Felix Mohr

Adrian El Baz, Ihsan Ullah, Edesio Alcobaça et al.

Although deep neural networks are capable of achieving performance superior to humans on various tasks, they are notorious for requiring large amounts of data and computing resources, restricting their success to domains where such resources are available. Metalearning methods can address this problem by transferring knowledge from related tasks, thus reducing the amount of data and computing resources needed to learn new tasks. We organize the MetaDL competition series, which provide opportunities for research groups all over the world to create and experimentally assess new meta-(deep)learning solutions for real problems. In this paper, authored collaboratively between the competition organizers and the top-ranked participants, we describe the design of the competition, the datasets, the best experimental results, as well as the top-ranked methods in the NeurIPS 2021 challenge, which attracted 15 active teams who made it to the final phase (by outperforming the baseline), making over 100 code submissions during the feedback phase. The solutions of the top participants have been open-sourced. The lessons learned include that learning good representations is essential for effective transfer learning.

Ihsan Ullah, Dustin Carrión-Ojeda, Sergio Escalera et al.

We introduce Meta-Album, an image classification meta-dataset designed to facilitate few-shot learning, transfer learning, meta-learning, among other tasks. It includes 40 open datasets, each having at least 20 classes with 40 examples per class, with verified licences. They stem from diverse domains, such as ecology (fauna and flora), manufacturing (textures, vehicles), human actions, and optical character recognition, featuring various image scales (microscopic, human scales, remote sensing). All datasets are preprocessed, annotated, and formatted uniformly, and come in 3 versions (Micro $\subset$ Mini $\subset$ Extended) to match users' computational resources. We showcase the utility of the first 30 datasets on few-shot learning problems. The other 10 will be released shortly after. Meta-Album is already more diverse and larger (in number of datasets) than similar efforts, and we are committed to keep enlarging it via a series of competitions. As competitions terminate, their test data are released, thus creating a rolling benchmark, available through OpenML.org. Our website https://meta-album.github.io/ contains the source code of challenge winning methods, baseline methods, data loaders, and instructions for contributing either new datasets or algorithms to our expandable meta-dataset.

Haozhe Sun, Isabelle Guyon, Felix Mohr et al.

It has become mainstream in computer vision and other machine learning domains to reuse backbone networks pre-trained on large datasets as preprocessors. Typically, the last layer is replaced by a shallow learning machine of sorts; the newly-added classification head and (optionally) deeper layers are fine-tuned on a new task. Due to its strong performance and simplicity, a common pre-trained backbone network is ResNet152.However, ResNet152 is relatively large and induces inference latency. In many cases, a compact and efficient backbone with similar performance would be preferable over a larger, slower one. This paper investigates techniques to reuse a pre-trained backbone with the objective of creating a smaller and faster model. Starting from a large ResNet152 backbone pre-trained on ImageNet, we first reduce it from 51 blocks to 5 blocks, reducing its number of parameters and FLOPs by more than 6 times, without significant performance degradation. Then, we split the model after 3 blocks into several branches, while preserving the same number of parameters and FLOPs, to create an ensemble of sub-networks to improve performance. Our experiments on a large benchmark of $40$ image classification datasets from various domains suggest that our techniques match the performance (if not better) of ``classical backbone fine-tuning'' while achieving a smaller model size and faster inference speed.

Tanja Tornede, Alexander Tornede, Lukas Fehring et al.

PyExperimenter is a tool to facilitate the setup, documentation, execution, and subsequent evaluation of results from an empirical study of algorithms and in particular is designed to reduce the involved manual effort significantly. It is intended to be used by researchers in the field of artificial intelligence, but is not limited to those.

Romain Egele, Felix Mohr, Tom Viering et al.

To reach high performance with deep learning, hyperparameter optimization (HPO) is essential. This process is usually time-consuming due to costly evaluations of neural networks. Early discarding techniques limit the resources granted to unpromising candidates by observing the empirical learning curves and canceling neural network training as soon as the lack of competitiveness of a candidate becomes evident. Despite two decades of research, little is understood about the trade-off between the aggressiveness of discarding and the loss of predictive performance. Our paper studies this trade-off for several commonly used discarding techniques such as successive halving and learning curve extrapolation. Our surprising finding is that these commonly used techniques offer minimal to no added value compared to the simple strategy of discarding after a constant number of epochs of training. The chosen number of epochs depends mostly on the available compute budget. We call this approach i-Epoch (i being the constant number of epochs with which neural networks are trained) and suggest to assess the quality of early discarding techniques by comparing how their Pareto-Front (in consumed training epochs and predictive performance) complement the Pareto-Front of i-Epoch.

Till Aust, Eduard Buss, Felix Mohr et al.

In our project WatchPlant, we propose to use a decentralized network of living plants as air-quality sensors by measuring their electrophysiology to infer the environmental state, also called phytosensing. We conducted in-lab experiments exposing ivy (Hedera helix) plants to ozone, an important pollutant to monitor, and measured their electrophysiological response. However, there is no well established automated way of detecting ozone exposure in plants. We propose a generic automatic toolchain to select a high-performance subset of features and highly accurate models for plant electrophysiology. Our approach derives plant- and stimulus-generic features from the electrophysiological signal using the tsfresh library. Based on these features, we automatically select and optimize machine learning models using AutoML. We use forward feature selection to increase model performance. We show that our approach successfully classifies plant ozone exposure with accuracies of up to 94.6% on unseen data. We also show that our approach can be used for other plant species and stimuli. Our toolchain automates the development of monitoring algorithms for plants as pollutant monitors. Our results help implement significant advancements for phytosensing devices contributing to the development of cost-effective, high-density urban air monitoring systems in the future.

Timo Löhr, Paul Hofman, Felix Mohr et al.

Predictions in the form of sets of probability distributions, so-called credal sets, provide a suitable means to represent a learner's epistemic uncertainty. In this paper, we propose a theoretically grounded approach to credal prediction based on the statistical notion of relative likelihood: The target of prediction is the set of all (conditional) probability distributions produced by the collection of plausible models, namely those models whose relative likelihood exceeds a specified threshold. This threshold has an intuitive interpretation and allows for controlling the trade-off between correctness and precision of credal predictions. We tackle the problem of approximating credal sets defined in this way by means of suitably modified ensemble learning techniques. To validate our approach, we illustrate its effectiveness by experiments on benchmark datasets demonstrating superior uncertainty representation without compromising predictive performance. We also compare our method against several state-of-the-art baselines in credal prediction.

Yanick Chistian Tchenko, Felix Mohr, Hicham Hadj Abdelkader et al.

A prevailing trend in neural network research suggests that model performance improves with increasing depth and capacity - often at the cost of integrability and efficiency. In this paper, we propose a strategy to optimize small, deployable models by enhancing their capabilities through structured knowledge inheritance. We introduce Hereditary Knowledge Transfer (HKT), a biologically inspired framework for modular and selective transfer of task-relevant features from a larger, pretrained parent network to a smaller child model. Unlike standard knowledge distillation, which enforces uniform imitation of teacher outputs, HKT draws inspiration from biological inheritance mechanisms - such as memory RNA transfer in planarians - to guide a multi-stage process of feature transfer. Neural network blocks are treated as functional carriers, and knowledge is transmitted through three biologically motivated components: Extraction, Transfer, and Mixture (ETM). A novel Genetic Attention (GA) mechanism governs the integration of inherited and native representations, ensuring both alignment and selectivity. We evaluate HKT across diverse vision tasks, including optical flow (Sintel, KITTI), image classification (CIFAR-10), and semantic segmentation (LiTS), demonstrating that it significantly improves child model performance while preserving its compactness. The results show that HKT consistently outperforms conventional distillation approaches, offering a general-purpose, interpretable, and scalable solution for deploying high-performance neural networks in resource-constrained environments.

Cheng Yan, Felix Mohr, Tom Viering

Sample-wise learning curves plot performance versus training set size. They are useful for studying scaling laws and speeding up hyperparameter tuning and model selection. Learning curves are often assumed to be well-behaved: monotone (i.e. improving with more data) and convex. By constructing the Learning Curves Database 1.1 (LCDB 1.1), a large-scale database with high-resolution learning curves including more modern learners (CatBoost, TabNet, RealMLP and TabPFN), we show that learning curves are less often well-behaved than previously thought. Using statistically rigorous methods, we observe significant ill-behavior in approximately 15% of the learning curves, almost twice as much as in previous estimates. We also identify which learners are to blame and show that specific learners are more ill-behaved than others. Additionally, we demonstrate that different feature scalings rarely resolve ill-behavior. We evaluate the impact of ill-behavior on downstream tasks, such as learning curve fitting and model selection, and find it poses significant challenges, underscoring the relevance and potential of LCDB 1.1 as a challenging benchmark for future research.

Felix Mohr, Jan N. van Rijn

Learning curves are a concept from social sciences that has been adopted in the context of machine learning to assess the performance of a learning algorithm with respect to a certain resource, e.g., the number of training examples or the number of training iterations. Learning curves have important applications in several machine learning contexts, most notably in data acquisition, early stopping of model training, and model selection. For instance, learning curves can be used to model the performance of the combination of an algorithm and its hyperparameter configuration, providing insights into their potential suitability at an early stage and often expediting the algorithm selection process. Various learning curve models have been proposed to use learning curves for decision making. Some of these models answer the binary decision question of whether a given algorithm at a certain budget will outperform a certain reference performance, whereas more complex models predict the entire learning curve of an algorithm. We contribute a framework that categorises learning curve approaches using three criteria: the decision-making situation they address, the intrinsic learning curve question they answer and the type of resources they use. We survey papers from the literature and classify them into this framework.

Felix Mohr, Marcel Wever

An essential task of Automated Machine Learning (AutoML) is the problem of automatically finding the pipeline with the best generalization performance on a given dataset. This problem has been addressed with sophisticated black-box optimization techniques such as Bayesian Optimization, Grammar-Based Genetic Algorithms, and tree search algorithms. Most of the current approaches are motivated by the assumption that optimizing the components of a pipeline in isolation may yield sub-optimal results. We present Naive AutoML, an approach that does precisely this: It optimizes the different algorithms of a pre-defined pipeline scheme in isolation. The finally returned pipeline is obtained by just taking the best algorithm of each slot. The isolated optimization leads to substantially reduced search spaces, and, surprisingly, this approach yields comparable and sometimes even better performance than current state-of-the-art optimizers.

Felix Mohr, Jan N. van Rijn

Common cross-validation (CV) methods like k-fold cross-validation or Monte-Carlo cross-validation estimate the predictive performance of a learner by repeatedly training it on a large portion of the given data and testing on the remaining data. These techniques have two major drawbacks. First, they can be unnecessarily slow on large datasets. Second, beyond an estimation of the final performance, they give almost no insights into the learning process of the validated algorithm. In this paper, we present a new approach for validation based on learning curves (LCCV). Instead of creating train-test splits with a large portion of training data, LCCV iteratively increases the number of instances used for training. In the context of model selection, it discards models that are very unlikely to become competitive. We run a large scale experiment on the 67 datasets from the AutoML benchmark and empirically show that in over 90% of the cases using LCCV leads to similar performance (at most 1.5% difference) as using 5/10-fold CV. However, it yields substantial runtime reductions of over 20% on average. Additionally, it provides important insights, which for example allow assessing the benefits of acquiring more data. These results are orthogonal to other advances in the field of AutoML.

Tanja Tornede, Alexander Tornede, Jonas Hanselle et al.

Automated machine learning (AutoML) strives for the automatic configuration of machine learning algorithms and their composition into an overall (software) solution - a machine learning pipeline - tailored to the learning task (dataset) at hand. Over the last decade, AutoML has developed into an independent research field with hundreds of contributions. At the same time, AutoML is being criticised for its high resource consumption as many approaches rely on the (costly) evaluation of many machine learning pipelines, as well as the expensive large scale experiments across many datasets and approaches. In the spirit of recent work on Green AI, this paper proposes Green AutoML, a paradigm to make the whole AutoML process more environmentally friendly. Therefore, we first elaborate on how to quantify the environmental footprint of an AutoML tool. Afterward, different strategies on how to design and benchmark an AutoML tool wrt. their "greenness", i.e. sustainability, are summarized. Finally, we elaborate on how to be transparent about the environmental footprint and what kind of research incentives could direct the community into a more sustainable AutoML research direction. Additionally, we propose a sustainability checklist to be attached to every AutoML paper featuring all core aspects of Green AutoML.

Eyke Hüllermeier, Felix Mohr, Alexander Tornede et al.

The notion of bounded rationality originated from the insight that perfectly rational behavior cannot be realized by agents with limited cognitive or computational resources. Research on bounded rationality, mainly initiated by Herbert Simon, has a longstanding tradition in economics and the social sciences, but also plays a major role in modern AI and intelligent agent design. Taking actions under bounded resources requires an agent to reflect on how to use these resources in an optimal way - hence, to reason and make decisions on a meta-level. In this paper, we will look at automated machine learning (AutoML) and related problems from the perspective of bounded rationality, essentially viewing an AutoML tool as an agent that has to train a model on a given set of data, and the search for a good way of doing so (a suitable "ML pipeline") as deliberation on a meta-level.

Felix Mohr, Marcel Wever

Automated Machine Learning (AutoML) is the problem of automatically finding the pipeline with the best generalization performance on some given dataset. AutoML has received enormous attention in the last decade and has been addressed with sophisticated black-box optimization techniques such as Bayesian Optimization, Grammar-Based Genetic Algorithms, and tree search algorithms. In contrast to those approaches, we present Naive AutoML, a very simple solution to AutoML that exploits important meta-knowledge about machine learning problems and makes simplifying, yet, effective assumptions to quickly come to high-quality solutions. While Naive AutoML can be considered a baseline for the highly sophisticated black-box solvers, we empirically show that those solvers are not able to outperform Naive AutoML; sometimes the contrary is true. On the other hand, Naive AutoML comes with strong advantages such as interpretability and flexibility and poses a strong challenge to current tools.

Felix Mohr, Gonzalo Mejía, Francisco Yuraszeck

In this paper we study a simple extension of the total weighted flowtime minimization problem for single and identical parallel machines. While the standard problem simply defines a set of jobs with their processing times and weights and assumes that all jobs have release date 0 and have no deadline, we assume that the release date of each job is a decision variable that is only constrained by a single global latest arrival deadline. To our knowledge, this simple yet practically highly relevant extension has never been studied. Our main contribution is that we show the NP- completeness of the problem even for the single machine case and provide an exhaustive empirical study of different typical approaches including genetic algorithms, tree search, and constraint programming.

Alexander Tornede, Marcel Wever, Stefan Werner et al.

Algorithm selection (AS) deals with the automatic selection of an algorithm from a fixed set of candidate algorithms most suitable for a specific instance of an algorithmic problem class, where "suitability" often refers to an algorithm's runtime. Due to possibly extremely long runtimes of candidate algorithms, training data for algorithm selection models is usually generated under time constraints in the sense that not all algorithms are run to completion on all instances. Thus, training data usually comprises censored information, as the true runtime of algorithms timed out remains unknown. However, many standard AS approaches are not able to handle such information in a proper way. On the other side, survival analysis (SA) naturally supports censored data and offers appropriate ways to use such data for learning distributional models of algorithm runtime, as we demonstrate in this work. We leverage such models as a basis of a sophisticated decision-theoretic approach to algorithm selection, which we dub Run2Survive. Moreover, taking advantage of a framework of this kind, we advocate a risk-averse approach to algorithm selection, in which the avoidance of a timeout is given high priority. In an extensive experimental study with the standard benchmark ASlib, our approach is shown to be highly competitive and in many cases even superior to state-of-the-art AS approaches.

Marcel Wever, Felix Mohr, Eyke Hüllermeier

Automated machine learning (AutoML) has received increasing attention in the recent past. While the main tools for AutoML, such as Auto-WEKA, TPOT, and auto-sklearn, mainly deal with single-label classification and regression, there is very little work on other types of machine learning tasks. In particular, there is almost no work on automating the engineering of machine learning applications for multi-label classification. This paper makes two contributions. First, it discusses the usefulness and feasibility of an AutoML approach for multi-label classification. Second, we show how the scope of ML-Plan, an AutoML-tool for multi-class classification, can be extended towards multi-label classification using MEKA, which is a multi-label extension of the well-known Java library WEKA. The resulting approach recursively refines MEKA's multi-label classifiers, which sometimes nest another multi-label classifier, up to the selection of a single-label base learner provided by WEKA. In our evaluation, we find that the proposed approach yields superb results and performs significantly better than a set of baselines.

Felix Mohr, Marcel Wever, Eyke Hüllermeier

Automated service composition as the process of creating new software in an automated fashion has been studied in many different ways over the last decade. However, the impact of automated service composition has been rather small as its utility in real-world applications has not been demonstrated so far. This paper presents \tool, an algorithm for automated service composition applied to the area of machine learning. Empirically, we show that \tool is competitive and sometimes beats algorithms that solve the same task but not benefit of the advantages of a service model. Thereby, we present a real-world example that demonstrates the utility of automated service composition in contrast to non-service oriented solutions in the same area.