Stefan Krämer

Tony Hauptmann, Stefan Kramer

Recent years have seen a surge of novel neural network architectures for the integration of multi-omics data for prediction. Most of the architectures include either encoders alone or encoders and decoders, i.e., autoencoders of various sorts, to transform multi-omics data into latent representations. One important parameter is the depth of integration: the point at which the latent representations are computed or merged, which can be either early, intermediate, or late. The literature on integration methods is growing steadily, however, close to nothing is known about the relative performance of these methods under fair experimental conditions and under consideration of different use cases. We developed a comparison framework that trains and optimizes multi-omics integration methods under equal conditions. We incorporated early integration and four recently published deep learning methods: MOLI, Super.FELT, OmiEmbed, and MOMA. Further, we devised a novel method, Omics Stacking, that combines the advantages of intermediate and late integration. Experiments were conducted on a public drug response data set with multiple omics data (somatic point mutations, somatic copy number profiles and gene expression profiles) that was obtained from cell lines, patient-derived xenografts, and patient samples. Our experiments confirmed that early integration has the lowest predictive performance. Overall, architectures that integrate triplet loss achieved the best results. Statistical differences can, overall, rarely be observed, however, in terms of the average ranks of methods, Super.FELT is consistently performing best in a cross-validation setting and Omics Stacking best in an external test set setting. The source code of all experiments is available under \url{https://github.com/kramerlab/Multi-Omics_analysis}

Derian Boer, Fabian Koch, Stefan Kramer

Large Language Models (LLMs) frequently lack domain-specific knowledge and even fine-tuned models tend to hallucinate. Hence, more reliable models that can include external knowledge are needed. We present a pipeline, 4StepFocus, and specifically a preprocessing step, that can substantially improve the answers of LLMs. This is achieved by providing guided access to external knowledge making use of the model's ability to capture relational context and conduct rudimentary reasoning by themselves. The method narrows down potentially correct answers by triplets-based searches in a semi-structured knowledge base in a direct, traceable fashion, before switching to latent representations for ranking those candidates based on unstructured data. This distinguishes it from related methods that are purely based on latent representations. 4StepFocus consists of the steps: 1) Triplet generation for extraction of relational data by an LLM, 2) substitution of variables in those triplets to narrow down answer candidates employing a knowledge graph, 3) sorting remaining candidates with a vector similarity search involving associated non-structured data, 4) reranking the best candidates by the LLM with background data provided. Experiments on a medical, a product recommendation, and an academic paper search test set demonstrate that this approach is indeed a powerful augmentation. It not only adds relevant traceable background information from information retrieval, but also improves performance considerably in comparison to state-of-the-art methods. This paper presents a novel, largely unexplored direction and therefore provides a wide range of future work opportunities. Used source code is available at https://github.com/kramerlab/4StepFocus.

Mattia Cerrato, Marius Köppel, Roberto Esposito et al.

Representation learning algorithms offer the opportunity to learn invariant representations of the input data with regard to nuisance factors. Many authors have leveraged such strategies to learn fair representations, i.e., vectors where information about sensitive attributes is removed. These methods are attractive as they may be interpreted as minimizing the mutual information between a neural layer's activations and a sensitive attribute. However, the theoretical grounding of such methods relies either on the computation of infinitely accurate adversaries or on minimizing a variational upper bound of a mutual information estimate. In this paper, we propose a methodology for direct computation of the mutual information between a neural layer and a sensitive attribute. We employ stochastically-activated binary neural networks, which lets us treat neurons as random variables. We are then able to compute (not bound) the mutual information between a layer and a sensitive attribute and use this information as a regularization factor during gradient descent. We show that this method compares favorably with the state of the art in fair representation learning and that the learned representations display a higher level of invariance compared to full-precision neural networks.

Lukas Pensel, Stefan Kramer

Multi-relational databases are the basis of most consolidated data collections in science and industry today. Most learning and mining algorithms, however, require data to be represented in a propositional form. While there is a variety of specialized machine learning algorithms that can operate directly on multi-relational data sets, propositionalization algorithms transform multi-relational databases into propositional data sets, thereby allowing the application of traditional machine learning and data mining algorithms without their modification. One prominent propositionalization algorithm is RELAGGS by Krogel and Wrobel, which transforms the data by nested aggregations. We propose a new neural network based algorithm in the spirit of RELAGGS that employs trainable composite aggregate functions instead of the static aggregate functions used in the original approach. In this way, we can jointly train the propositionalization with the prediction model, or, alternatively, use the learned aggegrations as embeddings in other algorithms. We demonstrate the increased predictive performance by comparing N-RELAGGS with RELAGGS and multiple other state-of-the-art algorithms.

Derian Boer, Stephen Roth, Stefan Kramer

In many real-world settings, machine learning models and interactive systems have access to both structured knowledge, e.g., knowledge graphs or tables, and unstructured content, e.g., natural language documents. However, most rely on either. Semi-Structured Knowledge Bases (SKBs) bridge this gap by linking unstructured content to nodes within structured data, thereby enabling new strategies for knowledge access and use. In this work, we present FocusedRetriever, a modular SKB-based framework for multi-hop question answering. It integrates components (VSS-based entity search, LLM-based generation of Cypher queries and pairwise re-ranking) in a way that enables it to outperform state-of-the-art methods across all three STaRK benchmark test sets, covering diverse domains and multiple performance metrics. The average first-hit rate exceeds that of the second-best method by 25.7%. FocusedRetriever leverages (1) the capacity of Large Language Models (LLMs) to extract relational facts and entity attributes from unstructured text, (2) node set joins to filter answer candidates based on these extracted triplets and constraints, (3) vector similarity search to retrieve and rank relevant unstructured content, and (4) the contextual capabilities of LLMs to finally rank the top-k answers. For generality, we only incorporate base LLMs in FocusedRetriever in our evaluation. However, our analysis of intermediate results highlights several opportunities for further upgrades including finetuning. The source code is publicly available at https://github.com/kramerlab/FocusedRetriever .

Julia Siekiera, Stefan Kramer

Artificial intelligence is increasingly leveraged across various domains to automate decision-making processes that significantly impact human lives. In medical image analysis, deep learning models have demonstrated remarkable performance. However, their inherent complexity makes them black box systems, raising concerns about reliability and interpretability. Counterfactual explanations provide comprehensible insights into decision processes by presenting hypothetical "what-if" scenarios that alter model classifications. By examining input alterations, counterfactual explanations provide patterns that influence the decision-making process. Despite their potential, generating plausible counterfactuals that adhere to similarity constraints providing human-interpretable explanations remains a challenge. In this paper, we investigate this challenge by a model-specific optimization approach. While deep generative models such as variational autoencoders (VAEs) exhibit significant generative power, probabilistic models like sum-product networks (SPNs) efficiently represent complex joint probability distributions. By modeling the likelihood of a semi-supervised VAE's latent space with an SPN, we leverage its dual role as both a latent space descriptor and a classifier for a given discrimination task. This formulation enables the optimization of latent space counterfactuals that are both close to the original data distribution and aligned with the target class distribution. We conduct experimental evaluation on the cheXpert dataset. To evaluate the effectiveness of the integration of SPNs, our SPN-guided latent space manipulation is compared against a neural network baseline. Additionally, the trade-off between latent variable regularization and counterfactual quality is analyzed.

Cedric Derstroff, Mattia Cerrato, Jannis Brugger et al.

Peer learning is a novel high-level reinforcement learning framework for agents learning in groups. While standard reinforcement learning trains an individual agent in trial-and-error fashion, all on its own, peer learning addresses a related setting in which a group of agents, i.e., peers, learns to master a task simultaneously together from scratch. Peers are allowed to communicate only about their own states and actions recommended by others: "What would you do in my situation?". Our motivation is to study the learning behavior of these agents. We formalize the teacher selection process in the action advice setting as a multi-armed bandit problem and therefore highlight the need for exploration. Eventually, we analyze the learning behavior of the peers and observe their ability to rank the agents' performance within the study group and understand which agents give reliable advice. Further, we compare peer learning with single agent learning and a state-of-the-art action advice baseline. We show that peer learning is able to outperform single-agent learning and the baseline in several challenging discrete and continuous OpenAI Gym domains. Doing so, we also show that within such a framework complex policies from action recommendations beyond discrete action spaces can evolve.

Stefan Kramer, Mattia Cerrato, Jannis Brugger et al.

The paper surveys automated scientific discovery, from equation discovery and symbolic regression to autonomous discovery systems and agents. It discusses the individual approaches from a "big picture" perspective and in context, but also discusses open issues and recent topics like the various roles of deep neural networks in this area, aiding in the discovery of human-interpretable knowledge. Further, we will present closed-loop scientific discovery systems, starting with the pioneering work on the Adam system up to current efforts in fields from material science to astronomy. Finally, we will elaborate on autonomy from a machine learning perspective, but also in analogy to the autonomy levels in autonomous driving. The maximal level, level five, is defined to require no human intervention at all in the production of scientific knowledge. Achieving this is one step towards solving the Nobel Turing Grand Challenge to develop AI Scientists: AI systems capable of making Nobel-quality scientific discoveries highly autonomously at a level comparable, and possibly superior, to the best human scientists by 2050.

Mattia Cerrato, Alesia Vallenas Coronel, Marius Köppel et al.

Neural network architectures have been extensively employed in the fair representation learning setting, where the objective is to learn a new representation for a given vector which is independent of sensitive information. Various representation debiasing techniques have been proposed in the literature. However, as neural networks are inherently opaque, these methods are hard to comprehend, which limits their usefulness. We propose a new framework for fair representation learning that is centered around the learning of "correction vectors", which have the same dimensionality as the given data vectors. Correction vectors may be computed either explicitly via architectural constraints or implicitly by training an invertible model based on Normalizing Flows. We show experimentally that several fair representation learning models constrained in such a way do not exhibit losses in ranking or classification performance. Furthermore, we demonstrate that state-of-the-art results can be achieved by the invertible model. Finally, we discuss the law standing of our methodology in light of recent legislation in the European Union.

Mattia Cerrato, Marius Köppel, Alexander Segner et al.

Neural network architectures have been extensively employed in the fair representation learning setting, where the objective is to learn a new representation for a given vector which is independent of sensitive information. Various "representation debiasing" techniques have been proposed in the literature. However, as neural networks are inherently opaque, these methods are hard to comprehend, which limits their usefulness. We propose a new framework for fair representation learning which is centered around the learning of "correction vectors", which have the same dimensionality as the given data vectors. The corrections are then simply summed up to the original features, and can therefore be analyzed as an explicit penalty or bonus to each feature. We show experimentally that a fair representation learning problem constrained in such a way does not impact performance.

Antoine Garcon, Julian Vexler, Dmitry Budker et al.

Deep neural networks (DNNs) are widely used in pattern-recognition tasks for which a human comprehensible, quantitative description of the data-generating process, e.g., in the form of equations, cannot be achieved. While doing so, DNNs often produce an abstract (entangled and non-interpretable) representation of the data-generating process. This is one of the reasons why DNNs are not extensively used in physics-signal processing: physicists generally require their analyses to yield quantitative information about the studied systems. In this article we use DNNs to disentangle components of oscillating time series, and recover meaningful information. We show that, because DNNs can find useful abstract feature representations, they can be used when prior knowledge about the signal-generating process exists, but is not complete, as it is particularly the case in "new-physics" searches. To this aim, we train our DNN on synthetic oscillating time series to perform two tasks: a regression of the signal latent parameters and signal denoising by an Autoencoder-like architecture. We show that the regression and denoising performance is similar to those of least-square curve fittings (LS-fit) with true latent parameters' initial guesses, in spite of the DNN needing no initial guesses at all. We then explore applications in which we believe our architecture could prove useful for time-series processing in physics, when prior knowledge is incomplete. As an example, we employ DNNs as a tool to inform LS-fits when initial guesses are unknown. We show that the regression can be performed on some latent parameters, while ignoring the existence of others. Because the Autoencoder needs no prior information about the physical model, the remaining unknown latent parameters can still be captured, thus making use of partial prior knowledge, while leaving space for data exploration and discoveries.

Sophie Burkhardt, Jannis Brugger, Nicolas Wagner et al.

Most deep neural networks are considered to be black boxes, meaning their output is hard to interpret. In contrast, logical expressions are considered to be more comprehensible since they use symbols that are semantically close to natural language instead of distributed representations. However, for high-dimensional input data such as images, the individual symbols, i.e. pixels, are not easily interpretable. We introduce the concept of first-order convolutional rules, which are logical rules that can be extracted using a convolutional neural network (CNN), and whose complexity depends on the size of the convolutional filter and not on the dimensionality of the input. Our approach is based on rule extraction from binary neural networks with stochastic local search. We show how to extract rules that are not necessarily short, but characteristic of the input, and easy to visualize. Our experiments show that the proposed approach is able to model the functionality of the neural network while at the same time producing interpretable logical rules.

Derian Boer, Stefan Kramer

Deep learning (DL) approaches are achieving extraordinary results in a wide range of domains, but often require a massive collection of private data. Hence, methods for training neural networks on the joint data of different data owners, that keep each party's input confidential, are called for. We address a specific setting in federated learning, namely that of deep learning from horizontally distributed data with a limited number of parties, where their vulnerable intermediate results have to be processed in a privacy-preserving manner. This setting can be found in medical and healthcare as well as industrial applications. The predominant scheme for this is based on homomorphic encryption (HE), and it is widely considered to be without alternative. In contrast to this, we demonstrate that a carefully chosen, less complex and computationally less expensive secure sum protocol in conjunction with default secure channels exhibits superior properties in terms of both collusion-resistance and runtime. Finally, we discuss several open research questions in the context of collaborative DL, especially regarding privacy risks caused by joint intermediate results.

Marius Köppel, Alexander Segner, Martin Wagener et al.

We present a pairwise learning to rank approach based on a neural net, called DirectRanker, that generalizes the RankNet architecture. We show mathematically that our model is reflexive, antisymmetric, and transitive allowing for simplified training and improved performance. Experimental results on the LETOR MSLR-WEB10K, MQ2007 and MQ2008 datasets show that our model outperforms numerous state-of-the-art methods, while being inherently simpler in structure and using a pairwise approach only.