arXiv subjects

Tilo Wettig

Publications and source records attributed to Tilo Wettig.

At least 37 records · Page 2Linked to original sources

arXiv2020

Low-rank tensor methods for Markov chains with applications to tumor progression models

Continuous-time Markov chains describing interacting processes exhibit a state space that grows exponentially in the number of processes. This state-space explosion renders the computation or storage of the time-marginal distribution, which is defined as the solution of a certain linear system, infeasible using classical methods. We consider Markov chains whose transition rates are separable functions, which allows for an efficient low-rank tensor representation of the operator of this linear system. Typically, the right-hand side also has low-rank structure, and thus we can reduce the cost for computation and storage from exponential to linear. Previously known iterative methods also allow for low-rank approximations of the solution but are unable to guarantee that its entries sum up to one as required for a probability distribution. We derive a convergent iterative method using low-rank formats satisfying this condition. We also perform numerical experiments illustrating that the marginal distribution is well approximated with low rank.

math.NA↗

arXiv2020

Lattice QCD on a novel vector architecture

The SX-Aurora TSUBASA PCIe accelerator card is the newest model of NEC's SX architecture family. Its multi-core vector processor features a vector length of 16 kbits and interfaces with up to 48 GB of HBM2 memory in the current models, available since 2018. The compute performance is up to 2.45 TFlop/s peak in double precision, and the memory throughput is up to 1.2 TB/s peak. New models with improved performance characteristics are announced for the near future. In this contribution we discuss key aspects of the SX-Aurora and describe how we enabled the architecture in the Grid Lattice QCD framework.

cs.DC↗

arXiv2019

Multigrid for Wilson Clover Fermions in Grid

With the ever-growing number of computing architectures, performance portability is an important aspect of (Lattice QCD) software. The Grid library provides a good framework for writing such code, as it thoroughly separates hardware-specific code from algorithmic functionality and already supports many modern architectures. We describe the implementation of a multigrid solver for Wilson clover fermions in Grid by the RQCD group. We present the features included in our implementation, discuss initial optimization efforts, and compare the performance with another multigrid implementation.

hep-lat↗

arXiv2019

Lattice QCD on upcoming Arm architectures

Recently Arm introduced a new instruction set called Scalable Vector Extension (SVE), which supports vector lengths up to 2048 bits. While SVE hardware will not be generally available until about 2021, we believe that future SVE-based architectures will have great potential for Lattice QCD. In this contribution we discuss key aspects of SVE and describe how we implemented SVE in the Grid Lattice QCD framework.

hep-lat↗

arXiv2019

Induced QCD II: Numerical results

We numerically explore an alternative discretization of continuum $\text{SU}(N_c)$ Yang-Mills theory on a Euclidean spacetime lattice, originally introduced by Budzcies and Zirnbauer for gauge group $\text{U}(N_c)$. This discretization can be reformulated such that the self-interactions of the gauge field are induced by a path integral over $N_b$ auxiliary bosonic fields, which couple linearly to the gauge field. In the first paper of the series we have shown that the theory reproduces continuum $\text{SU}(N_c)$ Yang-Mills theory in $d=2$ dimensions if $N_b$ is larger than $N_c-\frac{3}{4}$ and conjectured, following the argument of Budzcies and Zirnbauer, that this remains true for $d>2$. In the present paper, we test this conjecture by performing lattice simulations of the simplest nontrivial case, i.e., gauge group $\text{SU}(2)$ in three dimensions. We show that observables computed in the induced theory, such as the static $q\bar q$ potential and the deconfinement transition temperature, agree with the same observables computed from the ordinary plaquette action up to lattice artifacts. We also find that the bound for $N_b$ can be relaxed to $N_c-\frac{5}{4}$ as conjectured in our earlier paper. Studies of how the new discretization can be used to change the order of integration in the path integral to arrive at dual formulations of QCD are left for future work.

hep-lat↗

arXiv2019

SVE-enabling Lattice QCD Codes

Optimization of applications for supercomputers of the highest performance class requires parallelization at multiple levels using different techniques. In this contribution we focus on parallelization of particle physics simulations through vector instructions. With the advent of the Scalable Vector Extension (SVE) ISA, future ARM-based processors are expected to provide a significant level of parallelism at this level.

cs.DC↗

arXiv2018

Loss-function learning for digital tissue deconvolution

The gene expression profile of a tissue averages the expression profiles of all cells in this tissue. Digital tissue deconvolution (DTD) addresses the following inverse problem: Given the expression profile $y$ of a tissue, what is the cellular composition $c$ of that tissue? If $X$ is a matrix whose columns are reference profiles of individual cell types, the composition $c$ can be computed by minimizing $\mathcal L(y-Xc)$ for a given loss function $\mathcal L$. Current methods use predefined all-purpose loss functions. They successfully quantify the dominating cells of a tissue, while often falling short in detecting small cell populations. Here we learn the loss function $\mathcal L$ along with the composition $c$. This allows us to adapt to application-specific requirements such as focusing on small cell populations or distinguishing phenotypically similar cell populations. Our method quantifies large cell fractions as accurately as existing methods and significantly improves the detection of small cell populations and the distinction of similar cell types.

q-bio.QM↗

arXiv2017

DD-$α$AMG on QPACE 3

We describe our experience porting the Regensburg implementation of the DD-$α$AMG solver from QPACE 2 to QPACE 3. We first review how the code was ported from the first generation Intel Xeon Phi processor (Knights Corner) to its successor (Knights Landing). We then describe the modifications in the communication library necessitated by the switch from InfiniBand to Omni-Path. Finally, we present the performance of the code on a single processor as well as the scaling on many nodes, where in both cases the speedup factor is close to the theoretical expectations.

hep-lat↗

arXiv2017

Chiral condensate and Dirac spectrum of one- and two-flavor QCD at nonzero $θ$-angle

In the $ε$-domain of QCD we have obtained exact analytical expressions for the eigenvalue density of the Dirac operator at fixed $θ\ne 0$ for both one and two flavors. These results made it possible to explain how the different contributions to the spectral density conspire to give a chiral condensate at fixed $θ$ that does not change sign when the quark mass (or one of the quark masses for two flavors) crosses the imaginary axis, while the chiral condensate at fixed topological charge does change sign. From QCD at nonzero density we have learnt that the discontinuity of the chiral condensate may move to a different location when the spectral density increases exponentially with the volume with oscillations on the order of the inverse volume. This is indeed what happens when the product of the quark masses becomes negative, but the situation is more subtle in this case: the contribution of the "quenched" part of the spectral density diverges in the thermodynamic limit at nonzero $θ$, but this divergence is canceled exactly by the contribution from the zero modes. We conclude that the zero modes are essential for the continuity of the chiral condensate and that their contribution has to be perfectly balanced against the contribution from the nonzero modes. Lattice simulations at nonzero $θ$-angle can only be trusted if this is indeed the case.

hep-lat↗

arXiv2017

Complete random matrix classification of SYK models with $\mathcal{N}=0$, $1$ and $2$ supersymmetry

We present a complete symmetry classification of the Sachdev-Ye-Kitaev (SYK) model with $\mathcal{N}=0$, $1$ and $2$ supersymmetry (SUSY) on the basis of the Altland-Zirnbauer scheme in random matrix theory (RMT). For $\mathcal{N}=0$ and $1$ we consider generic $q$-body interactions in the Hamiltonian and find RMT classes that were not present in earlier classifications of the same model with $q=4$. We numerically establish quantitative agreement between the distributions of the smallest energy levels in the $\mathcal{N}=1$ SYK model and RMT. Furthermore, we delineate the distinctive structure of the $\mathcal{N}=2$ SYK model and provide its complete symmetry classification based on RMT for all eigenspaces of the fermion number operator. We corroborate our classification by detailed numerical comparisons with RMT and thus establish the presence of quantum chaotic dynamics in the $\mathcal{N}=2$ SYK model. We also introduce a new SYK-like model without SUSY that exhibits hybrid properties of the $\mathcal{N}=1$ and $\mathcal{N}=2$ SYK models and uncover its rich structure both analytically and numerically.

hep-th↗

arXiv2017

Scale-invariant biomarker discovery in urine and plasma metabolite fingerprints

Motivation: Metabolomics data is typically scaled to a common reference like a constant volume of body fluid, a constant creatinine level, or a constant area under the spectrum. Such normalization of the data, however, may affect the selection of biomarkers and the biological interpretation of results in unforeseen ways. Results: First, we study how the outcome of hypothesis tests for differential metabolite concentration is affected by the choice of scale. Furthermore, we observe this interdependence also for different classification approaches. Second, to overcome this problem and establish a scale-invariant biomarker discovery algorithm, we extend linear zero-sum regression to the logistic regression framework and show in two applications to ${}^1$H NMR-based metabolomics data how this approach overcomes the scaling problem. Availability: Logistic zero-sum regression is available as an R package as well as a high-performance computing implementation that can be downloaded at https://github.com/rehbergT/zeroSum

q-bio.QM↗

arXiv2017

pMR: A high-performance communication library

On many parallel machines, the time LQCD applications spent in communication is a significant contribution to the total wall-clock time, especially in the strong-scaling limit. We present a novel high-performance communication library that can be used as a de facto drop-in replacement for MPI in existing software. Its lightweight nature that avoids some of the unnecessary overhead introduced by MPI allows us to improve the communication performance of applications without any algorithmic or complicated implementation changes. As a first real-world benchmark, we make use of the pMR library in the coarse-grid solve of the Regensburg implementation of the DD-$α$AMG algorithm. On realistic lattices, we see an improvement of a factor 2x in pure communication time and total execution time savings of up to 20%.

hep-lat↗

arXiv2016

Induced QCD I: Theory

We explore an alternative discretization of continuum SU(N_c) Yang-Mills theory on a Euclidean spacetime lattice, originally introduced by Budzcies and Zirnbauer. In this discretization the self-interactions of the gauge field are induced by a path integral over N_b auxiliary boson fields, which are coupled linearly to the gauge field. The main progress compared to earlier approaches is that N_b can be as small as N_c. In the present paper we (i) extend the proof that the continuum limit of the new discretization reproduces Yang-Mills theory in two dimensions from gauge group U(N_c) to SU(N_c), (ii) derive refined bounds on N_b for non-integer values, and (iii) perform a perturbative calculation to match the bare parameter of the induced gauge theory to the standard lattice coupling. In follow-up papers we will present numerical evidence in support of the conjecture that the induced gauge theory reproduces Yang-Mills theory also in three and four dimensions, and explore the possibility to integrate out the gauge fields to arrive at a dual formulation of lattice QCD.

hep-lat↗

arXiv2016

Multiple right-hand-side setup for the DD-αAMG

The setup cost of a modern solver such as DD-αAMG (Wuppertal Multigrid) is a significant contribution to the total time spent on solving the Dirac equation, and in HMC it can even be dominant. We present an improved implementation of this algorithm with modified computation order in the setup procedure. By processing multiple right-hand sides simultaneously we can alleviate many of the performance issues of the default single right-hand-side setup. The main improvements are as follows: By combining multiple right-hand sides the message size for off-chip communication is larger, which leads to better utilization of the network bandwidth. Many matrix-vector products are replaced by matrix-matrix products, leading to better cache reuse. The synchronization overhead inflicted by on-chip parallelization (threading), which is becoming crucial on many-core architectures such as the Intel Xeon Phi, is effectively reduced. In the parts implemented so far, we observe a speedup of roughly 3x compared to the optimized version of the single right-hand-side setup on realistic lattices.

hep-lat↗

arXiv2015

Adaptive algebraic multigrid on SIMD architectures

We present details of our implementation of the Wuppertal adaptive algebraic multigrid code DD-$α$AMG on SIMD architectures, with particular emphasis on the Intel Xeon Phi processor (KNC) used in QPACE 2. As a smoother, the algorithm uses a domain-decomposition-based solver code previously developed for the KNC in Regensburg. We optimized the remaining parts of the multigrid code and conclude that it is a very good target for SIMD architectures. Some of the remaining bottlenecks can be eliminated by vectorizing over multiple test vectors in the setup, which is discussed in the contribution of Daniel Richtmann.

physics.comp-ph↗

arXiv2015

Induced YM theory with auxiliary bosons

We study pure SU(N) lattice gauge theory with a plaquette weight factor given by an inverse determinant which can be written as an integral over auxiliary bosonic fields (modifying a proposal of Budczies and Zirnbauer). We derive conditions for the existence of a continuum limit and its equivalence to Yang-Mills theory. Furthermore, we perturbatively compute the relation between the coupling constants of the `induced' gauge action and the familiar Wilson gauge action using the background-field technique. The perturbative relation agrees well with numerical results for N=2 in three dimensions.

hep-lat↗

arXiv2015

QPACE 2 and Domain Decomposition on the Intel Xeon Phi

We give an overview of QPACE 2, which is a custom-designed supercomputer based on Intel Xeon Phi processors, developed in a collaboration of Regensburg University and Eurotech. We give some general recommendations for how to write high-performance code for the Xeon Phi and then discuss our implementation of a domain-decomposition-based solver and present a number of benchmarks.

cs.DC↗

arXiv2014

The Chiral Condensate of One-Flavor QCD and the Dirac Spectrum at θ=0

In a sector of fixed topological charge, the chiral condensate has a discontinuity given by the Banks-Casher formula also in the case of one-flavor QCD. However, at fixed θ-angle, the chiral condensate remains constant when the quark mass crosses zero. To reconcile these contradictory observations, we have evaluated the spectral density of one-flavor QCD at θ=0. For negative quark mass, it becomes a strongly oscillating function with a period that scales as the inverse space-time volume and an amplitude that increases exponentially with the space-time volume. As we have learned from QCD at nonzero chemical potential, if this is the case, an alternative to the Banks-Casher formula applies, and as we will demonstrate in this talk, for one-flavor QCD this results in a continuous chiral condensate. A special role is played by the topological zero modes which have to be taken into account exactly in order to get a finite chiral condensate in the thermodynamic limit.

hep-lat↗