Conspiracies between Learning Algorithms, Circuit Lower Bounds and Pseudorandomness

2016/11/03 by Oliveira, Igor C., Santhanam, Rahul
#Computational Complexity (cs.CC) #Cryptography and Security (cs.CR) #Data Structures and Algorithms (cs.DS) #FOS: Computer and information sciences #Machine Learning (cs.LG)

paper · doi:10.48550/arxiv.1611.01190

Abstract

We prove several results giving new and stronger connections between learning, circuit lower bounds and pseudorandomness. Among other results, we show a generic learning speedup lemma, equivalences between various learning models in the exponential time and subexponential time regimes, a dichotomy between learning and pseudorandomness, consequences of non-trivial learning for circuit lower bounds, Karp-Lipton theorems for probabilistic exponential time, and NC¹-hardness for the Minimum Circuit Size Problem.

Related

• Learning algorithms from circuit lower bounds
• Quantum learning algorithms imply circuit lower bounds
• Algorithms versus Circuit Lower Bounds
• From Local Pseudorandom Generators to Hardness of Learning
• A Circuit Complexity Formulation of Algorithmic Information Theory