Andreas Fischer

Marco Peer, Thomas Gorges, Mathias Seuret et al.

Popular initiatives and referendums are central to Swiss democracy, yet the validation of handwritten signature lists remains a labor-intensive manual process. This paper investigates the potential of automated document analysis methods, including OCR and AI-based handwriting analysis, to support this task. We propose a pipeline combining template-based line segmentation with text recognition and writer retrieval techniques, evaluated on a dataset of 443 handwritten entries from 418 writers. Results show that OCR struggles with out-of-vocabulary handwriting, with a CER of 29.6% for first names. In contrast, writer retrieval performs more robustly, reaching an mAP of 50.6%. Furthermore, our experiments indicate that off-the-shelf OCR systems are not sufficiently reliable for transcription of handwritten signature data, particularly for short, out-of-vocabulary entries such as names or addresses. However, writer retrieval methods can effectively identify visually similar entries across signature lists, making them a suitable tool for supporting the detection of potential duplicate submissions based on handwriting similarity.

Marco Peer, Anna-Scius Bertrand, Patricia Scheurer et al.

We present BullingerDB, a large-scale benchmark dataset for historical document analysis based on the correspondence of Heinrich Bullinger (1504-1575). The corpus comprises 20,898 pages and 499,222 text lines written by 796 writers over six decades, featuring stylistic variation, multilingual content (mostly Latin and Early New High German) as well as meta-information such as writer identity and time. We evaluate BullingerDB on text recognition and writer retrieval. TrOCR, the best performing model, achieves a CER of 9.1%. For writer retrieval, we introduce a temporal nDCG metric to assess time-aware retrieval. While temporally coherent retrieval is achievable, mAP (78.3%) scores indicate challenges due to long-term stylistic variation. With BullingerDB, we aim to establish a new benchmark for multilingual historical text recognition and temporally-aware writer analysis.

Alicia Fornés, Asma Bensalah, Cristina Carmona-Duarte et al.

This project explores the feasibility of remote patient monitoring based on the analysis of 3D movements captured with smartwatches. We base our analysis on the Kinematic Theory of Rapid Human Movement. We have validated our research in a real case scenario for stroke rehabilitation at the Guttmann Institute5 (neurorehabilitation hospital), showing promising results. Our work could have a great impact in remote healthcare applications, improving the medical efficiency and reducing the healthcare costs. Future steps include more clinical validation, developing multi-modal analysis architectures (analysing data from sensors, images, audio, etc.), and exploring the application of our technology to monitor other neurodegenerative diseases.

Andreas Fischer, Bernhard Eidel

A cornerstone of numerical homogenization is the equivalence of the microscopic and the macroscopic energy densities, which is referred to as Hill-Mandel condition. Among these coupling conditions, the cases of periodic, linear displacement and constant traction conditions are most prominent in engineering applications. While the stiffness hierarchy of these coupling conditions is a theoretically established and numerically verified result, very little is known about the numerical errors and convergence properties for each of them in various norms. The present work addresses these aspects both on the macroscale and the microscale for linear as well as quadratic finite element shape functions. The analysis addresses aspects of (i) regularity and how its loss affects the convergence behavior on both scales compared with the a priori estimates, of (ii) error propagation from micro to macro and of (iii) optimal micro-macro mesh refinement strategy. For constant traction conditions two different approaches are compared. The performance of a recovery-type error estimation based on superconvergence is assessed. All results of the present work are valid for both the Finite Element Heterogeneous Multiscale Method FE-HMM and for FE$^2$.

Vedasri Nakka, Andreas Fischer, Rolf Ingold et al.

This thesis investigates the effectiveness of SimCLR, a contrastive learning technique, in Greek letter recognition, focusing on the impact of various augmentation techniques. We pretrain the SimCLR backbone using the Alpub dataset (pretraining dataset) and fine-tune it on a smaller ICDAR dataset (finetuning dataset) to compare SimCLR's performance against traditional baseline models, which use cross-entropy and triplet loss functions. Additionally, we explore the role of different data augmentation strategies, essential for the SimCLR training process. Methodologically, we examine three primary approaches: (1) a baseline model using cross-entropy loss, (2) a triplet embedding model with a classification layer, and (3) a SimCLR pretrained model with a classification layer. Initially, we train the baseline, triplet, and SimCLR models using 93 augmentations on ResNet-18 and ResNet-50 networks with the ICDAR dataset. From these, the top four augmentations are selected using a statistical t-test. Pretraining of SimCLR is conducted on the Alpub dataset, followed by fine-tuning on the ICDAR dataset. The triplet loss model undergoes a similar process, being pretrained on the top four augmentations before fine-tuning on ICDAR. Our experiments show that SimCLR does not outperform the baselines in letter recognition tasks. The baseline model with cross-entropy loss demonstrates better performance than both SimCLR and the triplet loss model. This study provides a detailed evaluation of contrastive learning for letter recognition, highlighting SimCLR's limitations while emphasizing the strengths of traditional supervised learning models in this task. We believe SimCLR's cropping strategies may cause a semantic shift in the input image, reducing training effectiveness despite the large pretraining dataset. Our code is available at https://github.com/DIVA-DIA/MT_augmentation_and_contrastive_learning/.

Michael Jungo, Beat Wolf, Andrii Maksai et al.

On-line handwritten character segmentation is often associated with handwriting recognition and even though recognition models include mechanisms to locate relevant positions during the recognition process, it is typically insufficient to produce a precise segmentation. Decoupling the segmentation from the recognition unlocks the potential to further utilize the result of the recognition. We specifically focus on the scenario where the transcription is known beforehand, in which case the character segmentation becomes an assignment problem between sampling points of the stylus trajectory and characters in the text. Inspired by the $k$-means clustering algorithm, we view it from the perspective of cluster assignment and present a Transformer-based architecture where each cluster is formed based on a learned character query in the Transformer decoder block. In order to assess the quality of our approach, we create character segmentation ground truths for two popular on-line handwriting datasets, IAM-OnDB and HANDS-VNOnDB, and evaluate multiple methods on them, demonstrating that our approach achieves the overall best results.

Christophe Stammet, Prisca Dotti, Ulrich Ultes-Nitsche et al.

Büchi Automata on infinite words present many interesting problems and are used frequently in program verification and model checking. A lot of these problems on Büchi automata are computationally hard, raising the question if a learning-based data-driven analysis might be more efficient than using traditional algorithms. Since Büchi automata can be represented by graphs, graph neural networks are a natural choice for such a learning-based analysis. In this paper, we demonstrate how graph neural networks can be used to reliably predict basic properties of Büchi automata when trained on automatically generated random automata datasets.

Matthias Thurnbauer, Johannes Reisinger, Christoph Goller et al.

Ambiguity is ubiquitous in natural language. Resolving ambiguous meanings is especially important in information retrieval tasks. While word embeddings carry semantic information, they fail to handle ambiguity well. Transformer models have been shown to handle word ambiguity for complex queries, but they cannot be used to identify ambiguous words, e.g. for a 1-word query. Furthermore, training these models is costly in terms of time, hardware resources, and training data, prohibiting their use in specialized environments with sensitive data. Word embeddings can be trained using moderate hardware resources. This paper shows that applying DBSCAN clustering to the latent space can identify ambiguous words and evaluate their level of ambiguity. An automatic DBSCAN parameter selection leads to high-quality clusters, which are semantically coherent and correspond well to the perceived meanings of a given word.

Adrian Iste, Kazuki Nishizawa, Chisa Tanaka et al.

Digital handwriting acquisition enables the capture of detailed temporal and kinematic signals reflecting the motor processes underlying writing behavior. While handwriting analysis has been extensively explored in clinical or adult populations, its potential for studying developmental and educational characteristics in children remains less investigated. In this work, we examine whether handwriting dynamics encode information related to student characteristics using a large-scale online dataset collected from Japanese students from elementary school to junior high school. We systematically compare three families of handwriting-derived features: basic statistical descriptors of kinematic signals, entropy-based measures of variability, and parameters obtained from the sigma-lognormal model. Although the dataset contains dense stroke-level recordings, features are aggregated at the student level to enable a controlled comparison between representations. These features are evaluated across three prediction tasks: grade prediction, gender classification, and academic performance classification, using Linear or Logistic Regression and Random Forest models under consistent experimental settings. The results show that handwriting dynamics contain measurable signals related to developmental stage and individual differences, especially for the grade prediction task. These findings highlight the potential of kinematic handwriting analysis and confirm that through their development, children's handwriting evolves toward a lognormal motor organization.

Chisa Tanaka, Andrew Vargo, Anna Scius-Bertrand et al.

While handwriting has traditionally been studied for character recognition and disease classification, its potential to reflect day-to-day physiological fluctuations in healthy individuals remains unexplored. This study examines whether daily variations in sleep-related recovery states can be inferred from online handwriting dynamics. % We propose a personalized binary classification framework that detects low-recovery days using features derived from the Sigma-Lognormal model, which captures the neuromotor generation process of pen strokes. In a 28-day in-the-wild study involving 13 university students, handwriting was recorded three times daily, and nocturnal cardiac indicators were measured using a wearable ring. For each participant, the lowest (or highest) quartile of four sleep-related metrics -- HRV, lowest heart rate, average heart rate, and total sleep duration -- defined the positive class. Leave-One-Day-Out cross-validation showed that PR-AUC significantly exceeded the baseline (0.25) for all four variables after FDR correction, with the strongest performance observed for cardiac-related variables. Importantly, classification performance did not differ significantly across task types or recording timings, indicating that recovery-related signals are embedded in general movement dynamics. These results demonstrate that subtle within-person autonomic recovery fluctuations can be detected from everyday handwriting, opening a new direction for non-invasive, device-independent health monitoring.

Michael Jungo, Andreas Fischer

Rule-based reinforcement learning has been gaining popularity ever since DeepSeek-R1 has demonstrated its success through simple verifiable rewards. In the domain of document analysis, reinforcement learning is not as prevalent, even though many downstream tasks may benefit from the emerging properties of reinforcement learning, particularly the enhanced reason capabilities. We study the effects of rule-based reinforcement learning with the task of Document Image Classification which is one of the most commonly studied downstream tasks in document analysis. We find that reinforcement learning tends to have better generalisation capabilities to out-of-distritbution data, which we examine in three different scenarios, namely out-of-distribution images, unseen classes and different modalities. Our code is available at https://github.com/jungomi/vision-finetune.

Marco Peer, Anna Scius-Bertrand, Andreas Fischer

Handwritten text recognition for historical documents remains challenging due to handwriting variability, degraded sources, and limited layout-aware annotations. In this work, we address annotation errors - particularly hyphenation issues - in the Bullinger correspondence, a large 16th-century letter collection. We introduce a self-training method based on a CTC alignment algorithm that matches full transcriptions to text line images using dynamic programming and model output probabilities trained with the CTC loss. Our approach improves performance (e.g., by 1.1 percentage points CER with PyLaia) and increases alignment accuracy. Interestingly, we find that weaker models yield more accurate alignments, enabling an iterative training strategy. We release a new manually corrected subset of 100 pages from the Bullinger dataset, along with our code and benchmarks. Our approach can be applied iteratively to further improve the CER as well as the alignment quality for text recognition pipelines. Code and data are available via https://github.com/andreas-fischer-unifr/nntp.

Lars Vögtlin, Anna Scius-Bertrand, Paul Maergner et al.

Deep learning methods have shown strong performance in solving tasks for historical document image analysis. However, despite current libraries and frameworks, programming an experiment or a set of experiments and executing them can be time-consuming. This is why we propose an open-source deep learning framework, DIVA-DAF, which is based on PyTorch Lightning and specifically designed for historical document analysis. Pre-implemented tasks such as segmentation and classification can be easily used or customized. It is also easy to create one's own tasks with the benefit of powerful modules for loading data, even large data sets, and different forms of ground truth. The applications conducted have demonstrated time savings for the programming of a document analysis task, as well as for different scenarios such as pre-training or changing the architecture. Thanks to its data module, the framework also allows to reduce the time of model training significantly.

Christian Abbet, Linda Studer, Andreas Fischer et al.

Supervised learning is constrained by the availability of labeled data, which are especially expensive to acquire in the field of digital pathology. Making use of open-source data for pre-training or using domain adaptation can be a way to overcome this issue. However, pre-trained networks often fail to generalize to new test domains that are not distributed identically due to tissue stainings, types, and textures variations. Additionally, current domain adaptation methods mainly rely on fully-labeled source datasets. In this work, we propose Self-Rule to Multi-Adapt (SRMA), which takes advantage of self-supervised learning to perform domain adaptation, and removes the necessity of fully-labeled source datasets. SRMA can effectively transfer the discriminative knowledge obtained from a few labeled source domain's data to a new target domain without requiring additional tissue annotations. Our method harnesses both domains' structures by capturing visual similarity with intra-domain and cross-domain self-supervision. Moreover, we present a generalized formulation of our approach that allows the framework to learn from multiple source domains. We show that our proposed method outperforms baselines for domain adaptation of colorectal tissue type classification \new{in single and multi-source settings}, and further validate our approach on an in-house clinical cohort. The code and trained models are available open-source: https://github.com/christianabbet/SRA.

Zineddine Bettouche, Andreas Fischer

Modern performance on several natural language processing (NLP) tasks has been enhanced thanks to the Transformer-based pre-trained language model BERT. We employ this concept to investigate a local publication database. Research papers are encoded and clustered to form a landscape view of the scientific topics, in which research is active. Authors working on similar topics can be identified by calculating the similarity between their papers. Based on this, we define a similarity metric between authors. Additionally we introduce the concept of self-similarity to indicate the topical variety of authors.

Zineddine Bettouche, Andreas Fischer

The process of transforming a raster image into a vector representation is known as image tracing. This study looks into several processing methods that include high-pass filtering, autoencoding, and vectorization to extract an abstract representation of an image. According to the findings, rebuilding an image with autoencoders, high-pass filtering it, and then vectorizing it can represent the image more abstractly while increasing the effectiveness of the vectorization process.

Michael Hollmer, Andreas Fischer

In this paper the concept of a machine learning based hands-on detection algorithm is proposed. The hand detection is implemented on the hardware side using a capacitive method. A sensor mat in the steering wheel detects a change in capacity as soon as the driver's hands come closer. The evaluation and final decision about hands-on or hands-off situations is done using machine learning. In order to find a suitable machine learning model, different models are implemented and evaluated. Based on accuracy, memory consumption and computational effort the most promising one is selected and ported on a micro controller. The entire system is then evaluated in terms of reliability and response time.

Anna Scius-Bertrand, Michael Jungo, Lars Vögtlin et al.

Classifying scanned documents is a challenging problem that involves image, layout, and text analysis for document understanding. Nevertheless, for certain benchmark datasets, notably RVL-CDIP, the state of the art is closing in to near-perfect performance when considering hundreds of thousands of training samples. With the advent of large language models (LLMs), which are excellent few-shot learners, the question arises to what extent the document classification problem can be addressed with only a few training samples, or even none at all. In this paper, we investigate this question in the context of zero-shot prompting and few-shot model fine-tuning, with the aim of reducing the need for human-annotated training samples as much as possible.

Zineddine Bettouche, Anas Safi, Andreas Fischer

Managing the semantic quality of the categorization in large textual datasets, such as Wikipedia, presents significant challenges in terms of complexity and cost. In this paper, we propose leveraging transformer models to distill semantic information from texts in the Wikipedia dataset and its associated categories into a latent space. We then explore different approaches based on these encodings to assess and enhance the semantic identity of the categories. Our graphical approach is powered by Convex Hull, while we utilize Hierarchical Navigable Small Worlds (HNSWs) for the hierarchical approach. As a solution to the information loss caused by the dimensionality reduction, we modulate the following mathematical solution: an exponential decay function driven by the Euclidean distances between the high-dimensional encodings of the textual categories. This function represents a filter built around a contextual category and retrieves items with a certain Reconsideration Probability (RP). Retrieving high-RP items serves as a tool for database administrators to improve data groupings by providing recommendations and identifying outliers within a contextual framework.

Michael Jungo, Lars Vögtlin, Atefeh Fakhari et al.

Handwriting recognition is a key technology for accessing the content of old manuscripts, helping to preserve cultural heritage. Deep learning shows an impressive performance in solving this task. However, to achieve its full potential, it requires a large amount of labeled data, which is difficult to obtain for ancient languages and scripts. Often, a trade-off has to be made between ground truth quantity and quality, as is the case for the recently introduced Bullinger database. It contains an impressive amount of over a hundred thousand labeled text line images of mostly premodern German and Latin texts that were obtained by automatically aligning existing page-level transcriptions with text line images. However, the alignment process introduces systematic errors, such as wrongly hyphenated words. In this paper, we investigate the impact of such errors on training and evaluation and suggest means to detect and correct typical alignment errors.

Zineddine Bettouche, Khalid Ali, Andreas Fischer et al.

Traffic forecasting in cellular networks is a challenging spatiotemporal prediction problem due to strong temporal dependencies, spatial heterogeneity across cells, and the need for scalability to large network deployments. Traditional cell-specific models incur prohibitive training and maintenance costs, while global models often fail to capture heterogeneous spatial dynamics. Recent spatiotemporal architectures based on attention or graph neural networks improve accuracy but introduce high computational overhead, limiting their applicability in large-scale or real-time settings. We study spatiotemporal grid forecasting, where each time step is a 2D lattice of traffic values, and predict the next grid patch using previous patches. We propose NeST-S6, a convolutional selective state-space model (SSM) with a spatial PDE-aware core, implemented in a nested learning paradigm: convolutional local spatial mixing feeds a spatial PDE-aware SSM core, while a nested-learning long-term memory is updated by a learned optimizer when one-step prediction errors indicate unmodeled dynamics. On the mobile-traffic grid (Milan dataset) at three resolutions (202, 502, 1002), NeST-S6 attains lower errors than a strong Mamba-family baseline in both single-step and 6-step autoregressive rollouts. Under drift stress tests, our model's nested memory lowers MAE by 48-65% over a no-memory ablation. NeST-S6 also speeds full-grid reconstruction by 32 times and reduces MACs by 4.3 times compared to competitive per-pixel scanning models, while achieving 61% lower per-pixel RMSE.

Ahmad Alsheikh, Andreas Fischer

Accurate and efficient temperature prediction is critical for optimizing the preheating process of PET preforms in industrial microwave systems prior to blow molding. We propose a novel deep learning framework for generalized temperature prediction. Unlike traditional models that require extensive retraining for each material or design variation, our method introduces a data-efficient neural architecture that leverages transfer learning and model fusion to generalize across unseen scenarios. By pretraining specialized neural regressor on distinct conditions such as recycled PET heat capacities or varying preform geometries and integrating their representations into a unified global model, we create a system capable of learning shared thermal dynamics across heterogeneous inputs. The architecture incorporates skip connections to enhance stability and prediction accuracy. Our approach reduces the need for large simulation datasets while achieving superior performance compared to models trained from scratch. Experimental validation on two case studies material variability and geometric diversity demonstrates significant improvements in generalization, establishing a scalable ML-based solution for intelligent thermal control in manufacturing environments. Moreover, the approach highlights how data-efficient generalization strategies can extend to other industrial applications involving complex physical modeling with limited data.

Ahmad Alsheikh, Andreas Fischer

Predicting the final hardness of steel after heat treatment is a challenging regression task due to the many-to-one nature of the process -- different combinations of input parameters (such as temperature, duration, and chemical composition) can result in the same hardness value. This ambiguity makes the inverse problem, estimating input parameters from a desired hardness, particularly difficult. In this work, we propose a novel solution using a Teacher-Student learning framework. First, a forward model (Teacher) is trained to predict final hardness from 13 metallurgical input features. Then, a backward model (Student) is trained to infer plausible input configurations from a target hardness value. The Student is optimized by leveraging feedback from the Teacher in an iterative, supervised loop. We evaluate our method on a publicly available tempered steel dataset and compare it against baseline regression and reinforcement learning models. Results show that our Teacher-Student framework not only achieves higher inverse prediction accuracy but also requires significantly less computational time, demonstrating its effectiveness and efficiency for inverse process modeling in materials science.

Joshua Heisler, Johannes Reisinger, Andreas Fischer

SAP has released its own proprietary generative model SAP Joule, intended for various generative tasks, including serving as a code assistant for software engineers. While Joule is yet not focused on SAP-specific ABAP code generation, it can be used for other common languages, including Javascript. This paper compares SAP Joules Javascript coding capabilities against a total of 29 other models using the HumanEval-X Javascript benchmark. SAP Joule achieves a strict accuracy of 80.49% as the fifth best model in our evaluation. To the best of our knowledge, this is the first comparative evaluation of SAP Joule code generation capabilities.

Zineddine Bettouche, Khalid Ali, Andreas Fischer et al.

Cellular traffic forecasting is essential for network planning, resource allocation, or load-balancing traffic across cells. However, accurate forecasting is difficult due to intricate spatial and temporal patterns that exist due to the mobility of users. Existing AI-based traffic forecasting models often trade-off accuracy and computational efficiency. We present Hierarchical SpatioTemporal Mamba (HiSTM), which combines a dual spatial encoder with a Mamba-based temporal module and attention mechanism. HiSTM employs selective state space methods to capture spatial and temporal patterns in network traffic. In our evaluation, we use a real-world dataset to compare HiSTM against several baselines, showing a 29.4% MAE improvement over the STN baseline while using 94% fewer parameters. We show that the HiSTM generalizes well across different datasets and improves in accuracy over longer time-horizons.

Khalid Ali, Zineddine Bettouche, Andreas Kassler et al.

Accurate spatiotemporal traffic forecasting is vital for intelligent resource management in 5G and beyond. However, conventional AI approaches often fail to capture the intricate spatial and temporal patterns that exist, due to e.g., the mobility of users. We introduce a lightweight, dual-path Spatiotemporal Network that leverages a Scalar LSTM (sLSTM) for efficient temporal modeling and a three-layer Conv3D module for spatial feature extraction. A fusion layer integrates both streams into a cohesive representation, enabling robust forecasting. Our design improves gradient stability and convergence speed while reducing prediction error. Evaluations on real-world datasets show superior forecast performance over ConvLSTM baselines and strong generalization to unseen regions, making it well-suited for large-scale, next-generation network deployments. Experimental evaluation shows a 23% MAE reduction over ConvLSTM, with a 30% improvement in model generalization.

Krishna Sahukara, Zineddine Bettouche, Andreas Fischer

Document pages captured by smartphones or scanners often contain tables, yet manual extraction is slow and error-prone. We introduce an automated LaTeX-based pipeline that synthesizes realistic two-column pages with visually diverse table layouts and aligned ground-truth masks. The generated corpus augments the real-world Marmot benchmark and enables a systematic resolution study of TableNet. Training TableNet on our synthetic data achieves a pixel-wise XOR error of 4.04% on our synthetic test set with a 256x256 input resolution, and 4.33% with 1024x1024. The best performance on the Marmot benchmark is 9.18% (at 256x256), while cutting manual annotation effort through automation.

Pau Riba, Andreas Fischer, Josep Lladós et al.

The emergence of geometric deep learning as a novel framework to deal with graph-based representations has faded away traditional approaches in favor of completely new methodologies. In this paper, we propose a new framework able to combine the advances on deep metric learning with traditional approximations of the graph edit distance. Hence, we propose an efficient graph distance based on the novel field of geometric deep learning. Our method employs a message passing neural network to capture the graph structure, and thus, leveraging this information for its use on a distance computation. The performance of the proposed graph distance is validated on two different scenarios. On the one hand, in a graph retrieval of handwritten words~\ie~keyword spotting, showing its superior performance when compared with (approximate) graph edit distance benchmarks. On the other hand, demonstrating competitive results for graph similarity learning when compared with the current state-of-the-art on a recent benchmark dataset.

Lucy Linder, Michael Jungo, Jean Hennebert et al.

This paper presents SwissCrawl, the largest Swiss German text corpus to date. Composed of more than half a million sentences, it was generated using a customized web scraping tool that could be applied to other low-resource languages as well. The approach demonstrates how freely available web pages can be used to construct comprehensive text corpora, which are of fundamental importance for natural language processing. In an experimental evaluation, we show that using the new corpus leads to significant improvements for the task of language modeling. To capture new content, our approach will run continuously to keep increasing the corpus over time.

Noémien Kocher, Christian Scuito, Lorenzo Tarantino et al.

In sequence modeling tasks the token order matters, but this information can be partially lost due to the discretization of the sequence into data points. In this paper, we study the imbalance between the way certain token pairs are included in data points and others are not. We denote this a token order imbalance (TOI) and we link the partial sequence information loss to a diminished performance of the system as a whole, both in text and speech processing tasks. We then provide a mechanism to leverage the full token order information -Alleviated TOI- by iteratively overlapping the token composition of data points. For recurrent networks, we use prime numbers for the batch size to avoid redundancies when building batches from overlapped data points. The proposed method achieved state of the art performance in both text and speech related tasks.

Paul Maergner, Nicholas R. Howe, Kaspar Riesen et al.

Graphs provide a powerful representation formalism that offers great promise to benefit tasks like handwritten signature verification. While most state-of-the-art approaches to signature verification rely on fixed-size representations, graphs are flexible in size and allow modeling local features as well as the global structure of the handwriting. In this article, we present two recent graph-based approaches to offline signature verification: keypoint graphs with approximated graph edit distance and inkball models. We provide a comprehensive description of the methods, propose improvements both in terms of computational time and accuracy, and report experimental results for four benchmark datasets. The proposed methods achieve top results for several benchmarks, highlighting the potential of graph-based signature verification.

Linda Studer, Michele Alberti, Vinaychandran Pondenkandath et al.

Automatic analysis of scanned historical documents comprises a wide range of image analysis tasks, which are often challenging for machine learning due to a lack of human-annotated learning samples. With the advent of deep neural networks, a promising way to cope with the lack of training data is to pre-train models on images from a different domain and then fine-tune them on historical documents. In the current research, a typical example of such cross-domain transfer learning is the use of neural networks that have been pre-trained on the ImageNet database for object recognition. It remains a mostly open question whether or not this pre-training helps to analyse historical documents, which have fundamentally different image properties when compared with ImageNet. In this paper, we present a comprehensive empirical survey on the effect of ImageNet pre-training for diverse historical document analysis tasks, including character recognition, style classification, manuscript dating, semantic segmentation, and content-based retrieval. While we obtain mixed results for semantic segmentation at pixel-level, we observe a clear trend across different network architectures that ImageNet pre-training has a positive effect on classification as well as content-based retrieval.

Paul Maergner, Vinaychandran Pondenkandath, Michele Alberti et al.

Biometric authentication by means of handwritten signatures is a challenging pattern recognition task, which aims to infer a writer model from only a handful of genuine signatures. In order to make it more difficult for a forger to attack the verification system, a promising strategy is to combine different writer models. In this work, we propose to complement a recent structural approach to offline signature verification based on graph edit distance with a statistical approach based on metric learning with deep neural networks. On the MCYT and GPDS benchmark datasets, we demonstrate that combining the structural and statistical models leads to significant improvements in performance, profiting from their complementary properties.

Baptiste Wicht, Jean Hennebert, Andreas Fischer

Deep Learning Library (DLL) is a new library for machine learning with deep neural networks that focuses on speed. It supports feed-forward neural networks such as fully-connected Artificial Neural Networks (ANNs) and Convolutional Neural Networks (CNNs). It also has very comprehensive support for Restricted Boltzmann Machines (RBMs) and Convolutional RBMs. Our main motivation for this work was to propose and evaluate novel software engineering strategies with potential to accelerate runtime for training and inference. Such strategies are mostly independent of the underlying deep learning algorithms. On three different datasets and for four different neural network models, we compared DLL to five popular deep learning frameworks. Experimentally, it is shown that the proposed framework is systematically and significantly faster on CPU and GPU. In terms of classification performance, similar accuracies as the other frameworks are reported.

Andreas Fischer, Benny Fuhry, Florian Kerschbaum et al.

Encrypting data before sending it to the cloud protects it against hackers and malicious insiders, but requires the cloud to compute on encrypted data. Trusted (hardware) modules, e.g., secure enclaves like Intel's SGX, can very efficiently run entire programs in encrypted memory. However, it already has been demonstrated that software vulnerabilities give an attacker ample opportunity to insert arbitrary code into the program. This code can then modify the data flow of the program and leak any secret in the program to an observer in the cloud via SGX side-channels. Since any larger program is rife with software vulnerabilities, it is not a good idea to outsource entire programs to an SGX enclave. A secure alternative with a small trusted code base would be fully homomorphic encryption (FHE) -- the holy grail of encrypted computation. However, due to its high computational complexity it is unlikely to be adopted in the near future. As a result researchers have made several proposals for transforming programs to perform encrypted computations on less powerful encryption schemes. Yet, current approaches fail on programs that make control-flow decisions based on encrypted data. In this paper, we introduce the concept of data flow authentication (DFAuth). DFAuth prevents an adversary from arbitrarily deviating from the data flow of a program. Hence, an attacker cannot perform an attack as outlined before on SGX. This enables that all programs, even those including operations on control-flow decision variables, can be computed on encrypted data. We implemented DFAuth using a novel authenticated homomorphic encryption scheme, a Java bytecode-to-bytecode compiler producing fully executable programs, and SGX enclaves. A transformed neural network that performs machine learning on sensitive medical data can be evaluated on encrypted inputs and encrypted weights in 0.86 seconds.

Bernhard Eidel, Andreas Fischer

The Heterogeneous Multiscale Finite Element Method (FE-HMM) is a two-scale FEM based on asymptotic homogenization for solving multiscale partial differential equations. It was introduced in [W. E and B. Engquist, \emph{Commun. Math. Sci.}, 1 (2003), 87--132]. The objective of the present work is an FE-HMM formulation for the homogenization of linear elastic solids in a geometrical linear frame, and doing so, for the first time, of a vector-valued field problem. A key ingredient of FE-HMM is that macrostiffness is estimated by stiffness sampling on heterogeneous microdomains in terms a of modified quadrature formula, which implies an equivalence of energy densities of the microscale with the macroscale. Beyond this coincidence with the Hill-Mandel macrohomogeneity condition, which is the cornerstone of the FE$^2$ method, we elaborate a conceptual comparison with the latter method. After developing an algorithmic framework we (i) assess the existing a priori convergence estimates for the micro- and macro-errors in various norms, (ii) verify optimal strategies in uniform micro-macro mesh refinements based on the estimates, (iii) analyze superconvergence properties of FE-HMM, and (iv) compare the numerical results of FE-HMM with those of FE$^2$.