arXiv subjects

Matthias Bauer

Publications and source records attributed to Matthias Bauer.

28 records · Page 2Linked to original sources

arXiv2019

3-D Atomic Mapping of Interfacial Roughness and its Spatial Correlation Length in sub-10 nm Superlattices

The interfacial abruptness and uniformity in heterostructures are critical to control their electronic and optical properties. With this perspective, this work demonstrates the 3-D atomistic-level mapping of the roughness and uniformity of buried epitaxial interfaces in Si/SiGe superlattices with a layer thickness in the 1.5-7.5 nm range. Herein, 3-D atom-by-atom maps were acquired and processed to generate iso-concentration surfaces highlighting local fluctuations in content at each interface. These generated surfaces were subsequently utilized to map the interfacial roughness and its spatial correlation length. The analysis revealed that the root mean squared roughness of the buried interfaces in the investigated superlattices is sensitive to the growth temperature with a value varying from about 0.2 nm (+/- 13%) to about 0.3 nm (+/- 11.5%) in the temperature range of 500-650 Celsius. The estimated horizontal correlation lengths were found to be 8.1 nm (+/- 5.8%) at 650 Celsius and 10.1 nm (+/- 6.2%) at 500 Celsius. Additionally, reducing the growth temperature was found to improve the interfacial abruptness, with 30 % smaller interfacial width is obtained at 500 Celsius. This behavior is attributed to the thermally activated atomic exchange at the surface during the heteroepitaxy. Finally, by testing different optical models with increasing levels of interfacial complexity, it is demonstrated that the observed atomic-level roughening at the interface must be accounted for to accurately describe the optical response of Si/SiGe heterostructures.

cond-mat.mtrl-sci↗

arXiv2019

Resampled Priors for Variational Autoencoders

We propose Learned Accept/Reject Sampling (LARS), a method for constructing richer priors using rejection sampling with a learned acceptance function. This work is motivated by recent analyses of the VAE objective, which pointed out that commonly used simple priors can lead to underfitting. As the distribution induced by LARS involves an intractable normalizing constant, we show how to estimate it and its gradients efficiently. We demonstrate that LARS priors improve VAE performance on several standard datasets both when they are learned jointly with the rest of the model and when they are fitted to a pretrained model. Finally, we show that LARS can be combined with existing methods for defining flexible priors for an additional boost in performance.

stat.ML↗

arXiv2018

Learning Invariances using the Marginal Likelihood

Generalising well in supervised learning tasks relies on correctly extrapolating the training data to a large region of the input space. One way to achieve this is to constrain the predictions to be invariant to transformations on the input that are known to be irrelevant (e.g. translation). Commonly, this is done through data augmentation, where the training set is enlarged by applying hand-crafted transformations to the inputs. We argue that invariances should instead be incorporated in the model structure, and learned using the marginal likelihood, which correctly rewards the reduced complexity of invariant models. We demonstrate this for Gaussian process models, due to the ease with which their marginal likelihood can be estimated. Our main contribution is a variational inference scheme for Gaussian processes containing invariances described by a sampling procedure. We learn the sampling procedure by back-propagating through it to maximise the marginal likelihood.

cs.LG↗

arXiv2018

Automatic Estimation of Modulation Transfer Functions

The modulation transfer function (MTF) is widely used to characterise the performance of optical systems. Measuring it is costly and it is thus rarely available for a given lens specimen. Instead, MTFs based on simulations or, at best, MTFs measured on other specimens of the same lens are used. Fortunately, images recorded through an optical system contain ample information about its MTF, only that it is confounded with the statistics of the images. This work presents a method to estimate the MTF of camera lens systems directly from photographs, without the need for expensive equipment. We use a custom grid display to accurately measure the point response of lenses to acquire ground truth training data. We then use the same lenses to record natural images and employ a data-driven supervised learning approach using a convolutional neural network to estimate the MTF on small image patches, aggregating the information into MTF charts over the entire field of view. It generalises to unseen lenses and can be applied for single photographs, with the performance improving if multiple photographs are available.

cs.CV↗

arXiv2017

Discriminative k-shot learning using probabilistic models

This paper introduces a probabilistic framework for k-shot image classification. The goal is to generalise from an initial large-scale classification task to a separate task comprising new classes and small numbers of examples. The new approach not only leverages the feature-based representation learned by a neural network from the initial task (representational transfer), but also information about the classes (concept transfer). The concept information is encapsulated in a probabilistic model for the final layer weights of the neural network which acts as a prior for probabilistic k-shot learning. We show that even a simple probabilistic model achieves state-of-the-art on a standard k-shot learning dataset by a large margin. Moreover, it is able to accurately model uncertainty, leading to well calibrated classifiers, and is easily extensible and flexible, unlike many recent approaches to k-shot learning.

stat.ML↗

arXiv2017

Ecological feedback in quorum-sensing microbial populations can induce heterogeneous production of autoinducers

Autoinducers are small signaling molecules that mediate intercellular communication in microbial populations and trigger coordinated gene expression via "quorum sensing". Elucidating the mechanisms that control autoinducer production is, thus, pertinent to understanding collective microbial behavior, such as virulence and bioluminescence. Recent experiments have shown a heterogeneous promoter activity of autoinducer synthase genes, suggesting that some of the isogenic cells in a population might produce autoinducers, whereas others might not. However, the mechanism underlying this phenotypic heterogeneity in quorum-sensing microbial populations has remained elusive. In our theoretical model, cells synthesize and secrete autoinducers into the environment, up-regulate their production in this self-shaped environment, and non-producers replicate faster than producers. We show that the coupling between ecological and population dynamics through quorum sensing can induce phenotypic heterogeneity in microbial populations, suggesting an alternative mechanism to stochastic gene expression in bistable gene regulatory circuits.

q-bio.PE↗

arXiv2017

Understanding Probabilistic Sparse Gaussian Process Approximations

Good sparse approximations are essential for practical inference in Gaussian Processes as the computational cost of exact methods is prohibitive for large datasets. The Fully Independent Training Conditional (FITC) and the Variational Free Energy (VFE) approximations are two recent popular methods. Despite superficial similarities, these approximations have surprisingly different theoretical properties and behave differently in practice. We thoroughly investigate the two methods for regression both analytically and through illustrative examples, and draw conclusions to guide practical application.

stat.ML↗

arXiv2006

Emerging Markets for RFID Traces

RFID tags are held to become ubiquitous in logistics in the near future, and item-level tagging will pave the way for Ubiquitous Computing, for example in application fields like smart homes. Our paper addresses the value and the production cost of information that can be gathered by observing these tags over time and different locations. We argue that RFID technology will induce a thriving market for such information, resulting in easy data access for analysts to infer business intelligence and individual profiles of unusually high detail. Understanding these information markets is important for many reasons: They represent new business opportunities, and market players need to be aware of their roles in these markets. Policy makers need to confirm that the market structure will not negatively affect overall welfare. Finally, though we are not addressing the complex issue of privacy, we are convinced that market forces will have a significant impact on the effectiveness of deployed security enhancements to RFID technology. In this paper we take a few first steps into a relatively new field of economic research and conclude with a list of research problems that promise deeper insights into the matter.

cs.CY↗

arXiv2004

Proofs of Zero Knowledge

We present a protocol for verification of ``no such entry'' replies from databases. We introduce a new cryptographic primitive as the underlying structure, the keyed hash tree, which is an extension of Merkle's hash tree. We compare our scheme to Buldas et al.'s Undeniable Attesters and Micali et al.'s Zero Knowledge Sets.

cs.CR↗

arXiv2004

New Covert Channels in HTTP

This paper presents new methods enabling anonymous communication on the Internet. We describe a new protocol that allows us to create an anonymous overlay network by exploiting the web browsing activities of regular users. We show that the overlay network provides an anonymity set greater than the set of senders and receivers in a realistic threat model. In particular, the protocol provides unobservability in our threat model.

cs.CR↗