Type of course: Seminar

Study programs:
Master Computer Science
Master Data Science
Master Software Systems Engineering

Offering chair:  Machine Learning and Reasoning (i6), RWTH

Description: Students will present recent papers in machine learning compiled by the professor. Topics include deep and reinforcement learning, transformer and GNN architectures, planning, LLMs, etc. 

Recommended prior knowledge: Bachelor degree in CS or Equivalent. Basic AI and ML courses.

Kickoff: 4/4/2024, 16:30h-18h, Room 228, Theaterstr. 35, 2nd floor

Format, organization, evaluation, dates, details: See slides below

Website: https://www-i6.informatik.rwth-aachen.de/~hector.geffner/Seminar-S2024.html

Introduction -- Slides 


Tentative list of papers
 
Theory Transformers, GNN, Embeddings, ...


The Expressive Power of Transformers with Chain of Thought. W Merrill, A Sabharwal. 2023, arXiv

On the Correspondence Between Monotonic Max-Sum GNNs and Datalog. DT Cucala, BC Grau, B Motik, EV Kostylev. 2023, arXiv

Explainable GNN-Based Models over Knowledge Graphs. DJ Tena Cucala, B Cuenca Grau, EV Kostylev, B Motik. 2022, ICLR

Recurrent Graph Neural Networks and Their Connections to Bisimulation and Logic. Maximilian Pfluger, David Tena Cucala, Egor V. Kostylev. AAAI 2024.

What relations are reliably embeddable in Euclidean space? R Bhattacharjee, S Dasgupta. 2020

Laplacian Eigenmaps for Dimensionality Reduction and Data Representation. M Belkin, P Niyogi. 2003, Neural Computation

Symbolic Learning

Learning to Play Using Low-Complexity Rule-Based Policies: Illustrations through Ms. Pac-Man I Szita, A Lorincz 2007, JAIR

Example-Guided Synthesis of Relational Queries. A Thakkar, A Naik, N Sands, R Alur, M Naik, M Raghothaman 2021, PLDI

Mobius: Synthesizing Relational Queries with Recursive and Invented Predicates A Thakkar, N Sands, G Petrou, R Alur, M Naik, M Raghothaman 2023, PACMPL

Learning programs by learning from failures. A Cropper, R Morel 2021, Machine Learning


Symbolic Regression, Variations

Learning Equations for Extrapolation and Control. S Sahoo, C Lampert, G Martius. 2018, ICML

Extrapolation and learning equations. G Martius, CH Lampert. 2017, ICLR


Neurosymbolic

INSIGHT: End-to-End Neuro-Symbolic Visual Reinforcement Learning with Language Explanations. L Luo, G Zhang, H Xu, Y Yang, C Fang, Q Li. 2024, arXiv

Discovering symbolic policies with deep reinforcement learning. M Landajuela, BK Petersen, S Kim, CP Santiago, R Glatt, N Mundhenk, JF Pettit, D Faissol. 2021, ICML

Scallop: A Language for Neurosymbolic Programming. Z Li, J Huang, M Naik. 2023, PACMPL.

Learning to Synthesize Programs as Interpretable and Generalizable Policies. D Trivedi, J Zhang, SH Sun, JJ Lim. 2021, NeurIPS

Synthesizing Programmatic Policies with Actor-Critic Algorithms and ReLU Networks. S Orfanos, LHS Lelis. 2023, arXiv

Reclaiming the Source of Programmatic Policies: Programmatic versus Latent Spaces. TH Carvalho, K Tjhia, L Lelis. 2024, ICLR

Solving olympiad geometry without human demonstrations. TH Trinh, Y Wu, QV Le, H He, T Luong. 2024, Nature

LLMs and the Abstraction and Reasoning Corpus: Successes, Failures, and the Importance of Object-based Representations. Y Xu, W Li, P Vaezipoor, S Sanner, EB Khalil. 2023

The CLRS Algorithmic Reasoning Benchmark. P Velickovic, AP Badia, D Budden, R Pascanu, A Banino, M Dashevskiy, R Hadsell, C Blundell. 2022, ICML

On the Markov Property of Neural Algorithmic Reasoning: Analyses and Methods. M Bohde, M Liu, A Saxton, S Ji. 2024, arXiv


Planning and Search

Optimize Planning Heuristics to Rank, not to Estimate Cost-to-Goal. L Chrestien, S Edelkamp, A Komenda, T Pevny. 2024, NeurIPS

Learning Discrete World Models for Classical Planning Problems. F Agostinelli, M Soltani. 2023, GenPlan

Acquiring planning domain models using LOCM. SN Cresswell, TL McCluskey, MM West 2013, KER

Solving Hard AI Planning Instances Using Curriculum-Driven Deep Reinforcement Learning. D Feng, CP Gomes, B Selman. 2020, arXiv


DRL: Deep Reinforcement Learning

Foundation Policies with Hilbert Representations. S Park, T Kreiman, S Levine. 2024, arXiv

Explore to Generalize in Zero-Shot RL. E Zisselman, I Lavie, D Soudry, A Tamar. 2024, NeurIPS

Transferable dynamics models for efficient object-oriented reinforcement learning. O Marom, B Rosman. 2024, AIJ

Behavior From the Void: Unsupervised Active Pre-Training. H Liu, P Abbeel. 2021, NeurIPS

Unveiling Options with Neural Network Decomposition. M Alikhasi, L Lelis. 2024, ICLR

Skill Machines: Temporal Logic Skill Composition in Reinforcement Learning. GN Tasse, D Jarvis, S James, B Rosman ICLR, 2024

Learning Hierarchical World Models with Adaptive Temporal Abstractions from Discrete Latent Dynamics. C Gumbsch, N Sajid, G Martius, MV Butz. 2024, ICLR

Mamba: Linear-Time Sequence Modeling with Selective State Spaces. Albert Gu, Tri Dao. 2023

Unsupervised Image Representation Learning with Deep Latent Particles. T Daniel, A Tamar 2022, arXiv

Entity-Centric Reinforcement Learning for Object Manipulation from Pixels. D Haramati, T Daniel, A Tamar. 2024, ICLR


Robotics

Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware. TZ Zhao, V Kumar, S Levine, C Finn. 2023, arXiv

M2T2: Multi-Task Masked Transformer for Object-centric Pick and Place. W Yuan, A Murali, A Mousavian, D Fox. 2023, arXiv

Constant-time Motion Planning with Anytime Refinement for Manipulation. I Mishani, H Feddock, M Likhachev. 2023, arXiv