In one sentence: what is an existentially closed Heyting algebra and what does it have to do with automata? We will also discuss possible open problems and future research directions in this area.

17.12.19 Lighting continued

The exam will cover material from the 4 overall areas: Each of the 4 areas will be scored about equally.
Students have to make sure that their solution is handed in at the submission date. Required formulas will be provided in the task descriptions.

Many existing post-correction approaches employ weighted finite-state transducers (WFST), which allow for compact storage and processing of symbol sequences.

However, rank-based semantics are generally not applicable for cyclic graphs and therefore, a second approach to ranking sets of arguments is being introduced, as well.      good book for first part: Polygonal Mesh Processing

28.01.20 Optimization (slidesupdated 03.02.20)

Feb 11, 2020.

The release date is usually also the debriefing date of the previous exercise, where the supervisors will walk through the theoretical assignments and explain the solution (Friday, 09:20, APB E023). In the talk, we also discuss some relations to subgraph counting algorithms as well as machine learning. The Temporal Stream-based Specification Language (TeSSLa) is a specification language developed for the specification of properties over real-time constraints and asynchronous arrival of events on the input streams. CMS students enroll for the exam in Selma; while exchange students should register by email to Benjamin Russig. Our starting point is a classical algorithm of Michael Karr for discovering affine invariants. Time & Place:   Friday, 09:20, APB E023 (Briefing + Debriefing, see below) Although optical character recognition (OCR) quality has improved substantially over the last decade, it still struggles on historic material. 29.10.19 Halfedge Data Structure (slidesupdated 29.10.19) Add event to calendar; Die offiziellen Einführungsveranstaltungen des Zentrums für Internationale Studien (ZIS) finden am 09. Current research on invariant generation employs an eclectic array of techniques, including abductive inference, abstract interpretation, constraint solving, interpolation, and machine learning. Which arguments are responsible for undecisiveness in argumentation semantics? Sliding-window streaming algorithms get as input a stream of data values and have to answer queries about the last n values for a certain window size n. In the talk we consider queries that are given by regular languages. Unfortunately, stable labellings are not guaranteed to exist, thus raising the question as to which parts of AFs are responsible for this non-existence.

On the theoretical side, we develop sufficient conditions to guarantee its termination (i.e., acyclicity notions), and study several restrictions that furthermore ensure its polynomiality.

For example, a polynomial time algorithm for model checking Markov chains against UBA-properties is known, which is not possible for properties specified by NBA. The talk will introduce this structure and define the Constraint Satisfaction Problem of its first-order reducts. Theoretical submissions can be submitted in three ways: The method allows the practitioner to start with an accurate and trustworthy application-domain-centric system model and to link such a `ground model' in a well documented and controllable way through intermediate design steps (called `refinements') to its implementation. We will show that the complexity of the satisfiability problem behaves well while extending the logic with threshold-counting or modulo-counting quantifiers. Surprisingly, for the case of disjoint unions, the fragment is the same as the one used in the Büchi-type result of weighted automata.

The aim of our work here is to gain a better understanding of such logical-algebraic structures by studying them from the perspective of model theory.

TU Dresden is co-financed by tax funds using the budget approved by the Landtag of the Free State of Saxony.

04.02.20 Optimization & Extro, A list of sample questions for oral exams can be found here: questionaire. The assignments have to be completed in teams of three students.

To get points for the practical submission, each team has to present their work to a tutor on the evaluation date. (a) in written form on A4 paper by 16:00 to chair staff - we recommend handing in solutions on physical paper at the end of the previous lecture.

One can, for example, restrict the type of cost functions that are allowed to appear in the description of an instance. Through its versatility the ASM approach is non-monolithic and integratable at any development level into current software design and analysis environments.

Debriefing: Supervisors present and explain solutions of the theoretical assignments (Friday 09:20 in APB E023)

A seemingly independent concept that has been studied in the context of probabilistic reachability is that of a witnessing subsystem. However, in an important particular case where the lattice is distributive, a subexponential algorithm can be proposed. A formula of TeSSLa is in the end a function mapping a set of input streams to a set of output streams.This talk will provide a short introduction to TeSSLa and its application domains at first and then focus on the theoretical results for TeSSLa and various fragments of it regarding other well known formalisms like Turing machines, automata and logics as well as complexity results for different decision problems like equivalence of formulas. In recent years, interest in AFT has gradually increased, with applications now ranging from foundations of database theory to abstract argumentation.In this presentation, we provide an overview of the field: we start from the motivations that drove Denecker et al to the conception of this field, discuss both theoretical developments and applications, and end with potential topics for future improvements.Of particular interest in this presentation is the  application of AFT for (weighted) argumentation. There are four exercises comprising theoretical and practical assignments. In particular we discuss how automated reasoning systems can be used for natural language question answering.

The idea is to reduce the question of whether a system satisfies a property, which is given as an LTL-formula, to the emptiness check of an automaton.

We also discuss the complexity of dualization for several partial cases, as well as complexity of dualization when the lattice is given by an implication base.

We show that model completeness also has an important role to play in logical algebra. In the model-theoretic study of usual algebra, the concept of model completeness plays a central role: it provides the correct abstraction of the concept of an algebraically closed field. TU Dresden, WS 2019/20 Introduction to Nonmonotonic Reasoning Slide 172. We present a complete classification of (infinite-domain) temporal CSPs that can be solved in fixed point logic (with or without counting). The talk will be about the two-variable fragment of First-Order Logic interpreted on words.Motivation: There are several reasons why logicians could be interested in FO2[<]. A well-known fact underlying Formal Concept Analysis (FCA) is that every lattice is determined, up to isomorphism, by the ordered set of its meet (infimum) and join (supremum) irreducible elements. TU Dresden Institut für Technische Informatik. Using prior knowledge on what mistakes OCR typically makes, how new words are formed grammatically, and which words are likely to appear next to each other,a post-correction system can be modelled as the composition of single transducers representing input character hypotheses, error model, lexicon, and word-level language model, each weighted with probabilities. Another possibility is to restrict the structure of the instance, that is, the way that cost functions are combined.


Moreover, for the constant and logarithmic space classes we provide very natural characterizations: For every regular language L the sliding window word problem can be solved in(i) constant space if and only if  L is a boolean combination of regular length languages and suffix-testable languages and in(ii)  logarithmic space if and only if  L is a boolean combination of regular length languages and regular left ideals. We show that a Farkas certificate with K non-zero entries can be translated into a witnessing subsystem with K states, and vice versa.

This logic will lead us to a new characterization of the robust class of so called regular transductions.