Pub Date : 2024-10-22DOI: 10.1016/j.artint.2024.104241
Cornelius Brand , Robert Ganian , Subrahmanyam Kalyanasundaram , Fionn Mc Inerney
Atomic congestion games are a classic topic in network design, routing, and algorithmic game theory, and are capable of modeling congestion and flow optimization tasks in various application areas. While both the price of anarchy for such games as well as the computational complexity of computing their Nash equilibria are by now well-understood, the computational complexity of computing a system-optimal set of strategies—that is, a centrally planned routing that minimizes the average cost of agents—is severely understudied in the literature. We close this gap by identifying the exact boundaries of tractability for the problem through the lens of the parameterized complexity paradigm. After showing that the problem remains highly intractable even on extremely simple networks, we obtain a set of results which demonstrate that the structural parameters which control the computational (in)tractability of the problem are not vertex-separator based in nature (such as, e.g., treewidth), but rather based on edge separators. We conclude by extending our analysis towards the (even more challenging) min-max variant of the problem.
{"title":"The complexity of optimizing atomic congestion","authors":"Cornelius Brand , Robert Ganian , Subrahmanyam Kalyanasundaram , Fionn Mc Inerney","doi":"10.1016/j.artint.2024.104241","DOIUrl":"10.1016/j.artint.2024.104241","url":null,"abstract":"<div><div>Atomic congestion games are a classic topic in network design, routing, and algorithmic game theory, and are capable of modeling congestion and flow optimization tasks in various application areas. While both the price of anarchy for such games as well as the computational complexity of computing their Nash equilibria are by now well-understood, the computational complexity of computing a <em>system-optimal</em> set of strategies—that is, a centrally planned routing that minimizes the average cost of agents—is severely understudied in the literature. We close this gap by identifying the exact boundaries of tractability for the problem through the lens of the parameterized complexity paradigm. After showing that the problem remains highly intractable even on extremely simple networks, we obtain a set of results which demonstrate that the structural parameters which control the computational (in)tractability of the problem are not vertex-separator based in nature (such as, e.g., treewidth), but rather based on edge separators. We conclude by extending our analysis towards the (even more challenging) min-max variant of the problem.</div></div>","PeriodicalId":8434,"journal":{"name":"Artificial Intelligence","volume":"338 ","pages":"Article 104241"},"PeriodicalIF":5.1,"publicationDate":"2024-10-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142533879","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-10-22DOI: 10.1016/j.artint.2024.104239
Chen Li, Yoshihiro Yamanishi
generation of bioactive and drug-like hit molecules is a pivotal goal in computer-aided drug discovery. While artificial intelligence (AI) has proven adept at generating molecules with desired chemical properties, previous studies often overlook the influence of disease-specific cellular environments. This study introduces GxVAEs, a novel AI-driven deep generative model designed to produce hit molecules from transcriptome profiles using dual variational autoencoders (VAEs). The first VAE, ProfileVAE, extracts latent features from transcriptome profiles to guide the second VAE, MolVAE, in generating hit molecules. GxVAEs aim to bridge the gap between molecule generation and the biological context of disease, producing molecules that are biologically relevant within specific cellular environments or pathological conditions. Experimental results and case studies focused on hit molecule generation demonstrate that GxVAEs surpass current state-of-the-art methods, in terms of reproducibility of known ligands. This approach is expected to effectively find potential molecular structures with bioactivities across diverse disease contexts.
{"title":"AI-driven transcriptome profile-guided hit molecule generation","authors":"Chen Li, Yoshihiro Yamanishi","doi":"10.1016/j.artint.2024.104239","DOIUrl":"10.1016/j.artint.2024.104239","url":null,"abstract":"<div><div><span><math><mi>D</mi><mi>e</mi><mspace></mspace><mi>n</mi><mi>o</mi><mi>v</mi><mi>o</mi></math></span> generation of bioactive and drug-like hit molecules is a pivotal goal in computer-aided drug discovery. While artificial intelligence (AI) has proven adept at generating molecules with desired chemical properties, previous studies often overlook the influence of disease-specific cellular environments. This study introduces GxVAEs, a novel AI-driven deep generative model designed to produce hit molecules from transcriptome profiles using dual variational autoencoders (VAEs). The first VAE, ProfileVAE, extracts latent features from transcriptome profiles to guide the second VAE, MolVAE, in generating hit molecules. GxVAEs aim to bridge the gap between molecule generation and the biological context of disease, producing molecules that are biologically relevant within specific cellular environments or pathological conditions. Experimental results and case studies focused on hit molecule generation demonstrate that GxVAEs surpass current state-of-the-art methods, in terms of reproducibility of known ligands. This approach is expected to effectively find potential molecular structures with bioactivities across diverse disease contexts.</div></div>","PeriodicalId":8434,"journal":{"name":"Artificial Intelligence","volume":"338 ","pages":"Article 104239"},"PeriodicalIF":5.1,"publicationDate":"2024-10-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142534018","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-10-18DOI: 10.1016/j.artint.2024.104236
Juan Pablo Aguilera , Martín Diéguez , David Fernández-Duque , Brett McLean
We investigate a version of linear temporal logic whose propositional fragment is Gödel–Dummett logic (which is well known both as a superintuitionistic logic and a t-norm fuzzy logic). We define the logic using two natural semantics: first a real-valued semantics, where statements have a degree of truth in the real unit interval, and second a ‘bi-relational’ semantics. We then show that these two semantics indeed define one and the same logic: the statements that are valid for the real-valued semantics are the same as those that are valid for the bi-relational semantics. This Gödel temporal logic does not have any form of the finite model property for these two semantics: there are non-valid statements that can only be falsified on an infinite model. However, by using the technical notion of a quasimodel, we show that every falsifiable statement is falsifiable on a finite quasimodel, yielding an algorithm for deciding if a statement is valid or not. Later, we strengthen this decidability result by giving an algorithm that uses only a polynomial amount of memory, proving that Gödel temporal logic is PSPACE-complete. We also provide a deductive calculus for Gödel temporal logic, and show this calculus to be sound and complete for the above-mentioned semantics, so that all (and only) the valid statements can be proved with this calculus.
我们研究了线性时间逻辑的一个版本，其命题片段是哥德尔-杜梅特逻辑（作为超直觉逻辑和 t 规范模糊逻辑而闻名）。我们使用两种自然语义来定义该逻辑：第一种是实值语义，其中语句在实数单位区间内具有真度；第二种是 "双关系 "语义。然后，我们证明这两种语义确实定义了同一个逻辑：实值语义中有效的语句与双关系语义中有效的语句是一样的。对于这两种语义，哥德尔时间逻辑不具有任何形式的有限模型属性：有一些非有效语句只能在无限模型上被证伪。然而，通过使用准模型的技术概念，我们证明了每个可证伪语句在有限准模型上都是可证伪的，从而得出了一种判定语句是否有效的算法。随后，我们通过给出一种只使用多项式内存的算法来加强这一可证实性结果，从而证明哥德尔时间逻辑是 PSPACE-完备的。我们还为哥德尔时间逻辑提供了一个演绎微积分，并证明这个微积分对于上述语义是健全和完备的，因此所有（且仅有）有效语句都可以用这个微积分证明。
{"title":"Gödel–Dummett linear temporal logic","authors":"Juan Pablo Aguilera , Martín Diéguez , David Fernández-Duque , Brett McLean","doi":"10.1016/j.artint.2024.104236","DOIUrl":"10.1016/j.artint.2024.104236","url":null,"abstract":"<div><div>We investigate a version of linear temporal logic whose propositional fragment is Gödel–Dummett logic (which is well known both as a superintuitionistic logic and a t-norm fuzzy logic). We define the logic using two natural semantics: first a real-valued semantics, where statements have a degree of truth in the real unit interval, and second a ‘bi-relational’ semantics. We then show that these two semantics indeed define one and the same logic: the statements that are valid for the real-valued semantics are the same as those that are valid for the bi-relational semantics. This Gödel temporal logic does not have any form of the finite model property for these two semantics: there are non-valid statements that can only be falsified on an infinite model. However, by using the technical notion of a quasimodel, we show that every falsifiable statement is falsifiable on a finite quasimodel, yielding an algorithm for deciding if a statement is valid or not. Later, we strengthen this decidability result by giving an algorithm that uses only a polynomial amount of memory, proving that Gödel temporal logic is PSPACE-complete. We also provide a deductive calculus for Gödel temporal logic, and show this calculus to be sound and complete for the above-mentioned semantics, so that all (and only) the valid statements can be proved with this calculus.</div></div>","PeriodicalId":8434,"journal":{"name":"Artificial Intelligence","volume":"338 ","pages":"Article 104236"},"PeriodicalIF":5.1,"publicationDate":"2024-10-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142577891","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-10-18DOI: 10.1016/j.artint.2024.104233
Yanjun Li , Yanjing Wang
Automated planning in AI and the logics of knowing how have close connections. In the recent literature, various planning-based know-how logics have been proposed and studied, making use of several notions of planning in AI. In this paper, we explore the reverse direction by using a multi-agent logic of knowing how to do know-how-based planning via model checking and theorem proving/satisfiability checking. Based on our logical framework, we propose two new classes of related planning problems: higher-order epistemic planning and meta-level epistemic planning, which generalize the current genre of epistemic planning in the literature. The former is for planning about planning, i.e., planning with higher-order goals that are again about epistemic planning, e.g., finding a plan for an agent to make sure p such that the adversary does not know how to make p false in the future. The latter is about planning at the meta-level by abstract reasoning combining knowledge-how from different agents, e.g., given that i knows how to prove a lemma and i knows j knows how to prove the theorem once the lemma is proved, we should derive that i knows how to let j knows how to prove the theorem. To make these possible, our framework features not only the operators of know-that and know-how but also a temporal operator □, which can help in capturing both the local and global knowledge-how. We axiomatize this powerful logic over finite models with perfect recall and show its decidability. We also give a PTIME algorithm for the model checking problem over finite models.
人工智能中的自动规划与 "知道如何 "的逻辑有着密切的联系。在最近的文献中，人们利用人工智能中的几个规划概念，提出并研究了各种基于规划的诀窍逻辑。在本文中，我们从相反的方向进行了探索，通过模型检查和定理证明/可满足性检查，利用多代理的 "知道如何 "逻辑来进行基于 "诀窍 "的规划。基于我们的逻辑框架，我们提出了两类新的相关规划问题：高阶认识规划和元级认识规划，它们概括了目前文献中的认识规划流派。前者是关于规划的规划，即具有高阶目标的规划，而高阶目标又是关于认识论规划的，例如，为代理人找到一个确保 p 的规划，使对手不知道如何在未来使 p 变成假的。后者则是通过结合不同代理的知识诀窍进行抽象推理，在元层面上进行规划，例如，考虑到 i 知道如何证明一个lemma，并且 i 知道 j 知道如何证明该定理，一旦该lemma 被证明，我们就应该推导出 i 知道如何让 j 知道如何证明该定理。为了使这些成为可能，我们的框架不仅有 "知道--那 "和 "知道--诀窍 "算子，还有一个时态算子□，它有助于捕捉局部和全局的知识诀窍。我们在具有完全召回能力的有限模型上对这一强大的逻辑进行了公理化，并证明了它的可解性。我们还给出了有限模型上模型检查问题的 PTIME 算法。
{"title":"Knowing how to plan about planning: Higher-order and meta-level epistemic planning","authors":"Yanjun Li , Yanjing Wang","doi":"10.1016/j.artint.2024.104233","DOIUrl":"10.1016/j.artint.2024.104233","url":null,"abstract":"<div><div>Automated planning in AI and the logics of knowing how have close connections. In the recent literature, various <em>planning-based know-how logics</em> have been proposed and studied, making use of several notions of planning in AI. In this paper, we explore the <em>reverse</em> direction by using a multi-agent logic of knowing how to do <em>know-how-based planning</em> via model checking and theorem proving/satisfiability checking. Based on our logical framework, we propose two new classes of related planning problems: <em>higher-order epistemic planning</em> and <em>meta-level epistemic planning</em>, which generalize the current genre of epistemic planning in the literature. The former is for planning about planning, i.e., planning with higher-order goals that are again about epistemic planning, e.g., finding a plan for an agent to make sure <em>p</em> such that the adversary does not know how to make <em>p</em> false in the future. The latter is about planning at the meta-level by abstract reasoning combining knowledge-how from different agents, e.g., given that <em>i</em> knows how to prove a lemma and <em>i</em> knows <em>j</em> knows how to prove the theorem once the lemma is proved, we should derive that <em>i</em> knows how to let <em>j</em> knows how to prove the theorem. To make these possible, our framework features not only the operators of know-that and know-how but also a temporal operator □, which can help in capturing both the <em>local</em> and <em>global</em> knowledge-how. We axiomatize this powerful logic over finite models with perfect recall and show its decidability. We also give a PTIME algorithm for the model checking problem over finite models.</div></div>","PeriodicalId":8434,"journal":{"name":"Artificial Intelligence","volume":"337 ","pages":"Article 104233"},"PeriodicalIF":5.1,"publicationDate":"2024-10-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142527870","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-10-11DOI: 10.1016/j.artint.2024.104232
Kai Li , Hang Xu , Haobo Fu , Qiang Fu , Junliang Xing
Strategic decision-making in imperfect-information games is an important problem in artificial intelligence. Counterfactual regret minimization (CFR), a family of iterative algorithms, has been the workhorse for solving these types of games since its inception. In recent years, a series of novel CFR variants have been proposed, significantly improving the convergence rate of vanilla CFR. However, most of these new variants are hand-designed by researchers through trial and error, often based on different motivations, which generally requires a tremendous amount of effort and insight. This work proposes AutoCFR, a systematic framework that meta-learns novel CFR algorithms through evolution, easing the burden of manual algorithm design. We first design a search language that is rich enough to represent various CFR variants. We then exploit a scalable regularized evolution algorithm with a set of acceleration techniques to efficiently search over the combinatorial space of algorithms defined by this language. The learned novel CFR algorithm can generalize to new imperfect-information games not seen during training and performs on par with or better than existing state-of-the-art CFR variants. In addition to superior empirical performance, we also theoretically show that the learned algorithm converges to an approximate Nash equilibrium. Extensive experiments across diverse imperfect-information games highlight the scalability, extensibility, and generalizability of AutoCFR, establishing it as a general-purpose framework for solving imperfect-information games.
{"title":"Automatically designing counterfactual regret minimization algorithms for solving imperfect-information games","authors":"Kai Li , Hang Xu , Haobo Fu , Qiang Fu , Junliang Xing","doi":"10.1016/j.artint.2024.104232","DOIUrl":"10.1016/j.artint.2024.104232","url":null,"abstract":"<div><div>Strategic decision-making in imperfect-information games is an important problem in artificial intelligence. Counterfactual regret minimization (CFR), a family of iterative algorithms, has been the workhorse for solving these types of games since its inception. In recent years, a series of novel CFR variants have been proposed, significantly improving the convergence rate of vanilla CFR. However, most of these new variants are hand-designed by researchers through trial and error, often based on different motivations, which generally requires a tremendous amount of effort and insight. This work proposes AutoCFR, a systematic framework that meta-learns novel CFR algorithms through evolution, easing the burden of manual algorithm design. We first design a search language that is rich enough to represent various CFR variants. We then exploit a scalable regularized evolution algorithm with a set of acceleration techniques to efficiently search over the combinatorial space of algorithms defined by this language. The learned novel CFR algorithm can generalize to new imperfect-information games not seen during training and performs on par with or better than existing state-of-the-art CFR variants. In addition to superior empirical performance, we also theoretically show that the learned algorithm converges to an approximate Nash equilibrium. Extensive experiments across diverse imperfect-information games highlight the scalability, extensibility, and generalizability of AutoCFR, establishing it as a general-purpose framework for solving imperfect-information games.</div></div>","PeriodicalId":8434,"journal":{"name":"Artificial Intelligence","volume":"337 ","pages":"Article 104232"},"PeriodicalIF":5.1,"publicationDate":"2024-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142527869","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-10-02DOI: 10.1016/j.artint.2024.104227
Pedro Zuidberg Dos Martires , Luc De Raedt , Angelika Kimmig
Over the past three decades, the logic programming paradigm has been successfully expanded to support probabilistic modeling, inference and learning. The resulting paradigm of probabilistic logic programming (PLP) and its programming languages owes much of its success to a declarative semantics, the so-called distribution semantics. However, the distribution semantics is limited to discrete random variables only. While PLP has been extended in various ways for supporting hybrid, that is, mixed discrete and continuous random variables, we are still lacking a declarative semantics for hybrid PLP that not only generalizes the distribution semantics and the modeling language but also the standard inference algorithm that is based on knowledge compilation. We contribute the measure semantics together with the hybrid PLP language DC-ProbLog (where DC stands for distributional clauses) and its inference engine infinitesimal algebraic likelihood weighting (IALW). These have the original distribution semantics, standard PLP languages such as ProbLog, and standard inference engines for PLP based on knowledge compilation as special cases. Thus, we generalize the state of the art of PLP towards hybrid PLP in three different aspects: semantics, language and inference. Furthermore, IALW is the first inference algorithm for hybrid probabilistic programming based on knowledge compilation.
{"title":"Declarative probabilistic logic programming in discrete-continuous domains","authors":"Pedro Zuidberg Dos Martires , Luc De Raedt , Angelika Kimmig","doi":"10.1016/j.artint.2024.104227","DOIUrl":"10.1016/j.artint.2024.104227","url":null,"abstract":"<div><div>Over the past three decades, the logic programming paradigm has been successfully expanded to support probabilistic modeling, inference and learning. The resulting paradigm of probabilistic logic programming (PLP) and its programming languages owes much of its success to a declarative semantics, the so-called distribution semantics. However, the distribution semantics is limited to discrete random variables only. While PLP has been extended in various ways for supporting hybrid, that is, mixed discrete and continuous random variables, we are still lacking a declarative semantics for hybrid PLP that not only generalizes the distribution semantics and the modeling language but also the standard inference algorithm that is based on knowledge compilation. We contribute the <em>measure semantics</em> together with the hybrid PLP language DC-ProbLog (where DC stands for distributional clauses) and its inference engine <em>infinitesimal algebraic likelihood weighting</em> (IALW). These have the original distribution semantics, standard PLP languages such as ProbLog, and standard inference engines for PLP based on knowledge compilation as special cases. Thus, we generalize the state of the art of PLP towards hybrid PLP in three different aspects: semantics, language and inference. Furthermore, IALW is the first inference algorithm for hybrid probabilistic programming based on knowledge compilation.</div></div>","PeriodicalId":8434,"journal":{"name":"Artificial Intelligence","volume":"337 ","pages":"Article 104227"},"PeriodicalIF":5.1,"publicationDate":"2024-10-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142421611","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-10-02DOI: 10.1016/j.artint.2024.104231
Yossi Azar , Amos Fiat , Federico Fusco
We study sequential bilateral trade where sellers and buyers valuations are completely arbitrary (i.e., determined by an adversary). Sellers and buyers are strategic agents with private valuations for the good and the goal is to design a mechanism that maximizes efficiency (or gain from trade) while being incentive compatible, individually rational and budget balanced. In this paper we consider gain from trade, which is harder to approximate than social welfare.
We consider a variety of feedback scenarios and distinguish the cases where the mechanism posts one price and when it can post different prices for buyer and seller. We show several surprising results about the separation between the different scenarios. In particular we show that (a) it is impossible to achieve sublinear α-regret for any , (b) but with full feedback sublinear 2-regret is achievable; (c) with a single price and partial feedback one cannot get sublinear α regret for any constant α (d) nevertheless, posting two prices even with one-bit feedback achieves sublinear 2-regret, and (e) there is a provable separation in the 2-regret bounds between full and partial feedback.
{"title":"An α-regret analysis of adversarial bilateral trade","authors":"Yossi Azar , Amos Fiat , Federico Fusco","doi":"10.1016/j.artint.2024.104231","DOIUrl":"10.1016/j.artint.2024.104231","url":null,"abstract":"<div><div>We study sequential bilateral trade where sellers and buyers valuations are completely arbitrary (<em>i.e.</em>, determined by an adversary). Sellers and buyers are strategic agents with private valuations for the good and the goal is to design a mechanism that maximizes efficiency (or gain from trade) while being incentive compatible, individually rational and budget balanced. In this paper we consider gain from trade, which is harder to approximate than social welfare.</div><div>We consider a variety of feedback scenarios and distinguish the cases where the mechanism posts one price and when it can post different prices for buyer and seller. We show several surprising results about the separation between the different scenarios. In particular we show that (a) it is impossible to achieve sublinear <em>α</em>-regret for any <span><math><mi>α</mi><mo><</mo><mn>2</mn></math></span>, (b) but with full feedback sublinear 2-regret is achievable; (c) with a single price and partial feedback one cannot get sublinear <em>α</em> regret for any constant <em>α</em> (d) nevertheless, posting two prices even with one-bit feedback achieves sublinear 2-regret, and (e) there is a provable separation in the 2-regret bounds between full and partial feedback.</div></div>","PeriodicalId":8434,"journal":{"name":"Artificial Intelligence","volume":"337 ","pages":"Article 104231"},"PeriodicalIF":5.1,"publicationDate":"2024-10-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142421612","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-23DOI: 10.1016/j.artint.2024.104229
Angelo Gilio , David E. Over , Niki Pfeifer , Giuseppe Sanfilippo
In this paper we first recall some results for conditional events, compound conditionals, conditional random quantities, p-consistency, and p-entailment. We discuss the equivalence between conditional bets and bets on conditionals, and review de Finetti's trivalent analysis of conditionals. But we go beyond de Finetti's early trivalent logical analysis and his later ideas, aiming to take his proposals to a higher level. We examine two recent articles that explore trivalent logics for conditionals and their definitions of logical validity and compare them with the approach to compound conditionals introduced by Gilio and Sanfilippo within the framework of conditional random quantities. As we use the notion of p-entailment, the full deduction theorem does not hold. We prove a Probabilistic Weak Deduction Theorem for conditional events. After that we study some variants of it, with further results, and we present several examples. Moreover, we illustrate how to derive new inference rules related to selected Aristotelian syllogisms. We focus on iterated conditionals and the invalidity of the Import-Export principle in the light of our Probabilistic Weak Deduction Theorem. We use the inference from a disjunction, A or B, to the conditional, if not-A then B, as an example to show the invalidity of this principle. We introduce a General Import-Export principle by examining examples and counterexamples. In particular, when considering the inference rules of System P, we find that a General Import-Export principle is satisfied, even if the assumptions of the Probabilistic Weak Deduction Theorem do not hold. We also deepen further aspects related to p-entailment and p-consistency. Finally, we briefly discuss some related work relevant to AI.
在本文中，我们首先回顾条件事件、复合条件、条件随机量、p 一致性和 p 尾数的一些结果。我们讨论了条件投注和条件投注之间的等价性，并回顾了德-菲尼提对条件的三价分析。但我们超越了德菲内蒂早期的三价逻辑分析和他后来的观点，旨在将他的提议提升到更高的层次。我们研究了最近两篇探讨条件的三价逻辑及其逻辑有效性定义的文章，并将它们与吉里奥和桑菲菲利波在条件随机量框架内引入的复合条件的方法进行了比较。由于我们使用 p-entailment 概念，完全演绎定理并不成立。我们证明了条件事件的概率弱演绎定理。之后，我们研究了它的一些变体，并给出了进一步的结果和几个例子。此外，我们还说明了如何推导出与选定的亚里士多德三段论相关的新推理规则。根据我们的概率弱演绎定理，我们将重点放在迭代条件和导入导出原则的无效性上。我们以从析取条件 A 或 B 到条件 If not-A then B 的推理为例，说明该原则的无效性。我们通过研究实例和反例，介绍了一般导入导出原理。特别是，在考虑系统 P 的推理规则时，我们发现即使概率弱演绎定理的假设不成立，一般导入导出原则也是满足的。我们还进一步深化了与 P 尾数和 P 一致性相关的方面。最后，我们简要讨论了一些与人工智能相关的工作。
{"title":"On trivalent logics, probabilistic weak deduction theorems, and a general import-export principle","authors":"Angelo Gilio , David E. Over , Niki Pfeifer , Giuseppe Sanfilippo","doi":"10.1016/j.artint.2024.104229","DOIUrl":"10.1016/j.artint.2024.104229","url":null,"abstract":"<div><div>In this paper we first recall some results for conditional events, compound conditionals, conditional random quantities, p-consistency, and p-entailment. We discuss the equivalence between conditional bets and bets on conditionals, and review de Finetti's trivalent analysis of conditionals. But we go beyond de Finetti's early trivalent logical analysis and his later ideas, aiming to take his proposals to a higher level. We examine two recent articles that explore trivalent logics for conditionals and their definitions of logical validity and compare them with the approach to compound conditionals introduced by Gilio and Sanfilippo within the framework of conditional random quantities. As we use the notion of p-entailment, the full deduction theorem does not hold. We prove a Probabilistic Weak Deduction Theorem for conditional events. After that we study some variants of it, with further results, and we present several examples. Moreover, we illustrate how to derive new inference rules related to selected Aristotelian syllogisms. We focus on iterated conditionals and the invalidity of the Import-Export principle in the light of our Probabilistic Weak Deduction Theorem. We use the inference from a disjunction, <em>A or B</em>, to the conditional, <em>if not-A then B</em>, as an example to show the invalidity of this principle. We introduce a General Import-Export principle by examining examples and counterexamples. In particular, when considering the inference rules of System P, we find that a General Import-Export principle is satisfied, even if the assumptions of the Probabilistic Weak Deduction Theorem do not hold. We also deepen further aspects related to p-entailment and p-consistency. Finally, we briefly discuss some related work relevant to AI.</div></div>","PeriodicalId":8434,"journal":{"name":"Artificial Intelligence","volume":"337 ","pages":"Article 104229"},"PeriodicalIF":5.1,"publicationDate":"2024-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142328068","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-23DOI: 10.1016/j.artint.2024.104230
Thomas Eiter , Tobias Geibinger , Nelson Higuera Ruiz , Nysret Musliu , Johannes Oetsch , Dave Pfliegler , Daria Stepanova
Answer-set programming (ASP) is a prominent approach to declarative problem solving that is increasingly used to tackle challenging optimisation problems. We present an approach to leverage ASP optimisation by using large-neighbourhood search (LNS), which is a meta-heuristic where parts of a solution are iteratively destroyed and reconstructed in an attempt to improve an overall objective. In our LNS framework, neighbourhoods can be specified either declaratively as part of the ASP encoding or automatically generated by code. Furthermore, our framework is self-adaptive, i.e., it also incorporates portfolios for the LNS operators along with selection strategies to adjust search parameters on the fly. The implementation of our framework, the system ALASPO, currently supports the ASP solver clingo, as well as its extensions clingo-dl and clingcon that allow for difference and full integer constraints, respectively. It utilises multi-shot solving to efficiently realise the LNS loop and in this way avoids program regrounding. We describe our LNS framework for ASP as well as its implementation, discuss methodological aspects, and demonstrate the effectiveness of the adaptive LNS approach for ASP on different optimisation benchmarks, some of which are notoriously difficult, as well as real-world applications for shift planning, configuration of railway-safety systems, parallel machine scheduling, and test laboratory scheduling.
答案集编程（ASP）是一种著名的声明式问题求解方法，越来越多地用于解决具有挑战性的优化问题。我们提出了一种通过使用大型邻域搜索（LNS）来利用 ASP 优化的方法，LNS 是一种元启发式，在这种方法中，解决方案的某些部分会被反复破坏和重建，以试图改善总体目标。在我们的 LNS 框架中，邻域可以作为 ASP 编码的一部分以声明方式指定，也可以由代码自动生成。此外，我们的框架还具有自适应能力，也就是说，它还包含了 LNS 运算符的组合以及选择策略，以即时调整搜索参数。我们框架的实现系统 ALASPO 目前支持 ASP 求解器 clingo 及其扩展程序 clingo-dl 和 clingcon，它们分别支持差分和全整数约束。它利用多射求解来有效实现 LNS 循环，从而避免了程序的重新搁浅。我们介绍了用于 ASP 的 LNS 框架及其实现方法，讨论了方法论方面的问题，并在不同的优化基准（其中一些是众所周知的难题）以及轮班计划、铁路安全系统配置、并行机调度和测试实验室调度等实际应用中演示了用于 ASP 的自适应 LNS 方法的有效性。
{"title":"Adaptive large-neighbourhood search for optimisation in answer-set programming","authors":"Thomas Eiter , Tobias Geibinger , Nelson Higuera Ruiz , Nysret Musliu , Johannes Oetsch , Dave Pfliegler , Daria Stepanova","doi":"10.1016/j.artint.2024.104230","DOIUrl":"10.1016/j.artint.2024.104230","url":null,"abstract":"<div><div>Answer-set programming (ASP) is a prominent approach to declarative problem solving that is increasingly used to tackle challenging optimisation problems. We present an approach to leverage ASP optimisation by using large-neighbourhood search (LNS), which is a meta-heuristic where parts of a solution are iteratively destroyed and reconstructed in an attempt to improve an overall objective. In our LNS framework, neighbourhoods can be specified either declaratively as part of the ASP encoding or automatically generated by code. Furthermore, our framework is self-adaptive, i.e., it also incorporates portfolios for the LNS operators along with selection strategies to adjust search parameters on the fly. The implementation of our framework, the system ALASPO, currently supports the ASP solver clingo, as well as its extensions clingo-dl and clingcon that allow for difference and full integer constraints, respectively. It utilises multi-shot solving to efficiently realise the LNS loop and in this way avoids program regrounding. We describe our LNS framework for ASP as well as its implementation, discuss methodological aspects, and demonstrate the effectiveness of the adaptive LNS approach for ASP on different optimisation benchmarks, some of which are notoriously difficult, as well as real-world applications for shift planning, configuration of railway-safety systems, parallel machine scheduling, and test laboratory scheduling.</div></div>","PeriodicalId":8434,"journal":{"name":"Artificial Intelligence","volume":"337 ","pages":"Article 104230"},"PeriodicalIF":5.1,"publicationDate":"2024-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S0004370224001668/pdfft?md5=ab34b67efbd10758275677814fac17d7&pid=1-s2.0-S0004370224001668-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142316277","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-12DOI: 10.1016/j.artint.2024.104228
François Suro, Fabien Michel, Tiberiu Stratulat
Compared to autonomous agent learning, lifelong agent learning tackles the additional challenge of accumulating skills in a way favourable to long term development. What an agent learns at a given moment can be an element for the future creation of behaviours of greater complexity, whose purpose cannot be anticipated.
Beyond its initial low-level sensorimotor development phase, the agent is expected to acquire, in the same manner as skills, values and goals which support the development of complex behaviours beyond the reactive level. To do so, it must have a way to represent and memorize such information.
In this article, we identify the properties suitable for a representation system supporting the lifelong development of agents through a review of a wide range of memory systems and related literature. Following this analysis, our second contribution is the proposition and implementation of such a representation system in MIND, a modular architecture for lifelong development. The new variable module acts as a simple memory system which is strongly integrated to the hierarchies of skill modules of MIND, and allows for the progressive structuration of behaviour around persistent non-symbolic representations. Variable modules have many applications for the development and structuration of complex behaviours, but also offer designers and operators explicit models of values and goals facilitating human interaction, control and explainability.
We show through experiments two possible uses of variable modules. In the first experiment, skills exchange information by using a variable representing the concept of “target”, which allows the generalization of navigation behaviours. In the second experiment, we show how a non-symbolic representation can be learned and memorized to develop beyond simple reactive behaviour, and keep track of the steps of a process whose state cannot be inferred by observing the environment.
与自主代理学习相比，终身代理学习面临着以有利于长期发展的方式积累技能的额外挑战。除了最初的低级感知运动发展阶段，我们还期望代理能以与技能相同的方式获得价值观和目标，从而支持其发展出超越反应水平的复杂行为。在本文中，我们通过对各种记忆系统和相关文献的回顾，确定了适合支持代理终身发展的表征系统的属性。根据这一分析，我们的第二个贡献是在 MIND（一种用于终身发展的模块化架构）中提出并实现了这样一种表征系统。新的可变模块作为一个简单的记忆系统，与 MIND 的技能模块层次结构紧密结合，并允许围绕持久的非符号表征逐步构建行为结构。可变模块在复杂行为的开发和结构化方面有很多应用，同时也为设计者和操作者提供了明确的价值和目标模型，促进了人际互动、控制和可解释性。在第一个实验中，通过使用代表 "目标 "概念的变量来交换技能信息，从而实现导航行为的通用化。在第二个实验中，我们展示了如何通过学习和记忆非符号表征来超越简单的反应行为，并跟踪无法通过观察环境来推断状态的过程步骤。
{"title":"Integration of memory systems supporting non-symbolic representations in an architecture for lifelong development of artificial agents","authors":"François Suro, Fabien Michel, Tiberiu Stratulat","doi":"10.1016/j.artint.2024.104228","DOIUrl":"10.1016/j.artint.2024.104228","url":null,"abstract":"<div><p>Compared to autonomous agent learning, lifelong agent learning tackles the additional challenge of accumulating skills in a way favourable to long term development. What an agent learns at a given moment can be an element for the future creation of behaviours of greater complexity, whose purpose cannot be anticipated.</p><p>Beyond its initial low-level sensorimotor development phase, the agent is expected to acquire, in the same manner as skills, values and goals which support the development of complex behaviours beyond the reactive level. To do so, it must have a way to represent and memorize such information.</p><p>In this article, we identify the properties suitable for a representation system supporting the lifelong development of agents through a review of a wide range of memory systems and related literature. Following this analysis, our second contribution is the proposition and implementation of such a representation system in MIND, a modular architecture for lifelong development. The new <em>variable module</em> acts as a simple memory system which is strongly integrated to the hierarchies of skill modules of MIND, and allows for the progressive structuration of behaviour around persistent non-symbolic representations. <em>Variable modules</em> have many applications for the development and structuration of complex behaviours, but also offer designers and operators explicit models of values and goals facilitating human interaction, control and explainability.</p><p>We show through experiments two possible uses of <em>variable modules</em>. In the first experiment, skills exchange information by using a variable representing the concept of “target”, which allows the generalization of navigation behaviours. In the second experiment, we show how a non-symbolic representation can be learned and memorized to develop beyond simple reactive behaviour, and keep track of the steps of a process whose state cannot be inferred by observing the environment.</p></div>","PeriodicalId":8434,"journal":{"name":"Artificial Intelligence","volume":"337 ","pages":"Article 104228"},"PeriodicalIF":5.1,"publicationDate":"2024-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S0004370224001644/pdfft?md5=6e7585d9f3b58a33fd4b372451648a7b&pid=1-s2.0-S0004370224001644-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142241363","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号