We investigate how to learn functions that rate game situations on a soccer pitch according to their potential to lead to successful attacks. We follow a purely data-driven approach using techniques from deep reinforcement learning to valuate multiplayer positionings based on positional data. Empirically, the predicted scores highly correlate with dangerousness of actual situations and show that rating of player positioning without expert knowledge is possible.

We address the problem of learning decision functions from training data in which some attribute values are unobserved. This problem can arise, for instance, when training data is aggregated from multiple sources, and some sources record only a subset of attributes. We derive a generic joint optimization problem in which the distribution governing the missing values is a free parameter. We show that the optimal solution concentrates the density mass on finitely many imputations, and provide a corresponding algorithm for learning from incomplete data. We report on empirical results on benchmark data, and on the email spam application that motivates our work.

Inductive Logic Programming (ILP) [4] combines techniques from machine learning with the representation of logic programming. It aims at inducing logical clauses, i.e, general rules from specific observations and background knowledge. Because of focusing on logical clauses, traditional ILP systems do not model uncertainty explicitly. On the other hand, state-of-the-art probabilistic models such as Bayesian networks (BN) [5], hidden Markov models, and stochastic contextfree grammars have a rigid structure and therefore have problems representing a variable number of objects and relations among these objects. Recently, various relational extensions of traditional probabilistic models have been proposed, see [1] for an overview. The newly emerging field of stochastic relational learning (SRL) studies learning such rich probabilistic models. The Balios EngineBalios is an inference engine for Bayesian logic programs (BLPs) [3,2]. BLPs combine BNs with definite clause logic. The basic idea is to view logical atoms as sets of random variables which are similar to each other. Consider the modelling the inheritance of a single gene that determines a person's P blood type bt(P). Each person P has two copies of the chromosome containing this gene, one, mc(M), inherited from her mother mother(M, P), and one, pc(F), inherited from her father father(F, P). Such a general influence relation cannot be captured within BNs.Knowledge Representation: Like BNs, BLPs separate the qualitative, i.e., the influence relations among random variables, from the quantitative aspects of the world, i.e., the strength of influences. In contrast to BNs, however, they allow to capture general probabilistic regularities. Consider the BLP shown in Figure 1 modelling our genetic domain. The rule graph gives an overview of all interactions (boxes) among abstract random variables (ovals). For instance, the maternal information mc/1 is specified in terms of mothers mother/2, maternal mc/1 and paternal pc/1 information. Each interaction gives rise to a local probabilistic model which is composed of a qualitative and a quantitative part. For instance, rule 2 in Figure 1(a) encodes that "the maternal genetic information mc(P) of a person P is influenced by the maternal mc(M) and paternal pc(M) genetic information of P's mother M.

We focus on the problem of detecting clients that attempt to exhaust server resources by flooding a service with protocol-compliant HTTP requests. Attacks are usually coordinated by an entity that controls many clients. Modeling the application as a structuredprediction problem allows the prediction model to jointly classify a multitude of clients based on their cohesion of otherwise inconspicuous features. Since the resulting output space is too vast to search exhaustively, we employ greedy search and techniques in which a parametric controller guides the search. We apply a known method that sequentially learns the controller and the structured-prediction model. We then derive an online policy-gradient method that finds the parameters of the controller and of the structured-prediction model in a joint optimization problem; we obtain a convergence guarantee for the latter method. We evaluate and compare the various methods based on a large collection of traffic data of a web-hosting service.

In many team sports, the ability to control and generate space in dangerous areas on the pitch is crucial for the success of a team. This holds, in particular, for soccer. In this study, we revisit ideas from Fernandez and Bornn (2018) who introduced interesting space-related quantities including pitch control (PC) and pitch value. We identify influence of the player on the pitch with the movements of the player and turn their concepts into data-driven quantities that give rise to a variety of different applications. Furthermore, we devise a novel space generation measure to visualize the strategies of the team and player. We provide empirical evidence for the usefulness of our contribution and showcase our approach in the context of game analyses.