Steffen Illium

Jekyll2025-07-22T08:56:15+02:00http://localhost:4000/feed.xmlSteffen IlliumPersonal WebsiteSteffen IlliumMAS Emergence Safety2024-10-27T00:00:00+02:002024-10-27T00:00:00+02:00http://localhost:4000/research/mas-emergence-safety Visualization showing agents exhibiting emergent coin-collecting behavior

Visualization showing agents exhibiting emergent coin-collecting behavior

Example of emergent behavior (e.g., coin hoarding) due to specification misalignment.

Leveraging established concepts from system safety engineering, we analyze how such misalignments can lead to deviations from intended global behavior. To illustrate the practical implications, we examine two highly configurable gridworld scenarios. These demonstrate how inadequate or independently derived local specifications (rewards/observations) can predictably result in unintended emergent behaviors, such as resource hoarding or inefficient coordination. Visualization showing agents exhibiting emergent blocking behavior

Visualization showing agents exhibiting emergent blocking behavior

Example of emergent behavior (e.g., mutual blocking) due to specification misalignment.

Recognizing that achieving a perfectly aligned global specification might be impractical in complex systems, we propose strategies focused on **adjusting the underlying local parameterizations** (e.g., reward shaping, observation design) to mitigate harmful emergence. By carefully tuning these local components, system alignment can be improved, reducing the risk of emergent failures and enhancing overall safety. {% cite altmann2024emergence %}]]>Steffen Illium

Aquarium MARL Environment2024-01-13T00:00:00+01:002024-01-13T00:00:00+01:00http://localhost:4000/research/aquarium-marl-environment Diagram detailing the construction of the observation vector for an agent

Diagram detailing the construction of the observation vector for an agent

Construction details of the agent observation vector.

Graphs showing average captures or rewards per prey agent under different training regimes

Performance metrics (e.g., average captures/rewards) comparing training strategies.

To demonstrate its capabilities, we conducted preliminary studies using **Proximal Policy Optimization (PPO)** to train multiple prey agents learning to evade a predator within Aquarium. Consistent with findings in existing MARL literature, our results showed that training agents with **individual policies led to suboptimal performance**, whereas utilizing **parameter sharing** among prey agents significantly improved coordination, sample efficiency, and overall evasion success. {% cite kolle2024aquarium %}]]>Steffen Illium

LMU DevOps Admin2023-10-15T00:00:00+02:002023-10-15T00:00:00+02:00http://localhost:4000/projects/server-administration **Affiliation:** Chair for Mobile and Distributed Systems, LMU Munich
**Duration:** 2018 - 2023 (Concurrent with Research Role)
**Objective:** Continious maintenance of IT infrastructure --- **Key Initiatives & Achievements:** * **Infrastructure as Code & Orchestration:** * Leveraged **Ansible** extensively for automated configuration management and orchestration across a heterogeneous environment, ensuring consistency and reducing manual effort in managing diverse operating systems (Debian, Arch Linux, Windows), hardware configurations, and software libraries. * **Containerization & Kubernetes Migration:** * Spearheaded the migration of numerous internal services (including web applications, databases, and research tools) from traditional VMs and bare-metal deployments to a **Kubernetes (K3S)** cluster. This enhanced scalability, resilience, and resource utilization. * Implemented **Longhorn** for persistent, distributed block storage within the Kubernetes cluster. * **DevOps & GitOps Implementation:** * Established a modern DevOps workflow centered around a self-hosted **GitLab** instance, utilizing **GitLab CI** for automated testing and container building. * Implemented **Argo CD** for GitOps-based continuous deployment to the Kubernetes cluster, ensuring declarative state management and automated synchronization. * Managed sensitive information using **Sealed Secrets** for secure secret handling within the GitOps workflow. * **Networking & Security:** * Configured **Traefik** as the primary reverse proxy and ingress controller for the Kubernetes cluster, automating routing, service discovery, and TLS certificate management. * Implemented and managed a **WireGuard** VPN server to provide secure remote access for chair members to internal resources. * **ML Workflow Optimization:** * Re-architected the execution environment for machine learning experiments. Transitioned from managing dependencies directly on workstations or via a less reliable SLURM setup to a containerized approach using **Docker**. * Utilized the self-hosted **GitLab Container Registry** for storing ML environment images. --- **Outcomes & Philosophy:** This hands-on role provided deep practical experience in modern system administration, networking, Infrastructure as Code (IaC), and cloud-native technologies within an academic research setting. It fostered my preference for minimalist, reproducible, and microservice-oriented architectures. These principles and skills are actively applied in my personal projects, including the self-hosting and management of this website and various other containerized services. A more comprehensive list of the technologies I work with can be found on the [About Me](/about/) page.]]>Steffen Illium

Primate Subsegment Sorting2023-06-25T00:00:00+02:002023-06-25T00:00:00+02:00http://localhost:4000/research/primate-subsegment-sorting Visualization related to the thresholding or selection process for subsegment labeling

Visualization related to the thresholding or selection process for subsegment labeling

Thresholding or selection criteria for subsegment refinement.

The effectiveness of this methodology was evaluated on the challenging **ComParE 2021 Primate dataset**. The results demonstrate remarkable improvements in classification performance, achieving substantially higher accuracy and Unweighted Average Recall (UAR) scores compared to existing baseline methods. Graphs or tables showing improved classification results (accuracy, UAR) compared to baselines

Graphs or tables showing improved classification results (accuracy, UAR) compared to baselines

Comparative performance results on the ComParE 2021 dataset.

This work represents a significant advancement in handling difficult, real-world bioacoustic data, showcasing how careful data refinement prior to deep learning model training can dramatically enhance classification outcomes. {% cite koelle23primate %}]]>Steffen Illium

Emergent Social Dynamics2023-05-01T00:00:00+02:002023-05-01T00:00:00+02:00http://localhost:4000/research/emergent-social-dynamics Trajectories or state space visualization of the particle population dynamics over time

Trajectories or state space visualization of the particle population dynamics over time

Visualization of particle trajectories or population dynamics within the 'social soup'.

This study highlights how complex, seemingly goal-directed social behaviors and stable ecosystem structures can emerge from simple, local rules based on mutual prediction and surprise minimization among interacting agents, offering insights into the self-organization of complex adaptive systems. {% cite zorn23surprise %}]]>Steffen Illium

Autoencoder Trajectory Compression2023-02-25T00:00:00+01:002023-02-25T00:00:00+01:00http://localhost:4000/research/autoencoder-trajectory-compression Schematic diagram of the LSTM autoencoder model architecture used for trajectory compression

Schematic diagram of the LSTM autoencoder model architecture used for trajectory compression

Schematic of the LSTM Decoder Architecture.

This paper introduces a novel approach for **compressing and reconstructing GPS trajectories** using a **Long Short-Term Memory (LSTM) autoencoder**. The autoencoder learns a compressed latent representation of the trajectory sequence, which can then be decoded to reconstruct the original path. ![Graphs showing mean distance errors (e.g., Fréchet, DTW) for different compression ratios on the T-Drive dataset](\assets\figures\23_trajectory_scores.png){:style="display:block; width:50%" .align-right} Our method was evaluated on two distinct datasets: one from a gaming context and another real-world dataset (T-Drive). We assessed performance across a range of compression ratios and trajectory lengths, comparing it against the widely used traditional **Douglas-Peucker algorithm**. **Key findings:** * The LSTM autoencoder approach significantly **outperforms Douglas-Peucker** in terms of reconstruction accuracy, as measured by both **discrete Fréchet distance** and **Dynamic Time Warping (DTW)**. * Unlike point-reduction techniques like Douglas-Peucker, our method performs a **lossy reconstruction at every point** along the trajectory. This offers potential advantages in maintaining temporal resolution and providing greater flexibility for downstream analysis. Experimental results demonstrate the effectiveness and potential benefits of using deep learning, specifically LSTM autoencoders, for GPS trajectory compression, offering improved accuracy over conventional geometric algorithms. {% cite kolle2023compression %}]]>Steffen Illium

Voronoi Data Augmentation2023-02-24T00:00:00+01:002023-02-24T00:00:00+01:00http://localhost:4000/research/voronoi-data-augmentation Example of an image augmented with VoronoiPatches, showing polygon patches blended onto a lion image

Example of an image augmented with VoronoiPatches, showing polygon patches blended onto a lion image

Visual example of the VoronoiPatches augmentation applied to an image.

Unlike traditional methods that often apply uniform transformations or cutout regions, VP operates by: 1. Generating a random layout of points within an image. 2. Creating a Voronoi diagram based on these points, partitioning the image into unique, convex polygon-shaped patches. 3. Redistributing information between these patches or blending information across patch boundaries (specific mechanism detailed in the paper). This approach potentially allows for smoother transitions between augmented regions and the original image compared to sharp cutout methods. The core idea is to encourage the CNN to learn more robust features by exposing it to varied, non-linearly recombined versions of the input data. ---

:trophy: Best Poster Award - ICAART 2023 :trophy:
(Official Link)

--- Evaluations demonstrate that VoronoiPatches can effectively **reduce model variance and combat overfitting**. Comparative studies indicate that VP **outperforms several existing state-of-the-art data augmentation techniques** in improving the robustness and generalization performance of CNN models on unseen data across various benchmarks. {% cite illium2023voronoipatches %} Graphs showing performance comparison (e.g., accuracy, loss) of VoronoiPatches against other augmentation methods

Graphs showing performance comparison (e.g., accuracy, loss) of VoronoiPatches against other augmentation methods

Comparative results illustrating the performance benefits of VoronoiPatches.

]]>Steffen Illium

Organism Network Emergence2022-12-01T00:00:00+01:002022-12-01T00:00:00+01:00http://localhost:4000/research/organism-network-emergence Diagram showing individual self-replicating units interacting to form a larger Organism Network structure

Diagram showing individual self-replicating units interacting to form a larger Organism Network structure

Conceptual architecture of an Organism Network emerging from interacting self-replicators.

The core hypothesis is that through local interactions and potentially shared environmental feedback, initially homogeneous populations of self-replicators can spontaneously develop specialized roles or structures, leading to a collective entity with capabilities exceeding those of individual units. ![Visualization potentially related to network robustness, differentiation, or communication channels.](\assets\figures\16_on_dropout.jpg) {:style="display:block; width:45%" .align-right} Key aspects explored in this work include: * **Mechanisms for Collaboration:** Investigating how communication or resource sharing between self-replicating units can be established and influence collective behavior. * **Emergent Differentiation:** Analyzing scenarios where units within the population begin to specialize, adopting different internal states (weight configurations) or functions, analogous to cellular differentiation in biological organisms. * **Formation of Structure:** Studying how interactions lead to stable spatial or functional structures within the population, forming the basis of the Organism Network. * **Functional Advantages:** Assessing whether these emergent ONs exhibit novel collective functionalities or improved problem-solving capabilities compared to non-interacting populations. (The role of dropout, as suggested by the image, might relate to promoting robustness or specialization within this context). This study bridges the gap between single-unit self-replication and the emergence of complex, multi-unit systems in artificial life research, offering insights into how collaborative dynamics can lead to higher-order computational structures. For more detailed insights, refer to {% cite illium2022constructing %}.

]]>Steffen Illium

MSP Android Course2022-10-15T00:00:00+02:002022-10-15T00:00:00+02:00http://localhost:4000/teaching/android

Key Learning Areas

Students gained practical experience in:

Native Android App Development (Java/Kotlin)
Android SDK, Activity/Fragment Lifecycle, UI Design (XML Layouts, Jetpack Compose later)
Client-Server Architecture & Networking (e.g., Retrofit, Volley)
Using Wireless Local Networks (WiFi / Bluetooth APIs)
Implementing Location Services (GPS / Fused Location Provider)
Background Processing and Services
Data Persistence (SharedPreferences, SQLite, Room)
Teamwork and Collaborative Software Development (Git)
Agile Methodologies and Project Management Tools

]]>Steffen Illium

Extended Self-Replication2022-08-01T00:00:00+02:002022-08-01T00:00:00+02:00http://localhost:4000/research/extended-self-replication Visualization showing the evolution or diversity of 'child' networks generated through self-replication

Visualization showing the evolution or diversity of 'child' networks generated through self-replication

Analyzing the lineage and diversity in populations of self-replicating networks.

This journal article provides an extended and more in-depth exploration of self-replicating neural networks, building upon earlier foundational work ([Gabor et al., 2019](link-to-previous-paper-if-available)). The research further investigates the use of **backpropagation-like mechanisms** not for typical supervised learning, but as an effective means to enable **non-trivial self-replication** – where networks learn to reproduce their own connection weights. Key extensions and analyses presented in this work include: * **Robustness Analysis:** A systematic evaluation of the self-replicating networks' resilience and stability when subjected to various levels of **noise** during the replication process. * **Artificial Chemistry Environments:** Further development and analysis of simulated environments where populations of self-replicating networks interact, leading to observable **emergent collective behaviors** and ecosystem dynamics. * **Dynamical Systems Perspective:** A detailed theoretical analysis of the self-replication process viewed as a dynamical system. This includes identifying **fixpoint weight configurations** (networks that perfectly replicate themselves) and characterizing their **attractor basins** (the regions in weight space from which networks converge towards a specific fixpoint). Graph showing the impact of different noise levels on self-replication fidelity or population dynamics

Graph showing the impact of different noise levels on self-replication fidelity or population dynamics

Investigating the influence of noise on the self-replication process.

By delving deeper into the mechanisms, robustness, emergent properties, and underlying dynamics, this study significantly enhances the understanding of how self-replication can be achieved and analyzed within neural network models, contributing valuable insights to the fields of artificial life and complex systems. {% cite gabor2022self %}]]>Steffen Illium

Past Course Iterations

Key Learning Areas