Search Results for author: Kristian Kersting

Found 89 papers, 33 papers with code

The Biased Artist: Exploiting Cultural Biases via Homoglyphs in Text-Guided Image Generation Models

1 code implementation19 Sep 2022 Lukas Struppek, Dominik Hintersdorf, Kristian Kersting

Our results further indicate that text encoders trained on multilingual data provide a way to mitigate the effects of homoglyph replacements.

Image Generation

CLIPping Privacy: Identity Inference Attacks on Multi-Modal Machine Learning Models

1 code implementation15 Sep 2022 Dominik Hintersdorf, Lukas Struppek, Kristian Kersting

As deep learning is now used in many real-world applications, research has focused increasingly on the privacy of deep learning models and how to prevent attackers from obtaining sensitive information about the training data.

Inference Attack

Transformer-Boosted Anomaly Detection with Fuzzy Hashes

no code implementations24 Aug 2022 Frieder Uhlig, Lukas Struppek, Dominik Hintersdorf, Kristian Kersting

In this work, we propose deep learning approximate matching (DLAM), which achieves much higher accuracy in detecting anomalies in fuzzy hashes than conventional approaches.

Anomaly Detection

ILLUME: Rationalizing Vision-Language Models by Interacting with their Jabber

1 code implementation17 Aug 2022 Manuel Brack, Patrick Schramowski, Björn Deiseroth, Kristian Kersting

Bootstrapping from pre-trained language models has been proven to be an efficient approach for building foundation vision-language models (VLM) for tasks such as image captioning or visual question answering.

Image Captioning Question Answering +2

HANF: Hyperparameter And Neural Architecture Search in Federated Learning

no code implementations24 Jun 2022 Jonas Seng, Pooja Prasad, Devendra Singh Dhami, Kristian Kersting

We show that HANF efficiently finds the optimized neural architecture and also tunes the hyperparameters on data owner servers.

BIG-bench Machine Learning Federated Learning +2

Towards a Solution to Bongard Problems: A Causal Approach

no code implementations14 Jun 2022 Salahedine Youssef, Matej Zečević, Devendra Singh Dhami, Kristian Kersting

To date, Bongard Problems (BP) remain one of the few fortresses of AI history yet to be raided by the powerful models of the current era.


Can Foundation Models Talk Causality?

no code implementations14 Jun 2022 Moritz Willig, Matej Zečević, Devendra Singh Dhami, Kristian Kersting

Foundation models are subject to an ongoing heated debate, leaving open the question of progress towards AGI and dividing the community into two camps: the ones who see the arguably impressive results as evidence to the scaling hypothesis, and the others who are worried about the lack of interpretability and reasoning capabilities.

Machines Explaining Linear Programs

no code implementations14 Jun 2022 David Steinmann, Matej Zečević, Devendra Singh Dhami, Kristian Kersting

In this work, we extend the attribution methods for explaining neural networks to linear programs.

Attributions Beyond Neural Networks: The Linear Program Case

no code implementations14 Jun 2022 Florian Peter Busch, Matej Zečević, Kristian Kersting, Devendra Singh Dhami

We introduce an approach where we consider neural encodings for LPs that justify the application of attribution methods from explainable artificial intelligence (XAI) designed for neural learning systems.

Explainable artificial intelligence

Gradient-based Counterfactual Explanations using Tractable Probabilistic Models

no code implementations16 May 2022 Xiaoting Shao, Kristian Kersting

Counterfactual examples are an appealing class of post-hoc explanations for machine learning models.

Adaptable Adapters

1 code implementation NAACL 2022 Nafise Sadat Moosavi, Quentin Delfosse, Kristian Kersting, Iryna Gurevych

The resulting adapters (a) contain about 50% of the learning parameters of the standard adapter and are therefore more efficient at training and inference, and require less storage space, and (b) achieve considerably higher performances in low-data settings.

Finding Structure and Causality in Linear Programs

1 code implementation29 Mar 2022 Matej Zečević, Florian Peter Busch, Devendra Singh Dhami, Kristian Kersting

Linear Programs (LP) are celebrated widely, particularly so in machine learning where they have allowed for effectively solving probabilistic inference tasks or imposing structure on end-to-end learning systems.

BIG-bench Machine Learning

Do Multilingual Language Models Capture Differing Moral Norms?

no code implementations18 Mar 2022 Katharina Hämmerl, Björn Deiseroth, Patrick Schramowski, Jindřich Libovický, Alexander Fraser, Kristian Kersting

Massively multilingual sentence representations are trained on large corpora of uncurated data, with a very imbalanced proportion of languages included in the training.

A Typology to Explore and Guide Explanatory Interactive Machine Learning

no code implementations4 Mar 2022 Felix Friedrich, Wolfgang Stammer, Patrick Schramowski, Kristian Kersting

In addition to benchmarking these methods on their overall ability to revise a model, we perform additional benchmarks regarding wrong reason revision, interaction efficiency, robustness to feedback quality, and the ability to revise a strongly corrupted model.

BIG-bench Machine Learning

Neuro-Symbolic Verification of Deep Neural Networks

1 code implementation2 Mar 2022 Xuan Xie, Kristian Kersting, Daniel Neider

Formal verification has emerged as a powerful approach to ensure the safety and reliability of deep neural networks.

Adversarial Robustness Fairness

Right for the Right Latent Factors: Debiasing Generative Models via Disentanglement

no code implementations1 Feb 2022 Xiaoting Shao, Karl Stelzner, Kristian Kersting

A key assumption of most statistical machine learning methods is that they have access to independent samples from the distribution of data they encounter at test time.

BIG-bench Machine Learning Disentanglement

Plug & Play Attacks: Towards Robust and Flexible Model Inversion Attacks

1 code implementation28 Jan 2022 Lukas Struppek, Dominik Hintersdorf, Antonio De Almeida Correia, Antonia Adler, Kristian Kersting

Model inversion attacks (MIAs) aim to create synthetic images that reflect the class-wise characteristics from a target classifier's private training data by exploiting the model's learned knowledge.

Interactive Disentanglement: Learning Concepts by Interacting with their Prototype Representations

1 code implementation CVPR 2022 Wolfgang Stammer, Marius Memmel, Patrick Schramowski, Kristian Kersting

In this work, we show the advantages of prototype representations for understanding and revising the latent space of neural concept learners.


To Trust or Not To Trust Prediction Scores for Membership Inference Attacks

1 code implementation17 Nov 2021 Dominik Hintersdorf, Lukas Struppek, Kristian Kersting

Membership inference attacks (MIAs) aim to determine whether a specific sample was used to train a predictive model.

On the Tractability of Neural Causal Inference

no code implementations22 Oct 2021 Matej Zečević, Devendra Singh Dhami, Kristian Kersting

Roth (1996) proved that any form of marginal inference with probabilistic graphical models (e. g. Bayesian Networks) will at least be NP-hard.

Causal Inference

The Causal Loss: Driving Correlation to Imply Causation

no code implementations22 Oct 2021 Moritz Willig, Matej Zečević, Devendra Singh Dhami, Kristian Kersting

Most algorithms in classical and contemporary machine learning focus on correlation-based dependence between features to drive performance.

Explaining Deep Tractable Probabilistic Models: The sum-product network case

no code implementations19 Oct 2021 Athresh Karanam, Saurabh Mathur, Predrag Radivojac, David M. Haas, Kristian Kersting, Sriraam Natarajan

We consider the problem of explaining a class of tractable deep probabilistic models, the Sum-Product Networks (SPNs) and present an algorithm ExSPN to generate explanations.

Neuro-Symbolic Forward Reasoning

1 code implementation18 Oct 2021 Hikaru Shindo, Devendra Singh Dhami, Kristian Kersting

NSFR factorizes the raw inputs into the object-centric representations, converts them into probabilistic ground atoms, and finally performs differentiable forward-chaining inference using weighted rules for inference.

Inferring Offensiveness In Images From Natural Language Supervision

1 code implementation8 Oct 2021 Patrick Schramowski, Kristian Kersting

Probing or fine-tuning (large-scale) pre-trained models results in state-of-the-art performance for many NLP tasks and, more recently, even for computer vision tasks when combined with image data.

SLASH: Embracing Probabilistic Circuits into Neural Answer Set Programming

no code implementations7 Oct 2021 Arseny Skryagin, Wolfgang Stammer, Daniel Ochs, Devendra Singh Dhami, Kristian Kersting

The probability estimates resulting from NPPs act as the binding element between the logical program and raw input data, thereby allowing SLASH to answer task-dependent logical queries.

Probabilistic Programming

XAI Establishes a Common Ground Between Machine Learning and Causality

no code implementations5 Oct 2021 Matej Zečević, Devendra Singh Dhami, Constantin A. Rothkopf, Kristian Kersting

By recognizing how human mental models (HMM) are naturally represented by the Pearlian Structural Causal Model (SCM), we make two key observations through the construction of an example metric space for linear SCM: first, that the notion of a "true" data-underlying SCM is justified, and second, that an aggregation of human-derived SCM might point to said "true" SCM.

BIG-bench Machine Learning

Sum-Product-Attention Networks: Leveraging Self-Attention in Probabilistic Circuits

no code implementations14 Sep 2021 Zhongjie Yu, Devendra Singh Dhami, Kristian Kersting

Probabilistic circuits (PCs) have become the de-facto standard for learning and inference in probabilistic modeling.

Relating Graph Neural Networks to Structural Causal Models

no code implementations9 Sep 2021 Matej Zečević, Devendra Singh Dhami, Petar Veličković, Kristian Kersting

Causality can be described in terms of a structural causal model (SCM) that carries information on the variables of interest and their mechanistic relations.

Causal Inference

Generative Adversarial Neural Cellular Automata

no code implementations19 Jul 2021 Maximilian Otte, Quentin Delfosse, Johannes Czech, Kristian Kersting

Motivated by the interaction between cells, the recently introduced concept of Neural Cellular Automata shows promising results in a variety of tasks.

Leveraging Probabilistic Circuits for Nonparametric Multi-Output Regression

1 code implementation16 Jun 2021 Zhongjie Yu, Mingye Zhu, Martin Trapp, Arseny Skryagin, Kristian Kersting

Inspired by recent advances in the field of expert-based approximations of Gaussian processes (GPs), we present an expert-based approach to large-scale multi-output regression using single-output GP experts.

Gaussian Processes

RECOWNs: Probabilistic Circuits for Trustworthy Time Series Forecasting

no code implementations8 Jun 2021 Nils Thoma, Zhongjie Yu, Fabrizio Ventola, Kristian Kersting

Time series forecasting is a relevant task that is performed in several real-world scenarios such as product sales analysis and prediction of energy demand.

Time Series Forecasting

Can Linear Programs Have Adversarial Examples? A Causal Perspective

no code implementations26 May 2021 Matej Zečević, Devendra Singh Dhami, Kristian Kersting

The recent years have been marked by extended research on adversarial attacks, especially on deep neural networks.

Combinatorial Optimization

User-Level Label Leakage from Gradients in Federated Learning

2 code implementations19 May 2021 Aidmar Wainakh, Fabrizio Ventola, Till Müßig, Jens Keim, Carlos Garcia Cordero, Ephraim Zimmer, Tim Grube, Kristian Kersting, Max Mühlhäuser

Specifically, we investigate Label Leakage from Gradients (LLG), a novel attack to extract the labels of the users' training data from their shared gradients.

Federated Learning

Decomposing 3D Scenes into Objects via Unsupervised Volume Segmentation

no code implementations2 Apr 2021 Karl Stelzner, Kristian Kersting, Adam R. Kosiorek

We present ObSuRF, a method which turns a single image of a scene into a 3D model represented as a set of Neural Radiance Fields (NeRFs), with each NeRF corresponding to a different object.

Image Segmentation Semantic Segmentation

Large Pre-trained Language Models Contain Human-like Biases of What is Right and Wrong to Do

1 code implementation8 Mar 2021 Patrick Schramowski, Cigdem Turan, Nico Andersen, Constantin A. Rothkopf, Kristian Kersting

That is, we show that these norms can be captured geometrically by a direction, which can be computed, e. g., by a PCA, in the embedding space, reflecting well the agreement of phrases to social norms implicitly expressed in the training texts and providing a path for attenuating or even preventing toxic degeneration in LMs.

General Knowledge

Adaptive Rational Activations to Boost Deep Reinforcement Learning

3 code implementations18 Feb 2021 Quentin Delfosse, Patrick Schramowski, Martin Mundt, Alejandro Molina, Kristian Kersting

Latest insights from biology show that intelligence not only emerges from the connections between neurons but that individual neurons shoulder more computational responsibility than previously anticipated.

Ranked #3 on Atari Games on Atari 2600 Skiing (using extra training data)

Atari Games General Reinforcement Learning +2

Monte-Carlo Graph Search for AlphaZero

2 code implementations20 Dec 2020 Johannes Czech, Patrick Korus, Kristian Kersting

The AlphaZero algorithm has been successfully applied in a range of discrete domains, most notably board games.

Board Games

Right for the Right Concept: Revising Neuro-Symbolic Concepts by Interacting with their Explanations

3 code implementations CVPR 2021 Wolfgang Stammer, Patrick Schramowski, Kristian Kersting

Most explanation methods in deep learning map importance estimates for a model's prediction back to the original input space.

TUDataset: A collection of benchmark datasets for learning with graphs

2 code implementations16 Jul 2020 Christopher Morris, Nils M. Kriege, Franka Bause, Kristian Kersting, Petra Mutzel, Marion Neumann

We provide Python-based data loaders, kernel and graph neural network baseline implementations, and evaluation tools.

Graph Classification

Fitted Q-Learning for Relational Domains

no code implementations10 Jun 2020 Srijita Das, Sriraam Natarajan, Kaushik Roy, Ronald Parr, Kristian Kersting

We consider the problem of Approximate Dynamic Programming in relational domains.


Einsum Networks: Fast and Scalable Learning of Tractable Probabilistic Circuits

1 code implementation ICML 2020 Robert Peharz, Steven Lang, Antonio Vergari, Karl Stelzner, Alejandro Molina, Martin Trapp, Guy Van Den Broeck, Kristian Kersting, Zoubin Ghahramani

Probabilistic circuits (PCs) are a promising avenue for probabilistic modeling, as they permit a wide range of exact and efficient inference routines.

CryptoSPN: Privacy-preserving Sum-Product Network Inference

no code implementations3 Feb 2020 Amos Treiber, Alejandro Molina, Christian Weinert, Thomas Schneider, Kristian Kersting

AI algorithms, and machine learning (ML) techniques in particular, are increasingly important to individuals' lives, but have caused a range of privacy concerns addressed by, e. g., the European GDPR.

Privacy Preserving

Structured Object-Aware Physics Prediction for Video Modeling and Planning

1 code implementation ICLR 2020 Jannik Kossen, Karl Stelzner, Marcel Hussing, Claas Voelcker, Kristian Kersting

When humans observe a physical system, they can easily locate objects, understand their interactions, and anticipate future behavior, even in settings with complicated and previously unseen interactions.

Meta-Learning Runge-Kutta

no code implementations25 Sep 2019 Nadine Behrmann, Patrick Schramowski, Kristian Kersting

However, by studying the characteristics of the local error function we show that including the partial derivatives of the initial value problem is favorable.

Meta-Learning Numerical Integration

DeepDB: Learn from Data, not from Queries!

1 code implementation2 Sep 2019 Benjamin Hilprecht, Andreas Schmidt, Moritz Kulessa, Alejandro Molina, Kristian Kersting, Carsten Binnig

The typical approach for learned DBMS components is to capture the behavior by running a representative set of queries and use the observations to train a machine learning model.


Neural Networks for Relational Data

1 code implementation28 Aug 2019 Navdeep Kaur, Gautam Kunapuli, Saket Joshi, Kristian Kersting, Sriraam Natarajan

While deep networks have been enormously successful over the last decade, they rely on flat-feature vector representations, which makes them unsuitable for richly structured domains such as those arising in applications like social network analysis.

Learning to play the Chess Variant Crazyhouse above World Champion Level with Deep Neural Networks and Human Data

2 code implementations19 Aug 2019 Johannes Czech, Moritz Willig, Alena Beyer, Kristian Kersting, Johannes Fürnkranz

Crazyhouse is a game with a higher branching factor than chess and there is only limited data of lower quality available compared to AlphaGo.

Board Games

Random Sum-Product Forests with Residual Links

no code implementations8 Aug 2019 Fabrizio Ventola, Karl Stelzner, Alejandro Molina, Kristian Kersting

Tractable yet expressive density estimators are a key building block of probabilistic machine learning.

Padé Activation Units: End-to-end Learning of Flexible Activation Functions in Deep Networks

5 code implementations ICLR 2020 Alejandro Molina, Patrick Schramowski, Kristian Kersting

The performance of deep network learning strongly depends on the choice of the non-linear activation function associated with each neuron.

Declarative Learning-Based Programming as an Interface to AI Systems

no code implementations18 Jun 2019 Parisa Kordjamshidi, Dan Roth, Kristian Kersting

Data-driven approaches are becoming more common as problem-solving techniques in many areas of research and industry.

BIG-bench Machine Learning

Neural-Symbolic Argumentation Mining: an Argument in Favor of Deep Learning and Reasoning

no code implementations22 May 2019 Andrea Galassi, Kristian Kersting, Marco Lippi, Xiaoting Shao, Paolo Torroni

Deep learning is bringing remarkable contributions to the field of argumentation mining, but the existing approaches still need to fill the gap toward performing advanced reasoning tasks.

Component Classification Link Prediction +2

Conditional Sum-Product Networks: Imposing Structure on Deep Probabilistic Architectures

no code implementations21 May 2019 Xiaoting Shao, Alejandro Molina, Antonio Vergari, Karl Stelzner, Robert Peharz, Thomas Liebig, Kristian Kersting

In contrast, deep probabilistic models such as sum-product networks (SPNs) capture joint distributions in a tractable fashion, but still lack the expressive power of intractable models based on deep neural networks.

Image Classification

Was ist eine Professur fuer Kuenstliche Intelligenz?

no code implementations17 Feb 2019 Kristian Kersting, Jan Peters, Constantin Rothkopf

The Federal Government of Germany aims to boost the research in the field of Artificial Intelligence (AI).

SPFlow: An Easy and Extensible Library for Deep Probabilistic Learning using Sum-Product Networks

1 code implementation11 Jan 2019 Alejandro Molina, Antonio Vergari, Karl Stelzner, Robert Peharz, Pranav Subramani, Nicola Di Mauro, Pascal Poupart, Kristian Kersting

We introduce SPFlow, an open-source Python library providing a simple interface to inference, learning and manipulation routines for deep and tractable probabilistic models called Sum-Product Networks (SPNs).

Model-based Approximate Query Processing

no code implementations15 Nov 2018 Moritz Kulessa, Alejandro Molina, Carsten Binnig, Benjamin Hilprecht, Kristian Kersting

However, classical AQP approaches suffer from various problems that limit the applicability to support the ad-hoc exploration of a new data set: (1) Classical AQP approaches that perform online sampling can support ad-hoc exploration queries but yield low quality if executed over rare subpopulations.

Automatic Bayesian Density Analysis

no code implementations24 Jul 2018 Antonio Vergari, Alejandro Molina, Robert Peharz, Zoubin Ghahramani, Kristian Kersting, Isabel Valera

Classical approaches for {exploratory data analysis} are usually not flexible enough to deal with the uncertainty inherent to real-world data: they are often restricted to fixed latent interaction models and homogeneous likelihoods; they are sensitive to missing, corrupt and anomalous data; moreover, their expressiveness generally comes at the price of intractable inference.

Anomaly Detection Bayesian Inference +1

Probabilistic Deep Learning using Random Sum-Product Networks

no code implementations5 Jun 2018 Robert Peharz, Antonio Vergari, Karl Stelzner, Alejandro Molina, Martin Trapp, Kristian Kersting, Zoubin Ghahramani

The need for consistent treatment of uncertainty has recently triggered increased interest in probabilistic deep learning methods.

Probabilistic Deep Learning

Neural Conditional Gradients

no code implementations12 Mar 2018 Patrick Schramowski, Christian Bauckhage, Kristian Kersting

The move from hand-designed to learned optimizers in machine learning has been quite successful for gradient-based and -free optimizers.

Lifted Filtering via Exchangeable Decomposition

no code implementations31 Jan 2018 Stefan Lüdtke, Max Schröder, Sebastian Bader, Kristian Kersting, Thomas Kirste

We present a model for exact recursive Bayesian filtering based on lifted multiset states.

Coresets for Dependency Networks

no code implementations9 Oct 2017 Alejandro Molina, Alexander Munteanu, Kristian Kersting

Many applications infer the structure of a probabilistic graphical model from data to elucidate the relationships between variables.

Sum-Product Networks for Hybrid Domains

no code implementations9 Oct 2017 Alejandro Molina, Antonio Vergari, Nicola Di Mauro, Sriraam Natarajan, Floriana Esposito, Kristian Kersting

While all kinds of mixed data -from personal data, over panel and scientific data, to public and commercial data- are collected and stored, building probabilistic graphical models for these hybrid domains becomes more difficult.

Global Weisfeiler-Lehman Graph Kernels

no code implementations7 Mar 2017 Christopher Morris, Kristian Kersting, Petra Mutzel

Specifically, we introduce a novel graph kernel based on the $k$-dimensional Weisfeiler-Lehman algorithm.

General Classification Graph Classification

A Unifying View of Explicit and Implicit Feature Maps of Graph Kernels

no code implementations2 Mar 2017 Nils M. Kriege, Marion Neumann, Christopher Morris, Kristian Kersting, Petra Mutzel

On this basis we propose exact and approximative feature maps for widely used graph kernels based on the kernel trick.

Faster Kernels for Graphs with Continuous Attributes via Hashing

no code implementations1 Oct 2016 Christopher Morris, Nils M. Kriege, Kristian Kersting, Petra Mutzel

While state-of-the-art kernels for graphs with discrete labels scale well to graphs with thousands of nodes, the few existing kernels for graphs with continuous attributes, unfortunately, do not scale well.

Lifted Convex Quadratic Programming

no code implementations14 Jun 2016 Martin Mladenov, Leonard Kleinhans, Kristian Kersting

Symmetry is the essential element of lifted inference that has recently demon- strated the possibility to perform very efficient inference in highly-connected, but symmetric probabilistic models models.

How is a data-driven approach better than random choice in label space division for multi-label classification?

no code implementations7 Jun 2016 Piotr Szymański, Tomasz Kajdanowicz, Kristian Kersting

We show that fastgreedy and walktrap community detection methods on weighted label co-occurence graphs are 85-92% more likely to yield better F1 scores than random partitioning.

Community Detection General Classification +1

The Symbolic Interior Point Method

no code implementations26 May 2016 Martin Mladenov, Vaishak Belle, Kristian Kersting

A recent trend in probabilistic inference emphasizes the codification of models in a formal syntax, with suitable high-level features such as individuals, relations, and connectives, enabling descriptive clarity, succinctness and circumventing the need for the modeler to engineer a custom solver.

Decision Making

Propagation Kernels

1 code implementation13 Oct 2014 Marion Neumann, Roman Garnett, Christian Bauckhage, Kristian Kersting

We introduce propagation kernels, a general graph-kernel framework for efficiently measuring the similarity of structured data.

Relational Linear Programs

no code implementations12 Oct 2014 Kristian Kersting, Martin Mladenov, Pavel Tokmakov

A relational linear program (RLP) is a declarative LP template defining the objective and the constraints through the logical concepts of objects, relations, and quantified variables.

Mind the Nuisance: Gaussian Process Classification using Privileged Noise

no code implementations NeurIPS 2014 Daniel Hernández-Lobato, Viktoriia Sharmanska, Kristian Kersting, Christoph H. Lampert, Novi Quadrianto

That is, in contrast to the standard GPC setting, the latent function is not just a nuisance but a feature: it becomes a natural measure of confidence about the training data by modulating the slope of the GPC sigmoid likelihood function.

Classification General Classification

Dimension Reduction via Colour Refinement

no code implementations22 Jul 2013 Martin Grohe, Kristian Kersting, Martin Mladenov, Erkal Selman

We demonstrate empirically that colour refinement can indeed greatly reduce the cost of solving linear programs.

Dimensionality Reduction

Symbolic Dynamic Programming for Continuous State and Observation POMDPs

no code implementations NeurIPS 2012 Zahra Zamani, Scott Sanner, Pascal Poupart, Kristian Kersting

In recent years, point- based value iteration methods have proven to be extremely effective techniques for finding (approximately) optimal dynamic programming solutions to POMDPs when an initial set of belief states is known.

Decision Making

Bayesian Logic Programs

no code implementations23 Nov 2001 Kristian Kersting, Luc De Raedt

Theyare a probabilistic extension of propositional logic and, hence, inherit some of the limitations of propositional logic, such as the difficulties to represent objects and relations.

Cannot find the paper you are looking for? You can Submit a new open access paper.