3,775 Hits in 8.1 sec

Synthesising Reinforcement Learning Policies Through Set-Valued Inductive Rule Learning [chapter]

Youri Coppens, Denis Steckelmacher, Catholijn M. Jonker, Ann Nowé
2021 Lecture Notes in Computer Science  
Today's advanced Reinforcement Learning algorithms produce black-box policies, that are often difficult to interpret and trust for a person.  ...  At the core of our approach is the fact that an RL process does not just learn a policy, a mapping from states to actions, but also produces extra meta-information, such as action values indicating the  ...  Policies Learnt Through Reinforcement Learning Reinforcement Learning (RL) tackles sequential decision making problems in an interactive setting [23] .  ... 
doi:10.1007/978-3-030-73959-1_15 fatcat:n433aoyl2jbgvf76jlgjnbyfwq

From explanation to synthesis: Compositional program induction for learning from demonstration [article]

Michael Burke, Svetlin Penkov, Subramanian Ramamoorthy
2019 arXiv   pre-print
This work introduces an approach to learning hybrid systems from demonstrations, with an emphasis on extracting models that are explicitly verifiable and easily interpreted by robot operators.  ...  We argue that computer program-like control systems are more interpretable than alternative end-to-end learning approaches, and that hybrid systems inherently allow for better generalisation across task  ...  Here, a task may require a robot to pass through a set of goal states, and then return by reversing through the same set of goal states.  ... 
arXiv:1902.10657v1 fatcat:tqiyn22nbbacjd3gfzkb5atnru

From Explanation to Synthesis: Compositional Program Induction for Learning from Demonstration

Michael Burke, Svetlin Valentinov Penkov, Subramanian Ramamoorthy
2019 Robotics: Science and Systems XV  
This work introduces an approach to learning hybrid systems from demonstrations, with an emphasis on extracting models that are explicitly verifiable and easily interpreted by robot operators.  ...  We argue that computer programlike control systems are more interpretable than alternative endto-end learning approaches, and that hybrid systems inherently allow for better generalisation across task  ...  Here, a task may require a robot to pass through a set of goal states, and then return by reversing through the same set of goal states.  ... 
doi:10.15607/rss.2019.xv.015 dblp:conf/rss/BurkePR19 fatcat:xi5yuy5mgzgklg5m24dpntqzcu

Towards Distributed Adaptive Control for Road Traffic Junction Signals using Learning Classifier Systems [chapter]

L. Bull, J. Sha'Aban, A. Tomlinson, J. D. Addison, B. G. Heydecker
2004 Studies in Fuzziness and Soft Computing  
Here the reinforcement learning is used to generate a single set of rules to control traffic at each of the individual junctions, while the genetic algorithm optimises rules across the network using these  ...  Reinforcement Component Bull and Hurst (2002) showed the importance of the learning rate β in ZCS.  ... 
doi:10.1007/978-3-540-39925-4_12 fatcat:x25ooplqgbgtrkf3srplf2olpi

Semi-supervised Learning From Demonstration Through Program Synthesis: An Inspection Robot Case Study

Simón C. Smith, Subramanian Ramamoorthy
2020 Electronic Proceedings in Theoretical Computer Science  
Through the design of a system that enables a robot to learn inspection strategies from a human operator, we present a hybrid semi-supervised system capable of learning interpretable and verifiable models  ...  The system induces an interpretable computer program of the demonstration that can be synthesised to produce novel inspection behaviours.  ...  End-to-end learning has facilitated one-shot learning for domain transfer from human video demonstration [70] and the use of reinforcement learning for optimised control policies [53, 73] .  ... 
doi:10.4204/eptcs.319.7 fatcat:s3qt66bhwrdnzmdiz5afv2g5ni

Certification of Iterative Predictions in Bayesian Neural Networks [article]

Matthew Wicker, Luca Laurenti, Andrea Patane, Nicola Paoletti, Alessandro Abate, Marta Kwiatkowska
2021 arXiv   pre-print
We use the lower bounds in the context of control and reinforcement learning to provide safety certification for given control policies, as well as to synthesize control policies that improve the certification  ...  On a set of benchmarks, we demonstrate that our framework can be employed to certify policies over BNNs predictions for problems of more than 10 dimensions, and to effectively synthesize policies that  ...  In scenarios such as sequential planning, time-series forecasting/control and model-based reinforcement learning, to evaluate a model w.r.t. a control policy (or strategy) one often needs to be able to  ... 
arXiv:2105.10134v2 fatcat:kgl63kzlxjablfvsssnvuimd2y

Making sense of raw input

Richard Evans, Matko Bošnjak, Lars Buesing, Kevin Ellis, David Pfau, Pushmeet Kohli, Marek Sergot
2021 Artificial Intelligence  
This way, we are able to jointly learn how to perceive (mapping raw sensory information to concepts) and apperceive (combining concepts into declarative rules). work that transforms raw unprocessed sensory  ...  Self-supervised learning has emerged as a well-defined sub-field within unsupervised learning [6, 7] .  ...  Relational reinforcement learning. Another approach that learns first-order rules to achieve strong generalisation is relational reinforcement learning [57, 58] .  ... 
doi:10.1016/j.artint.2021.103521 fatcat:lrnnltqicza3rguud5einblafu

Inductive logic programming at 30 [article]

Andrew Cropper, Sebastijan Dumančić, Richard Evans, Stephen H. Muggleton
2021 arXiv   pre-print
Inductive logic programming (ILP) is a form of logic-based machine learning. The goal is to induce a hypothesis (a logic program) that generalises given training examples.  ...  We focus on (i) new meta-level search methods, (ii) techniques for learning recursive programs, (iii) new approaches for predicate invention, and (iv) the use of different technologies.  ...  Law M ( ) Inductive learning of answer set programs. PhD thesis, Imperial College London, UK . Law M, Russo A, Broda K ( ) Inductive learning of answer set programs.  ... 
arXiv:2102.10556v2 fatcat:kv7ktjbajng6jjfae3sq3ubbmu

Advanced machine-learning techniques in drug discovery

Moe Elbadawi, Simon Gaisford, Abdul W. Basit
2020 Drug Discovery Today  
The popularity of machine learning (ML) across drug discovery continues to grow, yielding impressive results. As their use increases, so do their limitations become apparent.  ...  Rather, RL empirically learns the optimal decision to take through receiving reinforcement signals from its environment.  ...  For the smallest set of 353 compounds, the R 2 values when gradient boosting and multitask learning were used were 0.472 and 0.721, respectively.  ... 
doi:10.1016/j.drudis.2020.12.003 pmid:33290820 fatcat:es4pvfn6xjemnluslaowd3x75u

The Diagonal Model of Job Satisfaction and Motivation: Extracted from the Logical Comparison of Content and Process Theories

Zafarullah Sahito, Pertti Vaisanen
2017 International Journal of Higher Education  
Because the measurement of every employee action and organisational policy would be calculated through different values, the evaluation generated through this model consists of top to bottom and bottom  ...  to top diagonal approaches as well as the central value of the model.  ...  Monetary or extrinsic rewards can positively reinforce employees' work behaviour (Rudge, 2011) , and growth-need-strength to job design theories transformed the extant views and altered the research practice  ... 
doi:10.5430/ijhe.v6n3p209 fatcat:rvjrbtd5bjempilyrfmz7coe6u

Logic and the 2-Simplicial Transformer [article]

James Clift, Dmitry Doryn, Daniel Murfet, James Wallbridge
2019 arXiv   pre-print
We show that this architecture is a useful inductive bias for logical reasoning in the context of deep reinforcement learning.  ...  extension of the Transformer which includes a form of higher-dimensional attention generalising the dot-product attention, and uses this attention to update entity representations with tensor products of value  ...  Consider a deep reinforcement learning agent with a policy network parametrised by a vector of weights w ∈ R D and a sequence of full-episode rollouts of this policy in the environment, each of which either  ... 
arXiv:1909.00668v1 fatcat:2heqijpptfc5nfas3rjnzz73sm

Neural Algorithmic Reasoners are Implicit Planners [article]

Andreea Deac, Petar Veličković, Ognjen Milinković, Pierre-Luc Bacon, Jian Tang, Mladen Nikolić
2021 arXiv   pre-print
Implicit planning has emerged as an elegant technique for combining learned models of the world with end-to-end model-free reinforcement learning.  ...  We study the class of implicit planners inspired by value iteration, an algorithm that is guaranteed to yield perfect policies in fully-specified tabular environments.  ...  Andreea Deac and Petar Veličković wish to dedicate this paper to their family-for always being by their side, through thick and thin.  ... 
arXiv:2110.05442v1 fatcat:lyujfxha4nbrhmioa5fw3ie32q

The democratic transformation of public policy through community activism in Brazil

Rosana de Freitas Boullosa, Janaína Lopes Pereira Peres
2022 Policy and politics (Print)  
This reinforces the need to pursue a policy research agenda attentive to sociocentric experiences, ordinary actors and the emotions and values underlying public action.  ...  This article explores transformational change in public policy through community-based governance and the collective design of experience-oriented policy and action.  ...  We finally thank the three anonymous reviewers for Policy & Politics, for their very constructive feedback and inspirational insights on our initial submission.  ... 
doi:10.1332/030557321x16498834538186 fatcat:gtlaqawsbjas3mwdle4fof5pyy

Chinese teachers��� conceptions of assessment for and of learning: Six competing and complementary purposes

Gavin T.L. Brown, Lingbiao Gao, Kris Gritter
2015 Cogent Education  
As China continues to involve teachers in the implementation of an assessment for learning or formative assessment policy, a clearer understanding of how they conceive of the purposes and functions of  ...  Making use of inductive analyses and factor analytic techniques, variations in the constructs identified in teachers' thinking are identified and aligned across the study methods.  ...  This practice, despite contrary beliefs, reinforces the powerful effect of accountability evaluation of schools through student performance.  ... 
doi:10.1080/2331186x.2014.993836 fatcat:rsuko2wtefdjzjqmbwaughpqte

Tensions and opportunities in social prescribing. developing a framework to facilitate its implementation and evaluation in primary care: a realist review

Sara Calderón-Larrañaga, Yasmin Milner, Megan Clinch, Trisha Greenhalgh, Sarah Finer
2021 British Journal of General Practice Open  
Specific individual, interpersonal, organisational and policy resources are needed to ensure SP best practice in primary care.  ...  Findings where synthesised in a multi-level, dynamic and usable SP Framework.ConclusionOur realist review and resulting framework revealed that SP is not inherently advantageous.  ...  COMCs were then synthesised in an initial framework that was further developed through iterative discussions within the research team. 3 . 3 It comprises three settings (general practice, link workers  ... 
doi:10.3399/bjgpo.2021.0017 pmid:33849895 fatcat:z7ekjyqpyvhhxm4bzg4ayxspiq
« Previous Showing results 1 — 15 out of 3,775 results