", "dreamt" not "dreamed", "burnt" not "burned