the proceedings contain 9 papers. the topics discussed include: separation logics: semantics and proofs;on the computational content of intuitionistic modal proofs;a fresh look at relevant number theory;implementing i...
the proceedings contain 9 papers. the topics discussed include: separation logics: semantics and proofs;on the computational content of intuitionistic modal proofs;a fresh look at relevant number theory;implementing intermediate logics;automated proof search in intuitionistic sentential logic;implementing the Fatio protocol for multi-agent argumentation in LogiKEy;bisequent calculi for neutral free logic with definite descriptions;on regular relations in parametric array theories;and a proof-theoretical approach to some extensions of first order quantification.
After fact finding on the disruption bought by today’s Generative AI tools to the education system, we outline the advantages of joining the disruption, motivated by the natural synergies between today’s Generative ...
详细信息
As we progress towards real-world deployment, the critical need for interpretability in reinforcement learning algorithms grows more pivotal, ensuring the safety and reliability of intelligent agents. this paper tackl...
详细信息
ISBN:
(纸本)9783031700736;9783031700743
As we progress towards real-world deployment, the critical need for interpretability in reinforcement learning algorithms grows more pivotal, ensuring the safety and reliability of intelligent agents. this paper tackles the challenge of acquiring task specifications in linear temporal logicthrough expert demonstrations, aiming to alleviate the burdensome task of specification engineering. the rich semantics of temporal logics serve as an interpretable framework for delineating intricate, multi-stage tasks. We propose a method which iteratively learns a task specification and a nominal policy solving this task. In each iteration, the task specification is refined to better distinguish expert trajectories from trajectories sampled from the nominal policy. Withthis process we obtain a concise and interpretable task specification. Unlike previous work, our method is capable of learning directly from trajectories in the original state space and does not require predefined atomic propositions. We showcase the effectiveness of our method on multiple tasks in both an office and a Minecraft-inspired environment.
the logic of "Inferable" L-DINF has been recently proposed as a declarative framework to formally model via epistemic logicthe group dynamics of cooperative agents. In this paper, we extend the framework by...
详细信息
An open problem in artificial intelligence is how systems can flexibly learn discrete abstractions that are useful for solving inherently continuous problems. Previous work in computational neuroscience has considered...
详细信息
ISBN:
(纸本)9783031771378;9783031771385
An open problem in artificial intelligence is how systems can flexibly learn discrete abstractions that are useful for solving inherently continuous problems. Previous work in computational neuroscience has considered this functional integration of discrete and continuous variables during decision-making under the formalism of active inference [13,29]. However, their focus is on the expressive physical implementation of categorical decisions and the hierarchical mixed generative model is assumed to be known. As a consequence, it is unclear how this framework might be extended to the learning of appropriate coarse-grained variables for a given task. In light of this, we present a novel hierarchical hybrid active inference agent in which a high-level discrete active inference planner sits above a low-level continuous active inference controller. We make use of recent work in recurrent switching linear dynamical systems (rSLDS) which learn meaningful discrete representations of complex continuous dynamics via piecewise linear decomposition [22]. the representations learnt by the rSLDS inform the structure of the hybrid decision-making agent and allow us to (1) lift decision-making into the discrete domain enabling us to exploit information-theoretic exploration bonuses (2) specify temporally-abstracted sub-goals in a method reminiscent of the options framework [34] and (3) 'cache' the approximate solutions to low-level problems in the discrete planner. We apply our model to the sparse Continuous Mountain Car task, demonstrating fast system identification via enhanced exploration and successful planning through the delineation of abstract sub-goals.
logic has been proved useful to model various aspects of the reasoning process of agents and multi-agentsystems (MAS). In this paper, we report about the last advances over a line of work aimed to explore social aspe...
详细信息
ISBN:
(数字)9783030974572
ISBN:
(纸本)9783030974572;9783030974565
logic has been proved useful to model various aspects of the reasoning process of agents and multi-agentsystems (MAS). In this paper, we report about the last advances over a line of work aimed to explore social aspects of such systems. the objective is to formally model (aspects of) the group dynamics of cooperative agents. We have proposed and here extend a particular logical framework (the logic of "Inferable" L-DINF), where a group of cooperative agents can jointly perform actions. I.e., at least one agent of the group can perform the action, either withthe approval of the group or on behalf of the group. We have been able to take into consideration actions' cost and the preferences that each agent can have for what concerns performing each action. Our focus here is on: (i) explainability, i.e., the syntax of our logic is especially devised to make it possible to transpose a proof into a natural language explanation, in the perspective of trustworthy Artificial Intelligence;(ii) the capability to construct and execute joint plans within a group of agents;(iii) the formalization of aspects of the theory of Mind, which is an important social-cognitive skill involving the ability to attribute mental states, including emotions, desires, beliefs, and knowledge to oneself and to others, and to reason about the practical consequences of such mental states;such capability is very relevant when agents have to interact with humans, and in particular in robotic applications;(iv) connection between theory and practice, so as to make our logic actually usable by a system's designers.
the proceedings contain 11 papers. the special focus in this conference is on Software Verification and Formal Methods for ML-Enables Autonomous systems. the topics include: Neural Network Precision Tuning Using Stoch...
ISBN:
(纸本)9783031212215
the proceedings contain 11 papers. the special focus in this conference is on Software Verification and Formal Methods for ML-Enables Autonomous systems. the topics include: Neural Network Precision Tuning Using Stochastic Arithmetic;MLTL multi-type (MLTLM): A logic for Reasoning About Signals of Different Types;a Cascade of Checkers for Run-time Certification of Local Robustness;CEG4N: Counter-Example Guided Neural Network Quantization Refinement;minimal multi-Layer Modifications of Deep Neural Networks;differentiable logics for Neural Network Training and Verification;neural Networks in Imandra: Matrix Representation as a Verification Choice;self-correcting Neural Networks for Safe Classification;formal Specification for Learning-Enabled Autonomous systems;verified Numerical Methods for Ordinary Differential Equations.
agents and multi-agentsystems (MAS) are a technology that has many fields of application, which extend also to human sciences and where computationallogic has been widely applied. In this paper, we join together two...
详细信息
Deep learning has been increasingly successful in the last few years, but its inherent limitations have recently become more evident, especially with respect to explainability and interpretability. Neural-symbolic app...
详细信息
ISBN:
(纸本)9783031155659;9783031155642
Deep learning has been increasingly successful in the last few years, but its inherent limitations have recently become more evident, especially with respect to explainability and interpretability. Neural-symbolic approaches to inductive logic programming have been recently proposed to synergistically combine the advantages of inductive logic programming in terms of explainability and interpretability withthe characteristic capability of deep learning to treat noisy, erroneous, and non-logical data. this paper surveys and briefly compares four relevant neural-symbolic approaches to inductive logic programming that have been proposed in the last five years and that use templates as an effective basis to learn logic programs from data.
In this paper, we further advance a line of work aimed to formally model via epistemic logic (aspects of) the group dynamics of cooperative agents. In fact, we have previously proposed and here extend a particular log...
详细信息
暂无评论