NewA calmer, faster ZeroHour — security assessments on articles, story timelines and search from any page.

#learning-dynamics

0 stories this week · 3 this month · 4 in 6 months · mostly AI

Stories per day · 30 days

Relatedconvolutional-networks1 feature-learning1 noise1 recurrent-neural-networks1 representation-learning1 sample-complexity1 statistical-mechanics1 stochastic-gradient-descent1 teacher-student1 theory1 two-layer-networks1 weight-sharing1

4 stories, newest first · one card per story

arXiv cs.AI / cs.LG / cs.CL16d ago

Theory: noise can shape what linear RNNs learn

In the rich regime, high noise can make a non-degenerate solution suboptimal when parameter scales are limited.

recurrent-neural-networks representation-learning

arXiv cs.AI / cs.LG / cs.CL21d ago

Theory separates locality from weight sharing in convolutional learning

With uncorrelated inputs a convolutional student learns exactly P times faster, but spatial correlation can erase that edge.

convolutional-networks weight-sharing

1

arXiv cs.AI / cs.LG / cs.CL26d ago

Mesoscopic theory describes SGD feature learning in wide two-layer networks

Gaussian overlap statistics track linear and quadratic learning until specialization, which quadratic shielding accelerates.

learning-dynamics two-layer-networks

1

arXiv cs.AI / cs.LG / cs.CLSep 24, 2026

Intrinsic-Extrinsic Coupling in Learning Dynamics

Researchers show replay can erase a classifier-head intervention's gains, formalizing intrinsic-extrinsic coupling in learning.

continual-learning learning-dynamics

2