A recent paradigm called Regeneration Learning addresses generative problems where the target data (e.g., images) is more complex than the available input source. While current cross-modal representation and regeneration learning rely on supervised deep learning models, this paper aims to revisit the adequacy of unsupervised models in this field. In this regard, we propose a new unsupervised approach that utilizes the SOM as a heteroassociative memory model to learn cross-modal representations in a topologically coherent map. This approach enables bidirectional predictive/regenerative mapping between domains. We evaluate the potential of this method on an unsolved (so far!) practical problem in petroleum geoscience.

In remote data collection from sampling stations, a vehicle must be within sufficient distance from the particular station for a predefined minimal time to retrieve all the required data from the site.

The planning task is to find a cost-efficient data collection trajectory, allowing the data collection vehicle to retrieve data from all sensing sites.

Having a fixed-wing aerial vehicle flying with a constant forward velocity, the problem is to determine the shortest feasible path that visits every sensing site and the vehicle is within a reliable communication distance from the station in a sufficient period.

We propose to formulate the planning problem as a~variant of the Close Enough Dubins Traveling Salesman Problem with Time Constraints (CEDTSP-TC) that is heuristically solved by unsupervised learning of the Growing Self-Organizing Array (GSOA) modified to address the minimal required time for the vehicle to be within the communication range of the station.

The proposed method is compared with a baseline based on a sampling-based decoupled approach.

The presented results support the feasibility of both proposed solvers on random instances and show that the GSOA-based approach outperforms the decoupled approach or provides similar results.

In this paper we propose the application of min-max-neurons for the use in generalized learning vector quantization (GLVQ) models, which correspond to min-max-prototypes. These prototypes can be identified with hyperboxes in the data space. Keeping the general GLVQ cost function, we redefine the Hebb-responsibilities for min-max-prototypes and derive consistent learning rules for stochastic gradient descent learning. We demonstrate that the resulting hyperbox-based GLVQ is capable to solve several classification tasks in robust manner, which can be dedicated to the use of robust min-max-prototypes. Finally, we give suggestions for future research for GLVQ based on min-max-prototypes.

While high dimensionality and the selection of meaningful features is usually a burden in machine learning, it is even more so in the case of unsupervised learning and particularly in clustering. The presence of uninformative features, sometimes correlated, may bias significantly the results of distance-based methods such as k-means for instance. Since the seminal work of Witten et al. (2010), different versions of sparse k-means have been introduced, building on the idea of adding some penalty terms in the loss function and resulting into automatic feature selection and/or weighting. This paper investigates the connections between some of these methods, and particularly the differences induced by the choices of the penalty terms. It also focuses on the case of mixed data, and how they may be handled by the sparse k-means approaches. Eventually, it presents the algorithms and model selection tools made available through a recently implemented R package, vimpclust.

Self-organizing maps (SOMs) have many advantages as a tool for exploratory data analysis. Combining vector quantization and topological relationships that are defined in a low-dimensional space, they can run on big data sets and are mostly immune to the curse of dimensionality in the data space.

SOMs are used mainly for dimensionality reduction and marginally for clustering; however, SOMs also suffer from some shortcomings.

Vector quantization makes them unable to embed all data points, only prototypes or centroid are mapped.

Being defined as a regular grid in the low-dimensional space, dimensionality reduction and clustering with SOMs are indirect, as compared to methods of direct embedding like multi-dimensional scaling.

Since 2008, t-SNE (t-distributed stochastic neighbor embedding) has raised growing interest, first in the machine learning community and now outside of it, with many applications in cell biology, for instance.

Primarily used as a 2D embedding and visualization technique, t-SNE is more and more used as a clustering technique, which is capable of identifying meaningful clusters that classical clustering tools struggle to see.

Quite counter-intuitively, t-SNE often better separates clusters in low-dimensional embeddings than clustering tools would do in the high-dimensional data space.

To understand this paradox, several mechanisms of t-SNE can be framed as a distance transformation with the possibility (i) to denoise distances in high-dimensional spaces and (ii) to impose a strong inductive bias on them, which magnifies inter-cluster gaps.

Despite these strengths, t-SNE is not free of drawbacks, which we quickly review to sketch perspectives of future developments.

There is a long history of formulation of machine learning models that have at their core "representative prototypes" living in the data space (examples include variants of Self Organizing Maps in the unsupervised setting or variants of Learning Vector Quantization in the supervised one). The prototype based formulations make such models intuitive and naturally amenable to interpretation. I will present a series of generalizations and extensions of this idea, arguing that prototype based learning still deserves attention in the current era of deep learning.

In particular, I will present extensions of prototype based learning to model spaces and Riemannian manifolds. I will also show how prototype based models can be naturally extended to the setting of ordinal regression and learning with privileged information.

Variable importance determination refers to the challenge to identify the most relevant input dimensions or features for a given learning task and quantify their relevance, either with respect to a local decision or a global model. Feature relevance determination constitutes a foundation for feature selection, and it enables an intuitive insight into the rational of model decisions. Indeed, it constitutes one of the oldest and most prominent explanation technologies for machine learning models with relevance for both, deep and shallow networks. A huge number of measures have been proposed such as mutual information, permutation feature importance, deep lift, LIME, GMLVQ, or Shapley values, to name just a few.

Within the talk, I will address recent extensions of feature relevance determination, which occur as machine learning models are increasingly used in everyday life. Here, models face an open environment, possibly changing dynamics, and the necessity of model adaptation to account for changes of the underlying distribution. At present, feature relevance determination almost solely focusses on static scenarios and batch training. In the talk, I will target the question of how to efficiently and effectively accompany a model which learns incrementally by feature relevance determination methods(1,2). As a second challenge, features are often not mutually independent, and the relevance of groups rather than single features should be judged. While mathematical models such as Shapley values take feature correlations into account for individual additive feature relevance terms, it is unclear how to efficiently and effectively extend those to groups of features. In the talk, I will discuss novel methods for the efficient computation of feature interaction indices (3,4).

(1) F Fumagalli, M Muschalik, E Hüllermeier, B Hammer: Incremental Permutation Feature Importance (iPFI): Towards Online Explanations on Data Streams. Mach. Learn. 112(12): 4863-4903, 2023.

(2) M Muschalik, F Fumagalli, B Hammer, E Hüllermeier: iSAGE: An Incremental Version of SAGE for Online Explanation on Data Streams. ECML/PKDD (3) 2023: 428-445.(3) F Fumagalli, M Muschalik, P Kolpaczki, E Hüllermeier, B Hammer: SHAP-IQ: Unified Approximation of Any-Order Shapley Interactions. NeurIPS 2023.(4) M Muschalik, F Fumagalli, B Hammer, E Hüllermeier: Beyond TreeSHAP: Efficient Computation of Any-Order Shapley Interactions for Tree Ensembles AAAI 2024.