arXiv research

New method controls bias in unadjusted Hamiltonian Monte Carlo and underdamped Langevin.

Improved sampling from complex distributions with reduced bias.

The unadjusted Langevin algorithm converges faster for some variables in high dimensions.

We define extensions of the $L^2$-analytic invariants of closed manifolds, called delocalized $L^2$-invariants. These delocalized invariants are constructed in terms of a nontrivial conjugacy class of the fundamental group. We show that in many cases, they are topological in nature. We show that the marked length spect…

In this paper we state and prove Morse type inequalities for Morse functions as well as for closed differential 1-forms. These inequalities involve delocalized Betti numbers. As an immediate consequence, we prove the vanishing of delocalized Betti numbers of manifolds fibering over the circle.

Study delocalized eta invariants for signature operators on proper manifolds.

The paper defines higher invariants for groups of polynomial growth and proves their convergence.

Study approximates delocalized eta invariants of covering spaces.

A new model explains protein interactions via electron delocalization.

The paper proves universality in optimization problems with i.i.d. random vectors.

Non-Abelian actions are resolved using equivariant K-theory and delocalized cohomology.

For any closed complex manifold $X$, we calculate the Poincaré and Hodge polynomials of the delocalized equivariant cohomology $H^*(X^n, S_n)$ with a grading specified by physicists. As a consequence, we recover a special case of a formula for the elliptic genera of symmetric products in Dijkgraaf-Moore-Verlinde-Verlin…

Let G be a finitely generated discrete group. In this paper we establish vanishing results for rho-invariants associated to (i) the spin-Dirac operator of a spin manifold with positive scalar curvature (ii) the signature operator of the disjoint union of a pair of homotopy equivalent oriented manifolds with fundamental…

New method detects global factors near BBP phase transition in high-dimensional data.

A refined form of the `Folk Theorem' that a smooth action by a compact Lie group can be (canonically) resolved, by iterated blow up, to have unique isotropy type was established by the authors in the context of manifolds with corners; the canonical construction induces fibrations on the boundary faces of the resolution…

We give a description of the delocalized twisted cohomology of an orbifold and the Chern character of a twisted vector bundle in terms of supersymmetric Euclidean field theories. This includes the construction of a twist functor for $1|1$-dimensional EFTs from the data of a gerbe with connection.

Defines rho numbers for metrics with positive scalar curvature.

A multi-scale model predicts atomic-scale properties using both local and long-range information.

Let G be a discrete group, and let M be a closed spin manifold of dimension m>3 with pi_1(M)=G. We assume that M admits a Riemannian metric of positive scalar curvature. We discuss how to use the L2-rho invariant and the delocalized eta invariant associated to the Dirac operator on M in order to get information about t…

Let $Γ$ be a discrete group. Assuming rational injectivity of the Baum-Connes assembly map, we provide new lower bounds on the rank of the positive scalar curvature bordism group and the relative group in Stolz' positive scalar curvature sequence for $\mathrm{B} Γ$. The lower bounds are formulated in terms of the part …

The smooth action of a compact Lie group on a compact manifold can be resolved to an iterated space, as made explicit by Pierre Albin and the second author. On the resolution the lifted action has fixed isotropy type, in an iterated sense, with connecting fibrations and this structure descends to a resolution of the qu…

We show that the emergence of systemic risk in complex systems can be understood from the evolution of functional networks representing interactions inferred from fluctuation correlations between macroscopic observables. Specifically, we analyze the long-term collective dynamics of the New York Stock Exchange between 1…

Let $Γ$ be a f.g. discrete group and let $\tilde M$ be a Galois $Γ$-covering of a smooth closed manifold $M$. Let $S_*^Γ(\tilde{M})$ be the analytic structure group, appearing in the Higson-Roe analytic surgery sequence $\to S_*^Γ(\tilde M)\to K_*(M)\to K_*(C_r^*Γ)\to$. We prove that for an arbitrary discrete group $Γ$…

Geometric formula derived for Lefschetz pairing on Γ-proper manifolds.

The `Folk Theorem' that a smooth action by a compact Lie group can be (canonically) resolved, by iterated blow up, to have unique isotropy type is proved in the context of manifolds with corners. This procedure is shown to capture the simultaneous resolution of all isotropy types in a `resolution tower' which projects …

We analyze a class of conical G_2 metrics admitting two commuting isometries, together with a certain one-parameter family of G_2 deformations which preserves these symmetries. Upon using recent results of Calderbank and Pedersen, we write down the explicit G_2 metric for the most general member of this family and extr…

We develop methods to study the singularities of certain $G_2$ cones related to toric hyperkahler spaces and Einstein selfdual orbifolds. This allows us to determine the low energy gauge groups of chiral N=1 compactifications of M-theory on a large family of such backgrounds, which includes the models recently studied …

We consider a generalized APS boundary problem for a G-invariant Dirac-type operator, which is not of product type near the boundary. We establish a delocalized version (a so-called Kirillov formula) of the equivariant index theorem for this operator. We obtain more explicit formulas for different geometric Dirac-type …

Research shows bias in machine learning can be due to algorithmic flaws, not just data.

The paper analyzes high-dimensional linear regression using parametric empirical Bayes methods.

Uniform K-homology theory applied to elliptic operators on manifolds with boundary.

Financial LLMs need explicit bias consideration to avoid invalid results.

We study the phenomenon of bias amplification in classifiers, wherein a machine learning model learns to predict classes with a greater disparity than the underlying ground truth. We demonstrate that bias amplification can arise via an inductive bias in gradient descent methods that results in the overestimation of the…

It has been noticed that some external CVIs exhibit a preferential bias towards a larger or smaller number of clusters which is monotonic (directly or inversely) in the number of clusters in candidate partitions. This type of bias is caused by the functional form of the CVI model. For example, the popular Rand index (R…

Mitigates gender bias amplification in model predictions.

Depth uncertainty networks don't improve with bias correction, contrary to expectations.

We quantify causal bias in continuous treatment settings.

The paper introduces Relative Bias to quantify LLM bias systematically.

Interpolated-MLPs control inductive bias for better performance in low-compute tasks.

SSMs have a built-in bias towards low-frequency components, which can be adjusted.

This paper assesses biases in contextualized word representations.

A bias classifier is introduced to resist adversarial attacks.

UBM transfers bias mitigation from upstream to downstream tasks efficiently.

Active learning introduces bias; this paper fixes it.

Study shows racial bias in health data, which can be reduced with simple techniques.

Bias correction improves language model training performance.

Ensembles improve classifier performance by reducing bias, not variance.

Study finds gender bias in human evaluators and shows how machine learning can mitigate it.