Filters








229,431 Hits in 4.7 sec

AMLSI: A Novel Accurate Action Model Learning Algorithm [article]

Maxence Grand, Humbert Fiorino, Damien Pellier
2020 arXiv   pre-print
Unlike other algorithms, we show that AMLSI is able to lift this lock by learning domains from partial and noisy observations with sufficient accuracy to allow planners to solve new problems.  ...  AMLSI proceeds by trial and error: it queries the system to learn with randomly generated action sequences, and it observes the state transitions of the system, then AMLSI returns a PDDL domain corresponding  ...  In addition, this refinement is able to learn some domains (Blocksworld and Gripper) when observations are partial and noiseless.  ... 
arXiv:2011.13277v1 fatcat:epdg76a2jvbrpebxcn6o2pjkn4

Learning hierarchical task network domains from partially observed plan traces

Hankz Hankui Zhuo, Héctor Muñoz-Avila, Qiang Yang
2014 Artificial Intelligence  
HTNLearn can learn methods and action models simultaneously from partially observed plan traces (i.e., plan traces where the intermediate states are partially observable).  ...  We test HTNLearn in several HTN domains. The experimental results show that our algorithm HTNLearn is both effective and efficient.  ...  Héctor Muñoz-Avila's work is supported in part by NSF grant 0642882. Qiang Yang thanks China National 973 project 2014CB340304 for kind support.  ... 
doi:10.1016/j.artint.2014.04.003 fatcat:nvuecit6izb2hp4vrg5p6wn6lm

Learning STRIPS Action Models with Classical Planning [article]

Diego Aineto, Sergio Jiménez, Eva Onaindia
2019 arXiv   pre-print
Moreover, the compilation accepts partially specified action models and it can be used to validate whether the observation of a plan execution follows a given STRIPS action model, even if this model is  ...  This paper presents a novel approach for learning STRIPS action models from examples that compiles this inductive learning task into a classical planning task.  ...  Diego Aineto is partially supported by the FPU16/03184 and Sergio Jiménez by the RYC15/18009, both programs funded by the Spanish government.  ... 
arXiv:1903.01153v1 fatcat:5hk2l3ed45dnhb6xaoi7j73rfq

Learning action models with minimal observability

Diego Aineto, Sergio Jiménez Celorrio, Eva Onaindia
2019 Artificial Intelligence  
Learning action models with minimal observability.  ...  FAMA is exhaustively evaluated over a wide range of IPC domains and its performance is compared to ARMS, a state-of-the-art benchmark in action model learning.  ...  Diego Aineto is partially supported by the FPU16/03184 and Sergio Jiménez by the RYC15/18009, both programs funded by the Spanish government.  ... 
doi:10.1016/j.artint.2019.05.003 fatcat:rs65cvp37ffi3chui633puvue4

A Review on Learning Planning Action Models for Socio-Communicative HRI [article]

Ankuj Arora and Humbert Fiorino and Damien Pellier and Sylvie Pesty
2018 arXiv   pre-print
In the same domain, this action sequence that stems from plan execution can be represented as a trace.  ...  Thus, this entire interaction can be treated as a sequence of actions propelling the interaction from its initial to goal state, also known as a plan in the domain of AI planning.  ...  12, 9] system. • Deterministic effects, partial state observability: In this family, the system may be in one of a set of 'belief states' after the execution of each action.  ... 
arXiv:1810.09245v1 fatcat:5ga2klfbcfaife7wylepgsxtzq

STRIPS Action Discovery [article]

Alejandro Suárez-Hernández and Javier Segovia-Aguas and Carme Torras and Guillem Alenyà
2021 arXiv   pre-print
These approaches can synthesize action schemas in Planning Domain Definition Language (PDDL) from a set of execution traces each consisting, at least, of an initial and final state.  ...  In addition, we contribute with a compilation to classical planning that mitigates the problem of learning static predicates in the action model preconditions, exploits the capabilities of SAT planners  ...  Learning action models consists in automatically generating the domain that rules any problem in a given environment, e.g. the move action with its preconditions and effects in the visitall domain.  ... 
arXiv:2001.11457v3 fatcat:y2tr2dhcgbdovfqvjj2ggs7imq

A review of machine learning for automated planning

Sergio Jiménez, Tomás De La Rosa, Susana Fernández, Fernando Fernández, Daniel Borrajo
2012 Knowledge engineering review (Print)  
On the one hand, the definition of accurate action models for planning is still a bottleneck. On the other hand, off-the-shelf planners fail to scale up and to provide good solutions in many domains.  ...  In these problematic domains, planners can exploit domain-specific control knowledge to improve their performance in terms of both speed and quality of the solutions.  ...  Learning Deterministic Models in Partially Observable Environments In this scenario actions also present deterministic effects, but the state is not fully observable.  ... 
doi:10.1017/s026988891200001x fatcat:slnkph7hyve3higgkjyd3wydlu

Learning STRIPS Operators from Noisy and Incomplete Observations [article]

Kira Mourao, Luke S. Zettlemoyer, Ronald P. A. Petrick, Mark Steedman
2012 arXiv   pre-print
Even in standard STRIPS domains, existing approaches cannot learn from noisy, incomplete observations typical of real-world domains.  ...  We propose a method which learns STRIPS action models in such domains, by decomposing the problem into first learning a transition function between states in the form of a set of classifiers, and then  ...  This work was partially funded by the European Commission through the EU Cognitive Systems project Xperience (FP7-ICT-270273) and the UK EPSRC/MRC through the Neuroinformatics and Computational Neuroscience  ... 
arXiv:1210.4889v1 fatcat:k66lyi4poff6bhdjbll32sedom

A review of learning planning action models

Ankuj Arora, Humbert Fiorino, Damien Pellier, Marc Métivier, Sylvie Pesty
2018 Knowledge engineering review (Print)  
In such domains, planners now leverage recent advancements in machine learning to learn action models, that is, blueprints of all the actions whose execution effectuates transitions in the system.  ...  However, as we move from toy domains closer to the complex real world, these actions become increasingly difficult to codify.  ...  Supervised Learning Based Approaches In (Mourão et al. (2008) ), the authors learn the effects of an agent's actions given a set of actions and their preconditions, in a partially observable and noisy  ... 
doi:10.1017/s0269888918000188 fatcat:yjwlfpz4efhnrbhxczr2mxh4km

The Thing That We Tried Didn't Work Very Well : Deictic Representation in Reinforcement Learning [article]

Sarah Finney, Natalia Gardiol, Leslie Pack Kaelbling, Tim Oates
2012 arXiv   pre-print
We conclude with a discussion of possible causes of these results and strategies for more effective learning in domains with objects.  ...  In this paper we explore the effectiveness of two forms of deictic representation and a naïve propositional representation in a simple blocks-world domain.  ...  Thus, no matter how large the trees get, the agent is still trying to learn in a partially observable domain.  ... 
arXiv:1301.0567v1 fatcat:opmeisb3qzdwnjaujalxlqvarq

Algorithmic Improvements for Deep Reinforcement Learning Applied to Interactive Fiction

Vishal Jain, William Fedus, Hugo Larochelle, Doina Precup, Marc G. Bellemare
2020 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
We first propose a contextualisation mechanism, based on accumulated reward, which simplifies the learning problem and mitigates partial observability.  ...  Their state and action spaces are combinatorially large, their reward function is sparse, and they are partially observable: the agent is informed of the consequences of its actions through textual feedback  ...  Most actions have no effect in a given state; • Memory as state. Remembering key past events is often sufficient to deal with partial observability.  ... 
doi:10.1609/aaai.v34i04.5857 fatcat:tlufceqrlfbunkccc6yeor2wx4

Domain-independent generation and classification of behavior traces [article]

Daniel Borrajo, Manuela Veloso
2020 arXiv   pre-print
In this work, the observer agent has partial and noisy observability of the environment (state and actions of the other agents).  ...  An observer agent tries to learn characterizing other agents by observing their behavior when taking actions in a given environment.  ...  any way for evaluating the merits of participating in any transaction, and shall not constitute a solicitation under any jurisdiction or to any person, if such solicitation under  ... 
arXiv:2011.02918v1 fatcat:z2522b2bfva2nnivu42e65p3ky

Reinforcement Learning with Efficient Active Feature Acquisition [article]

Haiyan Yin and Yingzhen Li and Sinno Jialin Pan and Cheng Zhang and Sebastian Tschiatschek
2020 arXiv   pre-print
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  ...  Solving real-life sequential decision making problems under partial observability involves an exploration-exploitation problem.  ...  the decision making with partial observability in Sepsis domain.  ... 
arXiv:2011.00825v1 fatcat:kjhlzuoe2ngplle2qikzgovfe4

Algorithmic Improvements for Deep Reinforcement Learning applied to Interactive Fiction [article]

Vishal Jain, William Fedus, Hugo Larochelle, Doina Precup, Marc G. Bellemare
2019 arXiv   pre-print
We first propose a contextualisation mechanism, based on accumulated reward, which simplifies the learning problem and mitigates partial observability.  ...  Their state and action spaces are combinatorially large, their reward function is sparse, and they are partially observable: the agent is informed of the consequences of its actions through textual feedback  ...  Most actions have no effect in a given state; • Memory as state. Remembering key past events is often sufficient to deal with partial observability.  ... 
arXiv:1911.12511v1 fatcat:jbjomsgubneyhaksbbwgyus33e

Learning Policies with External Memory [article]

Leonid Peshkin, Nicolas Meuleau, Leslie Kaelbling
2001 arXiv   pre-print
In order for an agent to perform well in partially observable domains, it is usually necessary for actions to depend on the history of observations.  ...  In this case, we need to learn a reactive policy in a highly non-Markovian domain.  ...  Acknowledgments This work was supported in part by DARPA/Rome Labs Planning Initiative grant F30602-95-1-0020.  ... 
arXiv:cs/0103003v1 fatcat:lprtd6kmknconlipih6anc7zqm
« Previous Showing results 1 — 15 out of 229,431 results