Search our Publications

Results

Fabio Muratore, Christian Eilers, Michael Gienger, Jan Peters, "Data-efficient Domain Randomization with Bayesian Optimization", IEEE Robotics and Automation Letters (RA-L), 2021.

Abstract

When learning policies for robot control, the required real-world data is typically prohibitively expensive to acquire, so learning in simulation is a popular strategy. Unfortunately, such polices are often not transferable to the real world due to a mismatch between the simulation and reality, called ‘reality gap’. Domain randomization methods tackle this problem by randomizing the physics simulator (source domain) during training according ...



Download Bibtex file Download PDF

Duc Anh Nguyen, Anna Kononova, Stefan Menzel, Bernhard Sendhoff, Thomas Bäck, "Efficient AutoML via Combinational Sampling", IEEE Symposium Series on Computational Intelligence, 2021.

Abstract

Automated machine learning (AutoML) aims to automatically produce the best machine learning pipeline, i.e., a sequence of operators and their optimized hyperparameter settings, to maximize the performance of an arbitrary machine learning problem. Typically, AutoML based Bayesian optimization (BO) approaches convert the AutoML optimization problem into a Hyperparameter Optimization (HPO) problem, where the choice of algorithms is modeled as an add...



Download Bibtex file Per Mail Request

Duc Anh Nguyen, Jiawen Kong, Hao Wang, Stefan Menzel, Bernhard Sendhoff, Anna Kononova, Thomas Bäck, "Improved Automated CASH Optimization with Tree Parzen Estimators for Class Imbalance Problems", IEEE International Conference on Data Science and Advanced Analytics (DSAA), 2021.

Abstract

The imbalanced classification problem is very relevant in both academic and industrial applications. The task of finding the best machine learning model to use for a specific imbalanced dataset is complicated due to the large number of existing algorithms and their hyperparameters. The Combined Algorithm Selection and Hyperparameter optimization (CASH) was introduced to tackle both aspects at the same time. However, CASH has not been studied in d...



Download Bibtex file Per Mail Request

Theodoros Stouraitis, "A Dyadic Collaborative Manipulation Formalism for Optimizing Human-Robot Teaming", Edinburgh University, 2021.

Abstract

Dyadic collaborative Manipulation (DcM) is a term we use to refer to a team of two individuals, the agent and the partner, jointly manipulating an object. The two indi- viduals partner together to form a distributed system, augmenting their manipula- tion abilities. Effective collaboration between the two individuals during joint action depends on: (i) the breadth of the agent’s action repertoire, (ii) the level of model ac- quaintance bet...



Download Bibtex file Per Mail Request

Andrea Castellani, Sebastian Schmitt, Barbara Hammer, "Task-Sensitive Concept Drift Detector with Constraint Embedding", IEEE Symposium Series on Computational Intelligence (SSCI), 2021.

Abstract

Detecting drifts in data is essential for machine learning applications, as changes in the statistics of processed data typically has a profound influence on the performance of trained models. Most of the available drift detection methods are either supervised and require access to the true labels during inference time, or they are completely unsupervised and aim for changes in distributions without taking label information into account. W...



Download Bibtex file Download PDF

Sneha Saha, Thiago Rios, Leandro Minku, Bas van Stein, Patricia Wollstadt, Xin Yao, Thomas Bäck, Bernhard Sendhoff, Stefan Menzel, "Exploiting Generative Models for Performance Predictions of 3D Car Designs", IEEE Symposium Series on Computational Intelligence (IEEE SSCI 2021), 2021.

Abstract

In automotive digital development, engineers utilize multiple virtual prototyping tools to design and assess the performance of 3D shapes. However, accurate performance simulations are computationally expensive and time-consuming, which may be prohibitive for design optimization tasks. To address this challenge, we envision a 3D design assistance system for design exploration with performance assessment in the automotive domain. Recent advances...



Download Bibtex file Per Mail Request

Nikolas Hohmann, Mariusz Bujny, Jürgen Adamy, Markus Olhofer, "Hybrid Evolutionary Approach to Multi-objective Path Planning for UAVs", IEEE Symposium Series on Computational Intelligence, 2021.

Abstract

The goal of Multi-Objective Path Planning (MOPP) is to find Pareto-optimal paths for autonomous agents with respect to several optimization goals like minimizing risk, path length, travel time, or energy consumption. In this work, we formulate a MOPP for Unmanned Aerial Vehicles (UAVs). We utilize a path representation based on Non-Uniform Rational B-Splines (NURBS) and propose a hybrid evolutionary approach combining an Evolution Strategy (ES) w...



Download Bibtex file Per Mail Request

Thiago Rios, Bas van Stein, Thomas Bäck, Bernhard Sendhoff, Stefan Menzel, "Point2FFD: Learning Shape Representations of Simulation-ready 3D Models for Engineering Design Optimization", International Conference on 3D Vision, 2021.

Abstract

Methods for learning on 3D point clouds became ubiquitous due to the popularization of 3D scanning technology and advances of machine learning techniques. Among these methods, point-based deep neural networks have been utilized to explore 3D designs in optimization tasks. However, engineering computer simulations require high-quality meshed models, which are challenging to automatically generate from unordered point clouds. In this work, we propo...



Download Bibtex file Download PDF

Patricia Wollstadt, Martina Hasenjäger, Christiane Wiebel, "Quantifying Predictability of Scan Path Data using Active Information Storage", Entropy, vol. 23, no. 2, pp. 167, 2021.

Abstract

Entropy-based measures are an important tool to study human gaze behavior under various conditions. Measures, such as the gaze transition entropy, are used to quantify the predictability of transitions between consecutive fixations. However, these measures do not account for temporal dependencies beyond interactions of order one. Therefore, we propose a novel approach to quantifying predictability in scanpath data by estimating the active informa...



Download Bibtex file Download PDF

Fabio Muratore, Theo Gruner, Florian Wiese, Boris Belousov, Michael Gienger, Jan Peters, "Neural Posterior Domain Randomization", Conference on Robbot Learning (CorL), 2021.

Abstract

Combining domain randomization and reinforcement learning is a widely used approach to train control policies that can bridge the gap between simulation and reality. However, existing methods make avoidable assumptions. Typically, one type of probability distribution (e.g., normal or uniform) is chosen a beforehand for every domain parameter. Another common assumption is the differentiability of the simulator. These design decisions simplif...



Download Bibtex file Download PDF

1 2 3 4 5 ... 130

Search