diffusion, entropy, and the geometry of small sets James R. Lee University of Washington Joint work with Ronen Eldan
a prelude
noise and smoothness
noise and smoothness Many applications: PCPs & hardness of approximation, statistical physics, threshold phenomena, social choice, circuit complexity,
noise and smoothness Hypercontractive inequality [Bonami, Gross, Nelson]:
noise and smoothness Hypercontractive inequality [Bonami, Gross, Nelson]:
noise and smoothness Hypercontractive inequality [Bonami, Gross, Nelson]:
the convolution conjecture
anti-concentration of temperature Convolution conjecture [Talagrand 1989]: Equivalent to the conjecture that “heat” cannot concentrate near a single (high) “temperature”
the Gaussian case
the Gaussian case
the Gaussian case Ornstein-Uhlenbeck semi-group:
the Gaussian case Theorem [Eldan-L 2014]:
the Gaussian case Theorem [Eldan-L 2014]:
the Gaussian case Theorem [Eldan-L 2014]:
some difficulties Good: noise insensitive Bad: boundary far from origin half space dust
proof ideas arguing about small-probability events is tough, but remember the prelude… random measure
conditioning
Explicit form: Theorem [Lehec 2010]:
an optimal coupling is a martingale
proof sketch
proof sketch
Girsanov’s theorem Then under the change of measure: