Search by speaker

Filter institution








Filter content






Wed, 16. Oct at 10:00
HVP 11 a, R.313
Privacy constrained semiparametric inference
Abstract. For semi-parametric problems differential private estimators are typically constructed in a case-by-case basis. In this work we develop a privacy constrained semi-parametric plug-in approach, which can be used in general, over a collection of semi-parametric problems. We derive minimax lower and matching upper bounds for this approach and provide an adaptive procedure in case of irregular (atomic) functionals. Joint work with Lukas Steinberger (Vienna) and Thibault Randrianarisoa (Toronto, Vector Institute).
Wed, 16. Oct at 14:00
fubma001
Learning to Compute Gröbner bases
Abstract. Solving a polynomial system, or computing an associated \gb basis, has been a fundamental task in computational algebra. However, it is also known for its notorious doubly exponential time complexity in the number of variables in the worst case. In this talk, I present a new paradigm for addressing such problems, i.e., a machine-learning approach using a Transformer. The learning approach does not require an explicit algorithm design and can return the solutions in (roughly) constant time. This talk covers our initial results on this approach and relevant computational algebraic and machine learning challenges.
Wed, 16. Oct at 14:15
WIAS, Erhard-Schm...
Formation of microstructure for singularly perturbed problems related to helimagnets and shape-memory alloys
Abstract
Wed, 16. Oct at 16:30
EN 058
Tropical Gradient Descent
Abstract. The field of tropical statistics - motivated by the identification of the tropical Grassmannian and the space of phylogenetic trees - has produced a range of unconstrained optimisation problems over the tropical projective torus. We will review the types of convexity exhibited by tropical loss functions in statistics, and we propose a new gradient descent method for solving tropical optimisation problems. Theoretical results establish global solvability for tropically star-quasi-convex problems, and numerical experiments demonstrate the method's superior performance over classical descent for tropical optimisation problems which exhibit tropical quasi-convexity but not classical convexity. Notably, tropical gradient descent seamlessly integrates into advanced optimisation methods, such as Adam, offering improved overall performance.
Thu, 17. Oct at 13:00
On the number of arrangements of pseudolines: upper and lower bounds
Thu, 17. Oct at 16:15
Arnimallee 3
Families of Kirchhoff Graphs
Abstract
Mon, 21. Oct at 13:00
Rudower Chaussee ...
Reversible saddle-node separatrix-loop bifurcation
Mon, 21. Oct at 14:15
T9/046
1-independent percolation on high dimensional lattices
Abstract
Mon, 21. Oct at 15:00
Rudower Chaussee ...
Multilevel methods for non-smooth minimization and variational inequalities
Tue, 22. Oct at 14:00
WIAS HVP5-7 R411 ...
A semismooth Newton method for obstacle-type quasivariational inequalities
Abstract. Quasivariational inequalities (QVIs) are ubiquitous but, in particular, arise in PDE-constrained optimization in cases where the constraint set depends on the solution itself. In obstacle-type QVIs, this manifests as an obstacle that bends according to the state of the system. QVIs are notoriously hard to analyze, especially in the infinite-dimensional setting, and developing fast solvers posed in infinite dimensions has proven particularly challenging. As such most solvers in the literature rely on fixed point algorithms which can be slow to converge. In this talk, we introduce the first semismooth Newton method, posed in a Banach space setting, for such problems. We will see that the solver enjoys favourable properties such as local superlinear convergence and mesh independence.
Wed, 23. Oct at 10:00
HVP 11 a, R.313
Conditional nonparametric variable screening by neural factor regression
Abstract. High-dimensional covariates often admit linear factor structure. To effectively screen correlated covariates in high-dimension, we propose a conditional variable screening test based on non-parametric regression using neural networks due to their representation power. We ask the question whether individual covariates have additional contributions given the latent factors or more generally a set of variables. Our test statistics are based on the estimated partial derivative of the regression function of the candidate variable for screening and a observable proxy for the latent factors. Hence, our test reveals how much predictors contribute additionally to the non-parametric regression after accounting for the latent factors. Our derivative estimator is the convolution of a deep neural network regression estimator and a smoothing kernel. We demonstrate that when the neural network size diverges with the sample size, unlike estimating the regression function itself, it is necessary to smooth the partial derivative of the neural network estimator to recover the desired convergence rate for the derivative. Moreover, our screening test achieves asymptotic normality under the null after finely centering our test statistics that makes the biases negligible, as well as consistency for local alternatives under mild conditions. We demonstrate the performance of our test in a simulation study and two real world applications.
Wed, 23. Oct at 11:30
online
An Inexact Generalized Conditional Gradient Method
Abstract
Wed, 23. Oct at 16:00
Wed, 23. Oct at 16:30
EN 058
Generalized chain polytopes, Grassmannians and representations
Abstract. For a finite poset P we introduce a two-parameter family X(m,M) of polytopes defined by the set of inequalities labeled by subposets of P. The polytopes enjoy Minkowski sum property and are important for geometric and algebraic applications. We will discuss the connection of X(m,M) with the geometry of the classical Grassmann varieties. We will also construct a family of representations of the degenerate sl(n) Lie algebra admitting monomial bases labeled by integer points of X(m,M) (for the Grassmann poset P). The talk is based on https://arxiv.org/abs/2403.10074 (joint with Wojciech Samotij).
Thu, 24. Oct at 13:00
Thu, 24. Oct at 14:00
Zuse Institute Be...
Quantum Algorithms for Optimization
Abstract. Faster algorithms for optimization problems are among the main potential applications for future quantum computers. There has been interesting progress in this area in recent years, for instance improved quantum algorithms for gradient descent and for solving linear and semidefinite programs. In this talk I will survey what we know about quantum speed-ups both for discrete and for continuous optimization, with a bit more detail about two speed-ups I worked on recently: for regularized linear regression and for Principal Component Analysis. I'll also discuss some issues with this line of work, in particular that quadratic or subquadratic quantum speed-ups will only kick in for very large instance sizes and that many of these algorithms require some kind of quantum RAM.
Fri, 25. Oct at 16:00
EN 058
The Two Lives of the Grassmannian
Abstract. The Grassmannian parametrizes linear subspaces of a real vector space. It is both a projective variety (via Plücker coordinates) and an affine variety (via orthogonal projections). We examine these two representations, through the lenses of linear algebra, commutative algebra, and statistics.
Tue, 29. Oct at 11:15
1.023 (BMS Room, ...
On the early history of quantum gravity
Abstract. Quantum gravity, in the sense of a formal quantization of general relativity, had a first beginning in the year 1930. That year, the Belgian physicist Léon Rosenfeld published a seminal paper called "Zur Quantelung der Wellenfelder" in which he developed Heisenberg and Pauli's recently constructed method to quantize the electromagnetic field in order to apply it to the tetrad formulation of general relativity. In my talk, I aim to shed light on a perhaps surprising crucial historical influence that made this piece of intellectual work possible: that of unified field theory. A purely classical program, most prominently pursued by Albert Einstein and Hermann Weyl, to formally reduce the (classical) electromagnetic field to the gravitational field as described by general relativity.
Wed, 30. Oct at 10:00
HVP 11 a, R.313
Wed, 30. Oct at 14:15
WIAS, Erhard-Schm...
Time discretization in visco-elastodynamics at large displacements and strains in the Eulerian frame
Abstract
Wed, 30. Oct at 16:30
EN 058
Thu, 31. Oct at 14:00
Fri, 01. Nov at 14:15
TU (C130)
Topology Data Analysis for Multiscale Biology
Tue, 05. Nov at 11:15
1.023 (BMS Room, ...
1D Landau-Ginzburg superpotential of big quantum cohomology of CP2
Abstract. Using the inverse period map of the Gauss-Manin connection associated with QH∗(CP2) and the Dubrovin construction of Landau-Ginzburg superpotential for Dubrovin Frobenius manifolds, we construct a one-dimensional Landau-Ginzburg superpotential for the quantum cohomology of CP2. In the case of small quantum cohomology, the Landau-Ginzburg superpotential is expressed in terms of the cubic root of the j-invariant function. For big quantum cohomology, the one-dimensional Landau-Ginzburg superpotential is given by Taylor series expansions whose coefficients are expressed in terms of quasi-modular forms. Furthermore, we express the Landau-Ginzburg superpotential for both small and big quantum cohomology of QH∗(CP2) in closed form as the composition of the Weierstrass ℘-function and the universal coverings of C \ (Z ⊕ jZ) and C \ (Z ⊕ zZ) respectively. This seminar is based on the results of arXiv/2402.09574.
Wed, 06. Nov at 11:30
online
Computing Multiple Solutions of Topology Optimization Problems
Wed, 06. Nov at 16:00
Wed, 06. Nov at 16:30
EN 058
Thu, 07. Nov at 14:00
Thu, 07. Nov at 16:15
TU Berlin, Instit...
Thu, 07. Nov at 17:15
TU Berlin, Instit...
Tue, 12. Nov at 11:15
1.023 (BMS Room, ...
Universality classes in non-Hermitian random matrices
Wed, 13. Nov at 14:15
WIAS, Erhard-Schm...
A scaling law for a model of epitaxial growth with dislocations
Abstract
Thu, 14. Nov at 14:00
Fri, 15. Nov at 14:15
HU (ESZ, 0'115 an...
A Day of Arithmetic Geometry (on the occasion of the retirement of Jürg Kramer)
Tue, 19. Nov at 11:00
Certified Reduced-Order Methods for Model Predictive Control of Time-Varying Evolution Processes
Abstract. In this talk model predictive control (MPC) is utilized to stabilize a class of linear time-varying parabolic partial differential equations (PDEs). In our first example the control input is only finite-dimensional, i.e., it enters as a time-depending linear combination of finitely many indicator functions whose total supports cover only a small part of the spatial domain. In the second example the PDE involve switching coefficient functions. We discuss stabilizability and the application of reduced-order models to derive algorithms with closed-loop guarantees. <p>This is joint work with , and Benjamin Unger (Stuttgart).</p>
Tue, 19. Nov at 11:15
1.023 (BMS Room, ...
Uniqueness of Malliavin-Kontsevich-Suhov measures
Abstract. About 20 years ago, Kontsevich & Suhov conjectured the existence and uniqueness of a family of measures on the set of Jordan curves, characterised by conformal invariance and another property called 'conformal restriction'. This conjecture was motivated by (seemingly unrelated) works of Schramm, Lawler & Werner on stochastic Loewner evolutions (SLE), and Malliavin, Airault & Thalmaier on 'unitarising measures'. The existence of this family was settled by works of Werner-Kemppainen and Zhan, using a loop version of SLE. The uniqueness was recently obtained in a joint work with Jego. I will start by reviewing the different notions involved before giving some ideas of our proof of uniqueness: in a nutshell, we construct a family of 'orthogonal polynomials' which completely characterise the measure. In the remaining time, I will discuss the broader context in which our construction fits, namely the conformal field theory associated with SLE.
Wed, 20. Nov at 10:00
HVP 11 a, R.313
Wed, 20. Nov at 11:30
online
Of Rigid and Flexible Polytopes
Wed, 20. Nov at 16:00
Thu, 21. Nov at 14:00
Thu, 21. Nov at 16:15
TU Berlin, Instit...
Wed, 27. Nov at 10:00
HVP 11 a, R.313
Wed, 04. Dec at 11:30
online
Calabi-Yau 3-Folds with Boundary
Wed, 04. Dec at 16:00
Wed, 04. Dec at 16:00
Thu, 05. Dec at 17:15
TU Berlin, Instit...
Wed, 11. Dec at 14:15
WIAS, Erhard-Schm...
Thu, 12. Dec at 14:00
Thu, 12. Dec at 14:00
BEL 301
Thesis defense
Abstract. Thesis defense Dante Luber
Fri, 13. Dec at 14:15
FU
Wed, 18. Dec at 11:30
online
Computational Aspects of Quadratic Forms in Determining the Representation Type of Quiver Algebras
Wed, 18. Dec at 16:00
Thu, 09. Jan at 14:00
Fri, 10. Jan at 14:15
FU
Wed, 15. Jan at 11:30
online
Data Transmission in Dynamical Random Networks
Abstract
Tue, 21. Jan at 11:15
1.023 (BMS Room, ...
Wed, 29. Jan at 11:30
online
Coherent Transport of Semiconductor Spin-Qubits: Modeling, Simulation and Optimal Control
Abstract
Thu, 30. Jan at 17:15
TU Berlin, Instit...
Wed, 12. Feb at 11:30
online
Hybrid Models for Large Scale Infection Spread Simulations
Abstract