Sebastian Tschiatschek

David Lindner, Sebastian Tschiatschek, Katja Hofmann et al.

We study sequential decision-making with known rewards and unknown constraints, motivated by situations where the constraints represent expensive-to-evaluate human preferences, such as safe and comfortable driving behavior. We formalize the challenge of interactively learning about these constraints as a novel linear bandit problem which we call constrained linear best-arm identification. To solve this problem, we propose the Adaptive Constraint Learning (ACOL) algorithm. We provide an instance-dependent lower bound for constrained linear best-arm identification and show that ACOL's sample complexity matches the lower bound in the worst-case. In the average case, ACOL's sample complexity bound is still significantly tighter than bounds of simpler approaches. In synthetic experiments, ACOL performs on par with an oracle solution and outperforms a range of baselines. As an application, we consider learning constraints to represent human preferences in a driving simulation. ACOL is significantly more sample efficient than alternatives for this application. Further, we find that learning preferences as constraints is more robust to changes in the driving scenario than encoding the preferences directly in the reward function.

Ahana Ghosh, Sebastian Tschiatschek, Sam Devlin et al.

Block-based programming environments are increasingly used to introduce computing concepts to beginners. However, novice students often struggle in these environments, given the conceptual and open-ended nature of programming tasks. To effectively support a student struggling to solve a given task, it is important to provide adaptive scaffolding that guides the student towards a solution. We introduce a scaffolding framework based on pop quizzes presented as multi-choice programming tasks. To automatically generate these pop quizzes, we propose a novel algorithm, PQuizSyn. More formally, given a reference task with a solution code and the student's current attempt, PQuizSyn synthesizes new tasks for pop quizzes with the following features: (a) Adaptive (i.e., individualized to the student's current attempt), (b) Comprehensible (i.e., easy to comprehend and solve), and (c) Concealing (i.e., do not reveal the solution code). Our algorithm synthesizes these tasks using techniques based on symbolic reasoning and graph-based code representations. We show that our algorithm can generate hundreds of pop quizzes for different student attempts on reference tasks from Hour of Code: Maze Challenge and Karel. We assess the quality of these pop quizzes through expert ratings using an evaluation rubric. Further, we have built an online platform for practicing block-based programming tasks empowered via pop quiz based feedback, and report results from an initial user study.

Timo Klein, Thomas Lang, Andrii Shkabrii et al.

The performance of reinforcement learning (RL) agents depends critically on the quality of the underlying feature representations. Hyperbolic feature spaces are well-suited for this purpose, as they naturally capture hierarchical and relational structure often present in complex RL environments. However, leveraging these spaces commonly faces optimization challenges due to the nonstationarity of RL. In this work, we identify key factors that determine the success and failure of training hyperbolic deep RL agents. By analyzing the gradients of core operations in the Poincaré Ball and Hyperboloid models of hyperbolic geometry, we show that large-norm embeddings destabilize gradient-based training, leading to trust-region violations in proximal policy optimization (PPO). Based on these insights, we introduce Hyper++, a new hyperbolic PPO agent that consists of three components: (i) stable critic training through a categorical value loss instead of regression; (ii) feature regularization guaranteeing bounded norms while avoiding the curse of dimensionality from clipping; and (iii) using a more optimization-friendly formulation of hyperbolic network layers. In experiments on ProcGen, we show that Hyper++ guarantees stable learning, outperforms prior hyperbolic agents, and reduces wall-clock time by approximately 30%. On Atari-5 with Double DQN, Hyper++ strongly outperforms Euclidean and hyperbolic baselines. We release our code at https://github.com/Probabilistic-and-Interactive-ML/hyper-rl .

Manh Hung Nguyen, Sebastian Tschiatschek, Adish Singla

Student modeling is central to many educational technologies as it enables predicting future learning outcomes and designing targeted instructional strategies. However, open-ended learning domains pose challenges for accurately modeling students due to the diverse behaviors and a large space of possible misconceptions. To approach these challenges, we explore the application of large language models (LLMs) for in-context student modeling in open-ended learning domains. More concretely, given a particular student's attempt on a reference task as observation, the objective is to synthesize the student's attempt on a target task. We introduce a novel framework, LLM for Student Synthesis (LLM-SS), that leverages LLMs for synthesizing a student's behavior. Our framework can be combined with different LLMs; moreover, we fine-tune LLMs to boost their student modeling capabilities. We instantiate several methods based on LLM-SS framework and evaluate them using an existing benchmark, StudentSyn, for student attempt synthesis in a visual programming domain. Experimental results show that our methods perform significantly better than the baseline method NeurSS provided in the StudentSyn benchmark. Furthermore, our method using a fine-tuned version of the GPT-3.5 model is significantly better than using the base GPT-3.5 model and gets close to human tutors' performance.

Sebastian Tschiatschek, Maria Knobelsdorf, Adish Singla

We consider the equity and fairness of curricula derived from Knowledge Tracing models. We begin by defining a unifying notion of an equitable tutoring system as a system that achieves maximum possible knowledge in minimal time for each student interacting with it. Realizing perfect equity requires tutoring systems that can provide individualized curricula per student. In particular, we investigate the design of equitable tutoring systems that derive their curricula from Knowledge Tracing models. We first show that many existing models, including classical Bayesian Knowledge Tracing (BKT) and Deep Knowledge Tracing (DKT), and their derived curricula can fall short of achieving equitable tutoring. To overcome this issue, we then propose a novel model, Bayesian-Bayesian Knowledge Tracing (BBKT), that naturally enables online individualization and, thereby, more equitable tutoring. We demonstrate that curricula derived from our model are more effective and equitable than those derived from classical BKT models. Furthermore, we highlight that improving models with a focus on the fairness of next-step predictions might be insufficient to develop equitable tutoring systems.

Timur Sudak, Sebastian Tschiatschek

We consider the problem of learning Variational Autoencoders (VAEs), i.e., a type of deep generative model, from data with missing values. Such data is omnipresent in real-world applications of machine learning because complete data is often impossible or too costly to obtain. We particularly focus on improving a VAE's amortized posterior inference, i.e., the encoder, which in the case of missing data can be susceptible to learning inconsistent posterior distributions regarding the missingness. To this end, we provide a formal definition of posterior consistency and propose an approach for regularizing an encoder's posterior distribution which promotes this consistency. We observe that the proposed regularization suggests a different training objective than that typically considered in the literature when facing missing values. Furthermore, we empirically demonstrate that our regularization leads to improved performance in missing value settings in terms of reconstruction quality and downstream tasks utilizing uncertainty in the latent space. This improved performance can be observed for many classes of VAEs including VAEs equipped with normalizing flows.

Lukas Miklautz, Timo Klein, Kevin Sidak et al.

This work investigates an important phenomenon in centroid-based deep clustering (DC) algorithms: Performance quickly saturates after a period of rapid early gains. Practitioners commonly address early saturation with periodic reclustering, which we demonstrate to be insufficient to address performance plateaus. We call this phenomenon the "reclustering barrier" and empirically show when the reclustering barrier occurs, what its underlying mechanisms are, and how it is possible to Break the Reclustering Barrier with our algorithm BRB. BRB avoids early over-commitment to initial clusterings and enables continuous adaptation to reinitialized clustering targets while remaining conceptually simple. Applying our algorithm to widely-used centroid-based DC algorithms, we show that (1) BRB consistently improves performance across a wide range of clustering benchmarks, (2) BRB enables training from scratch, and (3) BRB performs competitively against state-of-the-art DC algorithms when combined with a contrastive loss. We release our code and pre-trained models at https://github.com/Probabilistic-and-Interactive-ML/breaking-the-reclustering-barrier .

Gabriele Martino, Sebastian Tschiatschek

Gradient-based attribution methods are model-faithful and scalable, but Integrated Gradients (IG) can be brittle because explanations depend on heuristic baselines, straight-line paths, discretization, and saturation. We propose Fisher--Rao Integrated Gradients (FRInGe), which defines both the reference and interpolation schedule in predictive distribution space. FRInGe replaces input baselines with a maximum-entropy predictive reference and follows a Fisher-Rao geodesic on the probability simplex. The corresponding input-space trajectory is realized through the pullback Fisher metric and stabilized by KL and Euclidean trust regions; attributions are obtained by integrating input gradients along this trajectory. Across six ImageNet architectures, FRInGe most clearly improves calibration-oriented attribution metrics, especially MAS scores, while remaining competitive on perturbation AUC and infidelity.

Timo Klein, Lukas Miklautz, Kevin Sidak et al.

Akin to neuroplasticity in human brains, the plasticity of deep neural networks enables their quick adaption to new data. This makes plasticity particularly crucial for deep Reinforcement Learning (RL) agents: Once plasticity is lost, an agent's performance will inevitably plateau because it cannot improve its policy to account for changes in the data distribution, which are a necessary consequence of its learning process. Thus, developing well-performing and sample-efficient agents hinges on their ability to remain plastic during training. Furthermore, the loss of plasticity can be connected to many other issues plaguing deep RL, such as training instabilities, scaling failures, overestimation bias, and insufficient exploration. With this survey, we aim to provide an overview of the emerging research on plasticity loss for academics and practitioners of deep reinforcement learning. First, we propose a unified definition of plasticity loss based on recent works, relate it to definitions from the literature, and discuss metrics for measuring plasticity loss. Then, we categorize and discuss numerous possible causes of plasticity loss before reviewing currently employed mitigation strategies. Our taxonomy is the first systematic overview of the current state of the field. Lastly, we discuss prevalent issues within the literature, such as a necessity for broader evaluation, and provide recommendations for future research, like gaining a better understanding of an agent's neural activity and behavior.

Sebastian Tschiatschek, Eugenia Stamboliev, Timothée Schmude et al.

We discuss the role of humans in algorithmic decision-making (ADM) for socially relevant problems from a technical and philosophical perspective. In particular, we illustrate tensions arising from diverse expectations, values, and constraints by and on the humans involved. To this end, we assume that a strategic decision-maker (SDM) introduces ADM to optimize strategic and societal goals while the algorithms' recommended actions are overseen by a practical decision-maker (PDM) - a specific human-in-the-loop - who makes the final decisions. While the PDM is typically assumed to be a corrective, it can counteract the realization of the SDM's desired goals and societal values not least because of a misalignment of these values and unmet information needs of the PDM. This has significant implications for the distribution of power between the stakeholders in ADM, their constraints, and information needs. In particular, we emphasize the overseeing PDM's role as a potential political and ethical decision maker, who acts expected to balance strategic, value-driven objectives and on-the-ground individual decisions and constraints. We demonstrate empirically, on a machine learning benchmark dataset, the significant impact an overseeing PDM's decisions can have even if the PDM is constrained to performing only a limited amount of actions differing from the algorithms' recommendations. To ensure that the SDM's intended values are realized, the PDM needs to be provided with appropriate information conveyed through tailored explanations and its role must be characterized clearly. Our findings emphasize the need for an in-depth discussion of the role and power of the PDM and challenge the often-taken view that just including a human-in-the-loop in ADM ensures the 'correct' and 'ethical' functioning of the system.

Manh Hung Nguyen, Victor-Alexandru Pădurean, Alkis Gotovos et al.

Generative AI is transforming computing education by enabling the automatic generation of personalized content and feedback. We investigate its capabilities in providing high-quality programming tasks to students. Despite promising advancements in task generation, a quality gap remains between AI-generated and expert-created tasks. The AI-generated tasks may not align with target programming concepts, could be incomprehensible to students, or may contain critical issues such as incorrect tests. Existing works often require interventions from human teachers for validation. We address these challenges by introducing PyTaskSyn, a novel synthesis technique that first generates a programming task and then decides whether it meets certain quality criteria to be given to students. The key idea is to break this process into multiple stages performed by expert and student agents simulated using both strong and weaker generative models. Through extensive evaluation, we show that PyTaskSyn significantly improves task quality compared to baseline techniques and showcases the importance of each specialized agent type in our validation pipeline. Additionally, we conducted user studies using our publicly available web application and show that PyTaskSyn can deliver high-quality programming tasks comparable to expert-designed ones while reducing workload and costs, and being more engaging than programming tasks that are available in online resources.

Timo Klein, Susanna Weinberger, Adish Singla et al.

We consider the problem of third-person imitation learning with the additional challenge that the learner must select the perspective from which they observe the expert. In our setting, each perspective provides only limited information about the expert's behavior, and the learning agent must carefully select and combine information from different perspectives to achieve competitive performance. This setting is inspired by real-world imitation learning applications, e.g., in robotics, a robot might observe a human demonstrator via camera and receive information from different perspectives depending on the camera's position. We formalize the aforementioned active third-person imitation learning problem, theoretically analyze its characteristics, and propose a generative adversarial network-based active learning approach. Empirically, we demstrate that our proposed approach can effectively learn from expert demonstrations and explore the importance of different architectural choices for the learner's performance.

Manh Hung Nguyen, Sebastian Tschiatschek, Adish Singla

The difficulty and expense of obtaining large-scale human responses make Large Language Models (LLMs) an attractive alternative and a promising proxy for human behavior. However, prior work shows that LLMs often produce homogeneous outputs that fail to capture the rich diversity of human perspectives and behaviors. Thus, rather than trying to capture this diversity with a single LLM agent, we propose a novel framework to construct a set of agents that collectively capture the diversity of a given human population. Each agent is an LLM whose behavior is steered by conditioning on a small set of human demonstrations (task-response pairs) through in-context learning. The central challenge is therefore to select a representative set of LLM agents from the exponentially large space of possible agents. We tackle this selection problem from the lens of submodular optimization. In particular, we develop methods that offer different trade-offs regarding time complexity and performance guarantees. Extensive experiments in crowdsourcing and educational domains demonstrate that our approach constructs agents that more effectively represent human populations compared to baselines. Moreover, behavioral analyses on new tasks show that these agents reproduce the behavior patterns and perspectives of the students and annotators they are designed to represent.

Martin Tappler, Ignacio D. Lopez-Miguel, Sebastian Tschiatschek et al.

We consider the challenging problem of using domain knowledge to improve deep reinforcement learning policies. To this end, we propose LEGIBLE, a novel approach, following a multi-step process, which starts by mining rules from a deep RL policy, constituting a partially symbolic representation. These rules describe which decisions the RL policy makes and which it avoids making. In the second step, we generalize the mined rules using domain knowledge expressed as metamorphic relations. We adapt these relations from software testing to RL to specify expected changes of actions in response to changes in observations. The third step is evaluating generalized rules to determine which generalizations improve performance when enforced. These improvements show weaknesses in the policy, where it has not learned the general rules and thus can be improved by rule guidance. LEGIBLE supported by metamorphic relations provides a principled way of expressing and enforcing domain knowledge about RL environments. We show the efficacy of our approach by demonstrating that it effectively finds weaknesses, accompanied by explanations of these weaknesses, in eleven RL environments and by showcasing that guiding policy execution with rules improves performance w.r.t. gained reward.

Timothée Schmude, Laura Koesten, Torsten Möller et al.

Every AI system that makes decisions about people has a group of stakeholders that are personally affected by these decisions. However, explanations of AI systems rarely address the information needs of this stakeholder group, who often are AI novices. This creates a gap between conveyed information and information that matters to those who are impacted by the system's decisions, such as domain experts and decision subjects. To address this, we present the "XAI Novice Question Bank," an extension of the XAI Question Bank containing a catalog of information needs from AI novices in two use cases: employment prediction and health monitoring. The catalog covers the categories of data, system context, system usage, and system specifications. We gathered information needs through task-based interviews where participants asked questions about two AI systems to decide on their adoption and received verbal explanations in response. Our analysis showed that participants' confidence increased after receiving explanations but that their understanding faced challenges. These included difficulties in locating information and in assessing their own understanding, as well as attempts to outsource understanding. Additionally, participants' prior perceptions of the systems' risks and benefits influenced their information needs. Participants who perceived high risks sought explanations about the intentions behind a system's deployment, while those who perceived low risks rather asked about the system's operation. Our work aims to support the inclusion of AI novices in explainability efforts by highlighting their information needs, aims, and challenges. We summarize our findings as five key implications that can inform the design of future explanations for lay stakeholder audiences.

Timothée Schmude, Laura Koesten, Torsten Möller et al.

We argue that explanations for "algorithmic decision-making" (ADM) systems can profit by adopting practices that are already used in the learning sciences. We shortly introduce the importance of explaining ADM systems, give a brief overview of approaches drawing from other disciplines to improve explanations, and present the results of our qualitative task-based study incorporating the "six facets of understanding" framework. We close with questions guiding the discussion of how future studies can leverage an interdisciplinary approach.

David Lindner, Xin Chen, Sebastian Tschiatschek et al.

We propose Convex Constraint Learning for Reinforcement Learning (CoCoRL), a novel approach for inferring shared constraints in a Constrained Markov Decision Process (CMDP) from a set of safe demonstrations with possibly different reward functions. While previous work is limited to demonstrations with known rewards or fully known environment dynamics, CoCoRL can learn constraints from demonstrations with different unknown rewards without knowledge of the environment dynamics. CoCoRL constructs a convex safe set based on demonstrations, which provably guarantees safety even for potentially sub-optimal (but safe) demonstrations. For near-optimal demonstrations, CoCoRL converges to the true safe set with no policy regret. We evaluate CoCoRL in gridworld environments and a driving simulation with multiple constraints. CoCoRL learns constraints that lead to safe driving behavior. Importantly, we can safely transfer the learned constraints to different tasks and environments. In contrast, alternative methods based on Inverse Reinforcement Learning (IRL) often exhibit poor performance and learn unsafe policies.

Dongge Han, Sebastian Tschiatschek

Abstraction plays an important role in the generalisation of knowledge and skills and is key to sample efficient learning. In this work, we study joint temporal and state abstraction in reinforcement learning, where temporally-extended actions in the form of options induce temporal abstractions, while aggregation of similar states with respect to abstract options induces state abstractions. Many existing abstraction schemes ignore the interplay of state and temporal abstraction. Consequently, the considered option policies often cannot be directly transferred to new environments due to changes in the state space and transition dynamics. To address this issue, we propose a novel abstraction scheme building on successor features. This includes an algorithm for transferring abstract options across different environments and a state abstraction mechanism that allows us to perform efficient planning with the transferred options.

Angus Lamb, Evgeny Saveliev, Yingzhen Li et al.

While deep learning has obtained state-of-the-art results in many applications, the adaptation of neural network architectures to incorporate new output features remains a challenge, as neural networks are commonly trained to produce a fixed output dimension. This issue is particularly severe in online learning settings, where new output features, such as items in a recommender system, are added continually with few or no associated observations. As such, methods for adapting neural networks to novel features which are both time and data-efficient are desired. To address this, we propose the Contextual HyperNetwork (CHN), an auxiliary model which generates parameters for extending the base model to a new feature, by utilizing both existing data as well as any observations and/or metadata associated with the new feature. At prediction time, the CHN requires only a single forward pass through a neural network, yielding a significant speed-up when compared to re-training and fine-tuning approaches. To assess the performance of CHNs, we use a CHN to augment a partial variational autoencoder (P-VAE), a deep generative model which can impute the values of missing features in sparsely-observed data. We show that this system obtains improved few-shot learning performance for novel features over existing imputation and meta-learning baselines across recommender systems, e-learning, and healthcare tasks.

David Lindner, Matteo Turchetta, Sebastian Tschiatschek et al.

For many reinforcement learning (RL) applications, specifying a reward is difficult. This paper considers an RL setting where the agent obtains information about the reward only by querying an expert that can, for example, evaluate individual states or provide binary preferences over trajectories. From such expensive feedback, we aim to learn a model of the reward that allows standard RL algorithms to achieve high expected returns with as few expert queries as possible. To this end, we propose Information Directed Reward Learning (IDRL), which uses a Bayesian model of the reward and selects queries that maximize the information gain about the difference in return between plausibly optimal policies. In contrast to prior active reward learning methods designed for specific types of queries, IDRL naturally accommodates different query types. Moreover, it achieves similar or better performance with significantly fewer queries by shifting the focus from reducing the reward approximation error to improving the policy induced by the reward model. We support our findings with extensive evaluations in multiple environments and with different query types.

Haiyan Yin, Yingzhen Li, Sinno Jialin Pan et al.

Solving real-life sequential decision making problems under partial observability involves an exploration-exploitation problem. To be successful, an agent needs to efficiently gather valuable information about the state of the world for making rewarding decisions. However, in real-life, acquiring valuable information is often highly costly, e.g., in the medical domain, information acquisition might correspond to performing a medical test on a patient. This poses a significant challenge for the agent to perform optimally for the task while reducing the cost for information acquisition. In this paper, we propose a model-based reinforcement learning framework that learns an active feature acquisition policy to solve the exploration-exploitation problem during its execution. Key to the success is a novel sequential variational auto-encoder that learns high-quality representations from partially observed states, which are then used by the policy to maximize the task reward in a cost efficient manner. We demonstrate the efficacy of our proposed framework in a control domain as well as using a medical simulator. In both tasks, our proposed method outperforms conventional baselines and results in policies with greater cost efficiency.

Dongge Han, Michael Wooldridge, Alex Rogers et al.

Submodular functions have been a powerful mathematical model for a wide range of real-world applications. Recently, submodular functions are becoming increasingly important in machine learning (ML) for modelling notions such as information and redundancy among entities such as data and features. Among these applications, a key question is payoff allocation, i.e., how to evaluate the importance of each entity towards the collective objective? To this end, classic solution concepts from cooperative game theory offer principled approaches to payoff allocation. However, despite the extensive body of game-theoretic literature, payoff allocation in submodular games are relatively under-researched. In particular, an important notion that arises in the emerging submodular applications is redundancy, which may occur from various sources such as abundant data or malicious manipulations where a player replicates its resource and act under multiple identities. Though many game-theoretic solution concepts can be directly used in submodular games, naively applying them for payoff allocation in these settings may incur robustness issues against replication. In this paper, we systematically study the replication manipulation in submodular games and investigate replication robustness, a metric that quantitatively measures the robustness of solution concepts against replication. Using this metric, we present conditions which theoretically characterise the robustness of semivalues, a wide family of solution concepts including the Shapley and Banzhaf value. Moreover, we empirically validate our theoretical results on an emerging submodular ML application, i.e., the ML data market.

Chao Ma, Sebastian Tschiatschek, José Miguel Hernández-Lobato et al.

Deep generative models often perform poorly in real-world applications due to the heterogeneity of natural data sets. Heterogeneity arises from data containing different types of features (categorical, ordinal, continuous, etc.) and features of the same type having different marginal distributions. We propose an extension of variational autoencoders (VAEs) called VAEM to handle such heterogeneous data. VAEM is a deep generative model that is trained in a two stage manner such that the first stage provides a more uniform representation of the data to the second stage, thereby sidestepping the problems caused by heterogeneous data. We provide extensions of VAEM to handle partially observed data, and demonstrate its performance in data generation, missing data prediction and sequential feature selection tasks. Our results show that VAEM broadens the range of real-world applications where deep generative models can be successfully deployed.

Zichao Wang, Sebastian Tschiatschek, Simon Woodhead et al.

Online education platforms enable teachers to share a large number of educational resources such as questions to form exercises and quizzes for students. With large volumes of available questions, it is important to have an automated way to quantify their properties and intelligently select them for students, enabling effective and personalized learning experiences. In this work, we propose a framework for mining insights from educational questions at scale. We utilize the state-of-the-art Bayesian deep learning method, in particular partial variational auto-encoders (p-VAE), to analyze real students' answers to a large collection of questions. Based on p-VAE, we propose two novel metrics that quantify question quality and difficulty, respectively, and a personalized strategy to adaptively select questions for students. We apply our proposed framework to a real-world dataset with tens of thousands of questions and tens of millions of answers from an online education platform. Our framework not only demonstrates promising results in terms of statistical metrics but also obtains highly consistent results with domain experts' evaluation.

Wolfgang Roth, Günther Schindler, Bernhard Klein et al.

While machine learning is traditionally a resource intensive task, embedded systems, autonomous navigation, and the vision of the Internet of Things fuel the interest in resource-efficient approaches. These approaches aim for a carefully chosen trade-off between performance and resource consumption in terms of computation and energy. The development of such approaches is among the major challenges in current machine learning research and key to ensure a smooth transition of machine learning technology from a scientific environment with virtually unlimited computing resources into everyday's applications. In this article, we provide an overview of the current state of the art of machine learning techniques facilitating these real-world requirements. In particular, we focus on resource-efficient inference based on deep neural networks (DNNs), the predominant machine learning models of the past decade. We give a comprehensive overview of the vast literature that can be mainly split into three non-mutually exclusive categories: (i) quantized neural networks, (ii) network pruning, and (iii) structural efficiency. These techniques can be applied during training or as post-processing, and they are widely used to reduce the computational demands in terms of memory footprint, inference speed, and energy efficiency. We also briefly discuss different concepts of embedded hardware for DNNs and their compatibility with machine learning techniques as well as potential for energy and latency reduction. We substantiate our discussion with experiments on well-known benchmark data sets using compression techniques (quantization, pruning) for a set of resource-constrained embedded systems, such as CPUs, GPUs and FPGAs. The obtained results highlight the difficulty of finding good trade-offs between resource efficiency and prediction quality.

Olga Ohrimenko, Shruti Tople, Sebastian Tschiatschek

We study the problem of collaborative machine learning markets where multiple parties can achieve improved performance on their machine learning tasks by combining their training data. We discuss desired properties for these machine learning markets in terms of fair revenue distribution and potential threats, including data replication. We then instantiate a collaborative market for cases where parties share a common machine learning task and where parties' tasks are different. Our marketplace incentivizes parties to submit high quality training and true validation data. To this end, we introduce a novel payment division function that is robust-to-replication and customized output models that perform well only on requested machine learning tasks. In experiments, we validate the assumptions underlying our theoretical analysis and show that these are approximately satisfied for commonly used machine learning models.

Maximilian Igl, Kamil Ciosek, Yingzhen Li et al.

The ability for policies to generalize to new environments is key to the broad application of RL agents. A promising approach to prevent an agent's policy from overfitting to a limited set of training environments is to apply regularization techniques originally developed for supervised learning. However, there are stark differences between supervised learning and RL. We discuss those differences and propose modifications to existing regularization techniques in order to better adapt them to RL. In particular, we focus on regularization techniques relying on the injection of noise into the learned function, a family that includes some of the most widely used approaches such as Dropout and Batch Normalization. To adapt them to RL, we propose Selective Noise Injection (SNI), which maintains the regularizing effect the injected noise has, while mitigating the adverse effects it has on the gradient quality. Furthermore, we demonstrate that the Information Bottleneck (IB) is a particularly well suited regularization technique for RL as it is effective in the low-data regime encountered early on in training RL agents. Combining the IB with SNI, we significantly outperform current state of the art results, including on the recently proposed generalization benchmark Coinrun.

Ahana Ghosh, Sebastian Tschiatschek, Hamed Mahdavi et al.

We study the problem of designing AI agents that can robustly cooperate with people in human-machine partnerships. Our work is inspired by real-life scenarios in which an AI agent, e.g., a virtual assistant, has to cooperate with new users after its deployment. We model this problem via a parametric MDP framework where the parameters correspond to a user's type and characterize her behavior. In the test phase, the AI agent has to interact with a user of unknown type. Our approach to designing a robust AI agent relies on observing the user's actions to make inferences about the user's type and adapting its policy to facilitate efficient cooperation. We show that without being adaptive, an AI agent can end up performing arbitrarily bad in the test phase. We develop two algorithms for computing policies that automatically adapt to the user in the test phase. We demonstrate the effectiveness of our approach in solving a two-agent collaborative task.

Wenbo Gong, Sebastian Tschiatschek, Richard Turner et al.

In this paper we introduce the ice-start problem, i.e., the challenge of deploying machine learning models when only little or no training data is initially available, and acquiring each feature element of data is associated with costs. This setting is representative for the real-world machine learning applications. For instance, in the health-care domain, when training an AI system for predicting patient metrics from lab tests, obtaining every single measurement comes with a high cost. Active learning, where only the label is associated with a cost does not apply to such problem, because performing all possible lab tests to acquire a new training datum would be costly, as well as unnecessary due to redundancy. We propose Icebreaker, a principled framework to approach the ice-start problem. Icebreaker uses a full Bayesian Deep Latent Gaussian Model (BELGAM) with a novel inference method. Our proposed method combines recent advances in amortized inference and stochastic gradient MCMC to enable fast and accurate posterior inference. By utilizing BELGAM's ability to fully quantify model uncertainty, we also propose two information acquisition functions for imputation and active prediction problems. We demonstrate that BELGAM performs significantly better than the previous VAE (Variational autoencoder) based models, when the data set size is small, using both machine learning benchmarks and real-world recommender systems and health-care applications. Moreover, based on BELGAM, Icebreaker further improves the performance and demonstrate the ability to use minimum amount of the training data to obtain the highest test time performance.

Sebastian Tschiatschek, Ahana Ghosh, Luis Haug et al.

Inverse reinforcement learning (IRL) enables an agent to learn complex behavior by observing demonstrations from a (near-)optimal policy. The typical assumption is that the learner's goal is to match the teacher's demonstrated behavior. In this paper, we consider the setting where the learner has its own preferences that it additionally takes into consideration. These preferences can for example capture behavioral biases, mismatched worldviews, or physical constraints. We study two teaching approaches: learner-agnostic teaching, where the teacher provides demonstrations from an optimal policy ignoring the learner's preferences, and learner-aware teaching, where the teacher accounts for the learner's preferences. We design learner-aware teaching algorithms and show that significant performance improvements can be achieved over learner-agnostic teaching.

Franz Pernkopf, Wolfgang Roth, Matthias Zoehrer et al.

While machine learning is traditionally a resource intensive task, embedded systems, autonomous navigation and the vision of the Internet-of-Things fuel the interest in resource efficient approaches. These approaches require a carefully chosen trade-off between performance and resource consumption in terms of computation and energy. On top of this, it is crucial to treat uncertainty in a consistent manner in all but the simplest applications of machine learning systems. In particular, a desideratum for any real-world system is to be robust in the presence of outliers and corrupted data, as well as being `aware' of its limits, i.e.\ the system should maintain and provide an uncertainty estimate over its own predictions. These complex demands are among the major challenges in current machine learning research and key to ensure a smooth transition of machine learning technology into every day's applications. In this article, we provide an overview of the current state of the art of machine learning techniques facilitating these real-world requirements. First we provide a comprehensive review of resource-efficiency in deep neural networks with focus on techniques for model size reduction, compression and reduced precision. These techniques can be applied during training or as post-processing and are widely used to reduce both computational complexity and memory footprint. As most (practical) neural networks are limited in their ways to treat uncertainty, we contrast them with probabilistic graphical models, which readily serve these desiderata by means of probabilistic inference. In that way, we provide an extensive overview of the current state-of-the-art of robust and efficient machine learning for real-world systems.

Luis Haug, Sebastian Tschiatschek, Adish Singla

Learning near-optimal behaviour from an expert's demonstrations typically relies on the assumption that the learner knows the features that the true reward function depends on. In this paper, we study the problem of learning from demonstrations in the setting where this is not the case, i.e., where there is a mismatch between the worldviews of the learner and the expert. We introduce a natural quantity, the teaching risk, which measures the potential suboptimality of policies that look optimal to the learner in this setting. We show that bounds on the teaching risk guarantee that the learner is able to find a near-optimal policy using standard algorithms based on inverse reinforcement learning. Based on these findings, we suggest a teaching scheme in which the expert can decrease the teaching risk by updating the learner's worldview, and thus ultimately enable her to find a near-optimal policy.

David Janz, Jiri Hron, Przemysław Mazur et al.

Posterior sampling for reinforcement learning (PSRL) is an effective method for balancing exploration and exploitation in reinforcement learning. Randomised value functions (RVF) can be viewed as a promising approach to scaling PSRL. However, we show that most contemporary algorithms combining RVF with neural network function approximation do not possess the properties which make PSRL effective, and provably fail in sparse reward problems. Moreover, we find that propagation of uncertainty, a property of PSRL previously thought important for exploration, does not preclude this failure. We use these insights to design Successor Uncertainties (SU), a cheap and easy to implement RVF algorithm that retains key properties of PSRL. SU is highly effective on hard tabular exploration benchmarks. Furthermore, on the Atari 2600 domain, it surpasses human performance on 38 of 49 games tested (achieving a median human normalised score of 2.09), and outperforms its closest RVF competitor, Bootstrapped DQN, on 36 of those.

Chao Ma, Sebastian Tschiatschek, Konstantina Palla et al.

Many real-life decision-making situations allow further relevant information to be acquired at a specific cost, for example, in assessing the health status of a patient we may decide to take additional measurements such as diagnostic tests or imaging scans before making a final assessment. Acquiring more relevant information enables better decision making, but may be costly. How can we trade off the desire to make good decisions by acquiring further information with the cost of performing that acquisition? To this end, we propose a principled framework, named EDDI (Efficient Dynamic Discovery of high-value Information), based on the theory of Bayesian experimental design. In EDDI, we propose a novel partial variational autoencoder (Partial VAE) to predict missing data entries problematically given any subset of the observed ones, and combine it with an acquisition function that maximizes expected information gain on a set of target variables. We show cost reduction at the same decision quality and improved decision quality at the same cost in multiple machine learning benchmarks and two real-world health-care applications.

Martin Ratajczak, Sebastian Tschiatschek, Franz Pernkopf

We consider higher-order linear-chain conditional random fields (HO-LC-CRFs) for sequence modelling, and use sum-product networks (SPNs) for representing higher-order input- and output-dependent factors. SPNs are a recently introduced class of deep models for which exact and efficient inference can be performed. By combining HO-LC-CRFs with SPNs, expressive models over both the output labels and the hidden variables are instantiated while still enabling efficient exact inference. Furthermore, the use of higher-order factors allows us to capture relations of multiple input segments and multiple output labels as often present in real-world data. These relations can not be modelled by the commonly used first-order models and higher-order models with local factors including only a single output label. We demonstrate the effectiveness of our proposed models for sequence labeling. In extensive experiments, we outperform other state-of-the-art methods in optical character recognition and achieve competitive results in phone classification.

Sebastian Tschiatschek, Kai Arulkumaran, Jan Stühmer et al.

Partially observable Markov decision processes (POMDPs) are a powerful abstraction for tasks that require decision making under uncertainty, and capture a wide range of real world tasks. Today, effective planning approaches exist that generate effective strategies given black-box models of a POMDP task. Yet, an open question is how to acquire accurate models for complex domains. In this paper we propose DELIP, an approach to model learning for POMDPs that utilizes amortized structured variational inference. We empirically show that our model leads to effective control strategies when coupled with state-of-the-art planners. Intuitively, model-based approaches should be particularly beneficial in environments with changing reward structures, or where rewards are initially unknown. Our experiments confirm that DELIP is particularly effective in this setting.

Sebastian Tschiatschek, Aytunc Sahin, Andreas Krause

We consider learning of submodular functions from data. These functions are important in machine learning and have a wide range of applications, e.g. data summarization, feature selection and active learning. Despite their combinatorial nature, submodular functions can be maximized approximately with strong theoretical guarantees in polynomial time. Typically, learning the submodular function and optimization of that function are treated separately, i.e. the function is first learned using a proxy objective and subsequently maximized. In contrast, we show how to perform learning and optimization jointly. By interpreting the output of greedy maximization algorithms as distributions over sequences of items and smoothening these distributions, we obtain a differentiable objective. In this way, we can differentiate through the maximization algorithms and optimize the model to work well with the optimization algorithm. We theoretically characterize the error made by our approach, yielding insights into the tradeoff of smoothness and accuracy. We demonstrate the effectiveness of our approach for jointly learning and optimizing on synthetic maximum cut data, and on real world applications such as product recommendation and image collection summarization.

Christoph Hirnschall, Adish Singla, Sebastian Tschiatschek et al.

We study platforms in the sharing economy and discuss the need for incentivizing users to explore options that otherwise would not be chosen. For instance, rental platforms such as Airbnb typically rely on customer reviews to provide users with relevant information about different options. Yet, often a large fraction of options does not have any reviews available. Such options are frequently neglected as viable choices, and in turn are unlikely to be evaluated, creating a vicious cycle. Platforms can engage users to deviate from their preferred choice by offering monetary incentives for choosing a different option instead. To efficiently learn the optimal incentives to offer, we consider structural information in user preferences and introduce a novel algorithm - Coordinated Online Learning (CoOL) - for learning with structural information modeled as convex constraints. We provide formal guarantees on the performance of our algorithm and test the viability of our approach in a user study with data of apartments on Airbnb. Our findings suggest that our approach is well-suited to learn appropriate incentives and increase exploration on the investigated platform.

Andrew An Bian, Joachim M. Buhmann, Andreas Krause et al.

We investigate the performance of the standard Greedy algorithm for cardinality constrained maximization of non-submodular nondecreasing set functions. While there are strong theoretical guarantees on the performance of Greedy for maximizing submodular functions, there are few guarantees for non-submodular ones. However, Greedy enjoys strong empirical performance for many important non-submodular functions, e.g., the Bayesian A-optimality objective in experimental design. We prove theoretical guarantees supporting the empirical performance. Our guarantees are characterized by a combination of the (generalized) curvature $α$ and the submodularity ratio $γ$. In particular, we prove that Greedy enjoys a tight approximation guarantee of $\frac{1}α(1- e^{-γα})$ for cardinality constrained maximization. In addition, we bound the submodularity ratio and curvature for several important real-world objectives, including the Bayesian A-optimality objective, the determinantal function of a square submatrix and certain linear programs with combinatorial constraints. We experimentally validate our theoretical findings for both synthetic and real-world applications.

Christoph Hirnschall, Adish Singla, Sebastian Tschiatschek et al.

We study an online multi-task learning setting, in which instances of related tasks arrive sequentially, and are handled by task-specific online learners. We consider an algorithmic framework to model the relationship of these tasks via a set of convex constraints. To exploit this relationship, we design a novel algorithm -- COOL -- for coordinating the individual online learners: Our key idea is to coordinate their parameters via weighted projections onto a convex set. By adjusting the rate and accuracy of the projection, the COOL algorithm allows for a trade-off between the benefit of coordination and the required computation/communication. We derive regret bounds for our approach and analyze how they are influenced by these trade-off factors. We apply our results on the application of learning users' preferences on the Airbnb marketplace with the goal of incentivizing users to explore under-reviewed apartments.

Adish Singla, Sebastian Tschiatschek, Andreas Krause

Motivated by an application of eliciting users' preferences, we investigate the problem of learning hemimetrics, i.e., pairwise distances among a set of $n$ items that satisfy triangle inequalities and non-negativity constraints. In our application, the (asymmetric) distances quantify private costs a user incurs when substituting one item by another. We aim to learn these distances (costs) by asking the users whether they are willing to switch from one item to another for a given incentive offer. Without exploiting structural constraints of the hemimetric polytope, learning the distances between each pair of items requires $Θ(n^2)$ queries. We propose an active learning algorithm that substantially reduces this sample complexity by exploiting the structural constraints on the version space of hemimetrics. Our proposed algorithm achieves provably-optimal sample complexity for various instances of the task. For example, when the items are embedded into $K$ tight clusters, the sample complexity of our algorithm reduces to $O(n K)$. Extensive experiments on a restaurant recommendation data set support the conclusions of our theoretical analysis.

Adish Singla, Sebastian Tschiatschek, Andreas Krause

We address the problem of maximizing an unknown submodular function that can only be accessed via noisy evaluations. Our work is motivated by the task of summarizing content, e.g., image collections, by leveraging users' feedback in form of clicks or ratings. For summarization tasks with the goal of maximizing coverage and diversity, submodular set functions are a natural choice. When the underlying submodular function is unknown, users' feedback can provide noisy evaluations of the function that we seek to maximize. We provide a generic algorithm -- \submM{} -- for maximizing an unknown submodular function under cardinality constraints. This algorithm makes use of a novel exploration module -- \blbox{} -- that proposes good elements based on adaptively sampling noisy function evaluations. \blbox{} is able to accommodate different kinds of observation models such as value queries and pairwise comparisons. We provide PAC-style guarantees on the quality and sampling cost of the solution obtained by \submM{}. We demonstrate the effectiveness of our approach in an interactive, crowdsourced image collection summarization application.