Minimizing movements with general costs,
and nonnegative cross-curvature in
infinite dimensions
Flavien Léger
joint works with Pierre-Cyril Aubin-Frankowski,
Gabriele Todeschi, François-Xavier Vialard

1. Minimizing movements with general cost
2. NNCC spaces
Implicit minimizing movements
X,Y: two sets, c:X×Y→R∪{+∞}
g:X→R∪{+∞}: energy to minimize
I M P L I C I T S C H E M E
Goal: geometry in which convexity of g gives convergence rates
Explicit minimizing movements
f:X→R∪{+∞}: c-concave energy
E X P L I C I T S C H E M E

(majorize)
(minimize)
Explicit minimizing movements
X,Y smooth manifolds, c∈C1(X×Y), f∈C1(X) c-concave
Under assumptions (twist, non-empty c-subdifferentials...), the explicit scheme can be written as
EVI for Alternating Minimization
Alternating minimization (AM) of ϕ:X×Y→R∪{+∞}
(Csiszár–Tusnády ’84)
(L–Aubin-Frankowski ’23)
EVI (five-point property) for AM:
Implicit: ϕ(x,y)=g(x)+c(x,y)
Explicit: ϕ(x,y)=c(x,y)+fc(y)
D E F I N I T I O N
Convergence rates
If (xn,yn) satisfy the EVI then
sublinear rates
when μ=0
exponential rates
when μ>0
(L–Aubin-Frankowski ’23)
EVI from convexity
Suppose that for every y0∈Y there exists x0∈argminx∈Xϕ(x,y0), y1∈argminy∈Yϕ(x0,y), such that for each x∈X,
T H E O R E M (L–Aubin-Frankowski '24)
Then EVI
⏵ there exists a variational c-segment s↦(x(s),y0) on (X×Y,ϕ) with x(0)=x0 and x(1)=x
⏵ s↦F(x(s))−μD(x(s),y1) is convex
⏵ s→0+limsD(x(s),y1)=0
1. Minimizing movements with general cost
2. NNCC spaces
Introduction
X,Y n-dimensional manifolds, c∈C4(X×Y), ∇xy2c(x,y) nonsingular
MTW condition
Nonnegative cross-curvature
(Ma–Trudinger–Wang ’05)
(Trudinger–Wang ’09)
(Kim–McCann ’10)
s↦(x(s),yˉ) is a c-segment if ∇yc(x(s),yˉ)=(1−s)∇yc(x(0),yˉ)+s∇yc(x(1),yˉ).
D E F I N I T I O N
T H E O R E M (KIM–MCCANN '10)
Under assumptions (c-convex domains...) c has nonnegative cross-curvature iff for any c-segment s↦(x(s),yˉ),
Variational c-segments and NNCC spaces
D E F I N I T I O N (L–Todeschi–Vialard '24)
s↦(x(s),yˉ) is a variational c-segment if c(x(s),yˉ) is finite and
(X×Y,c) is a space with nonnegative cross-curvature (NNCC space) if variational c-segments always exist.
X,Y two arbitrary sets, c:X×Y→R∪{±∞}.
Properties of NNCC spaces
Stable by products
Stable by projections with “equidistant fibers”
Stable under Gromov–Hausdorff convergence
Metric cost c(x,y)=d2(x,y) NNCC⟹PC
NNCC ⟺
the set of c-concave function with nonempty c-subdifferentials is convex
(connect to: Ambrosio–Gigli–Savaré ’05)
(connect to: Kim–McCann '12)
(connect to: Figalli–Kim–McCann '11)
(connect to: Loeper ’09)
Examples
Gromov–Wasserstein
Kullback–Leibler divergence, Hellinger, Fisher–Rao costs are NNCC
Transport costs
(G×G,GW2) is NCCC
(P(X)×P(Y),Tc) NNCC ⟺ (X×Y,c) NNCC
(Polish spaces, lsc cost)
Ex: W22 on Rn, on Sn...
X=[X,f,μ] and Y=[Y,g,ν]∈G
(L–Todeschi–Vialard '24)
Rem: strengthens PC
Variational c-segments ≈ generalized geodesics
(Sturm)
Conclusion and perspectives
⏵ Variational c-segments: nonsmooth geometry for EVIs
⏵ Variational c-segments: applications to functional inequalities
⏵ NNCC spaces: applications to mechanism design
Thank you!
(Pisa 2024-12-03) NNCC
By Flavien Léger
(Pisa 2024-12-03) NNCC
- 111