DTAI Thesis Topics Dept. Computer Science KU Leuven 2020-2021 http://people.cs.kuleuven.be/~luc.deraedt/dtaithesis20-21.pdf Luc De Raedt
Lab for Declarative Languages and Artificial Intelligence Machine Learning 3 ZAP , 1 res. manager ± 5 post-docs ± 25 Ph.D. students Declarative Languages and Systems 4 ZAP ± 3 post-docs ±12 Ph.D. students Demoen retired Bruynooghe & still interested De Schreye � 2 in education in informatics retired
AI is hot! Self-driving cars - Eve (the robot scientist) Siri IBM Watson in Jeopardy and “Machine Reading” AlphaGo — (Deep) learning … � 3
Machine Learning, Reasoning Typically associated with deep/ Involves multi-step, multi-modal probabilistic learning . “ reasoning ”. Focus on Focus on patterns and data. knowledge and logical inference Machine Learning Reasoning System 1: instinct, reflexes System 2: deliberate, logical Fast Thinking Slow Thinking Associations: Seeing, observing Interventions: Doing, intervening Counterfactuals: Retrospection, understanding What if I see …? What if I do or had done …? Great results but only reaches 90% Required to be actionable, trustworthy, … “Is it a stop sign?” “Do you stop at a stop sign?” One of the grand challenges in AI: Combine reasoning with machine learning. � 4
Learning and Reasoning July 2019: January 2020: February 2020: Holger Hoos & Philipp AI For Europe — COM(2018) 237 Daniel Kahneman, Yoshua Bengio, Slusallek, CLAIRE: ICT-48-2020 “European network of AI Nobel prize laureate: Turing Award winner: “European Researchers look excellence centres”: “We need System I and “We need to expand Deep beyond deep learning” “Necessary competencies are: System II with a Learning from System I to learning and reasoning, XAI, unbiased representation of the System II” AI, safety, reliability and verifiability.” world” Ursula von der Leyen, Henry Kautz, Former Yann LeCun, Minister Muyters / Crevits, President of the president of Association for Turing Award winner: Flemish AI Action Plan: Commission: the Advancement of Artificial “Our next challenge for “Hybrid AI to support reasoning […] Intelligence: “Combining symbolic Deep Learning is and complex decision making” reasoning with deep “Violent agreement on the learning to reason.” neural networks may help need to bring together the us improve explainability neural and symbolic of AI outcomes” traditions” � 5
DTAI's focus on learning and reasoning Machine Learning & Data Mining how to extract knowledge from data Uncertainty reasoning how to represent and reason about uncertainty Knowledge Representation how to represent and reason about knowledge � 6
DTAI's focus on Declarative Languages Declarative = specify the what rather than the how Different types of languages Logic Functional Constraints Probabilistic Explainable / Understandable AI � 7
DTAI's methodology involves Fundamental research (theoretical as well as empirical) Systems, Solvers and Software Applications Thesis can focus on one or more aspects, depending on interests student This presentation does not go in depth about techniques but every thesis does � 8
This presentation Overview of research illustrations of possible thesis topics. List of contact persons for topics Full information — see online (needs update) ( dtai.cs.kuleuven.be/research ) i Own topic should be aligned with interests professor � 9
Research topics
Predictive learning and clustering Contact: Hendrik Blockeel, Jesse Davis “Standard” machine learning develop new algorithms for machine learning Decision Trees Predictive Clustering Probabilistic Graphical Models PU Learning evaluation of machine learning (ROC etc.) � 11
Predictive learning and clustering “fill in the Contact: Hendrik Blockeel A B C D missing values” A B C D ? ? ? ? X 2 X 3 model learner ABD → C learning online X 1 X 4 A B C D ! ? ? ? X 6 X 5 MERCS, see van Wolputte et al. IJCAI 18 � 12
Automated Data Science Contact: Luc De Raedt Can we (partly) automate data science ? One example : Can we automatically derive the right features ? the right representations ? learning Can we automatically discover what we can learn / predict ? constraints Can we learn constraints ? optimisation function Can we automatically wrangle the data ? (programming by example) Example database about students, professors, courses, and marks … we use a SpreadSheet Context The SYNTH project — https://synth.cs.kuleuven.be/ the democratisation of Data Science the automation of Data Science � 13
Contact: Luc De Raedt Automated Data Science AI: Can we automate X ? (where X requires Interpretation intelligence) and Evaluation Data Mining X = Science -> Knowledge Selection and Preprocessing X = Data Science Data Consolidation Would be pretty useful Pattern Prepared It is about automating the whole process ! Consolid KDD Process [Fayyad] <>AutoML SYNTH Project � 14
Probabilistic Programming and Statistical Relational Learning Contact: Luc De Raedt, Hendrik Blockeel, Jesse Davis Key open question in AI — integrate Probabilistic reasoning Logical or relational Machine learning representations statistical relational learning probabilistic programming � 15
Probabilistic Programming and Statistical Relational Learning Contact: Luc De Raedt E.g. ProbLog: a probabilistic Prolog i 0.05 :: burglary. alarm :- burglary. 0.01 :: earthquake. alarm :- earthquake. 0.7 :: hears_alarm(john). calls(Pers) :- alarm, hears_alarm(Pers). 0.6 :: hears_alarm(mary). P( hears_alarm(john) | burglary = true) ? Challenges on inference, learning, implementation, application, ... � 16
Logic + Probability + Neural Networks Contact: Hendrik Blockeel, Luc De Raedt + = 16 + = 3 + = ? + = 4 Data Query Answer + = ? Query Answer DeepProbLog [Manhaeve NeurIPS 2018] � 17
Probabilistic Programming and Statistical Relational Learning Action and activity learning / Dynamics Travian: A massively multiplayer real-time strategy game Commercial game run by TravianGames GmbH ~3.000.000 players spread over different “worlds” Can we build a model of this world ? Can we use it for playing better ? [Thon et al. ECML 08] � 18
Robotics (and Vision) Contact: Luc De Raedt Winograd’s SHRDLU Put the blue pyramid on the block in the box Bring me the tea pot and the sugar The CLEVR Dataset and Variations
Verifying AI & ML systems Contact: Luc De Raedt, Hendrik Blockeel, Jesse Davis, & Wannes Meert Verification of software has a long tradition (eg model checking techniques) How to verify systems that learn ? that use AI ? Our approach — combined principles of probabilistic logics with verification Topics inductive synthesis of specifications Markov Decision Processes (& reinforcement leanring) Derive properties of learned systems … � 20
Knowledge-Base Systems Contact: Marc Denecker, Gerda Janssens IDP Advanced KBS system developed by group FO(.) language rooted in predicate logic and logic programming separation of domain knowledge and problem solving Language extensions to increase expressivity E.g. design patterns for FO(.) (past thesis) Better solvers and more inference methods E.g. a solver for rational numbers (past thesis) � 21
Knowledge-Base Systems Contact: Marc Denecker, Gerda Janssens Three themes for students : logical modeling of interesting AI problem + expressing AI knowledge domains logical analysis and implementation of software systems and tasks + software by applying inference on specifications Advanced algorithmics and implementation + extending/optimising the IDP software package. � 22
Applications of the Knowledge Base System Paradigm Logical modeling of AI problems Analysing medieval manuscripts vocabulary Vms { extern vocabulary V IsSource(Manuscript ) } theory Tms : Vms { { ! x : IsSource(x) <- ~ ? y : CopiedBy(y, x) & VariantIn(y) = VariantIn(x). } } term NbOfSources : Vms { #{ x : IsSource(x) } - monks copied texts } - resulting in variants (colors) procedure minSources(feature) { setvocabulary(feature, Vms) return minimize(Tms, feature, NbOfSources)[1] - reconstruct history } - special-purpose datamining-program: 400 lines of Perl, bugs - description problem in IDP: 15 lines, correct, somewhat faster � 23
Applications of the Knowledge Base System Paradigm Contact: Marc Denecker, Gerda Janssens Software = Knowledge Base + Logical Inference + User Interface E.g., An interactive configuration system for an insurance company AIM : Build cheap, correct, reusable, maintainable software from a logical specification � 24
Applications of the Knowledge Winning the RuleML Challenge Base System Paradigm Insurance application Propagation constraints and choices Fill out necessary values � 25
Knowledge-Base Systems Contact: Marc Denecker, Gerda Janssens Advanced algorithmics and implementation + extending/ optimising the IDP software package. help us win the next CP or ASP competition + E.g., structuring search space as a hierarchy of search problems + E.g., linear programming techniques in IDP + E.g., improved computation of definitions + E.g., algorithms for revision inference (updating solutions) � 26
Recommend
More recommend
