United States

Autonomous systems are often deployed in the open world with incomplete specification of objectives or constraints, due to practical limitations in precise model specification. Operating based on such models may produce negative side effects (NSEs), which affect the safety and reliability of the system. We focus on mitigating NSEs assuming an underlying Markov decision process (MDP) based planning model. First, we learn a model of NSEs using observed data which contains state-action trajectories and the severity of associated .NSEs. Unlike previous work that associated NSEs with state-action pairs, our framework associates NSEs with entire trajectories, which is highly general and models non-Markovian dependence on states and actions. Second, we learn finite state controllers (FSCs) that predict the NSE severity for a given trajectory, and also generalize well to unseen data. Finally, we develop a constrained MDP model that uses information from both the underlying MDP and the learned FSC for planning while avoiding NSEs. Empirical evaluations show the effectiveness of our approach in learning and mitigating non-Markovian NSEs from a limited amount of historical data, and outperforming the previous best method.

AAAI 2023

Planning and Learning for Non-Markovian Negative Side Effects Using Finite State Controllers

safe ai systems

safe learning

planning with markov models

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. AAAI-23 is the Thirty-Seventh AAAI Conference on Artificial Intelligence. The theme of this conference is to create collaborative bridges within and beyond AI. Like previous AAAI conferences, AAAI-23 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and two new activities: a Bridge Program and a Lab Program. Many of these activities are tailored to the theme of bridges and all are selected according to the highest standards, with additional programs for students and young researchers. 
AAAI is providing you with a conference planner, which you can use to help organize your itinerary of activities. This includes talks to attend in person, talks to attend remotely, breaks with colleagues and your site seeing activities. To access this conference planner, please go to [https://aaai-2023.takemobi.io](https://aaai-2023.takemobi.io).

In order to access this site, you need to register. If you haven't already, please register [here](https://aaai.org/Conferences/AAAI-23/registration/).


The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines.

poster


We study the synthesis under environment specifications problem for LTL/LTLf which, in particular, generalizes FOND (strong) planning with these temporal goals. We consider the case where the agent cannot enforce its goal --- for which the argument for using best-effort strategies has been made  --- and study the intermediate ground, between enforcing and best-effort strategies, of dominant strategies. Intuitively, such strategies achieve the goal against any environment for which it is achievable.  

We show that dominant strategies may exist when enforcing ones do not, while still sharing with the latter many desirable properties such as being interchangeable with each other, and being monotone with respect to tightening of environment specifications. We give necessary and sufficient conditions for the existence of dominant strategies, and show that deciding if they exist is 2EXPTIME-complete --- the same as for enforcing strategies. Finally, we give a uniform, optimal, game-theoretic algorithm for simultaneously solving the three synthesis problems of enforcing, dominant, and best-effort strategies. 

Reactive Synthesis of Dominant Strategies

We introduce a class of strategic games in which agents are assigned to nodes of a topology graph and the utility of an agent depends on both the agent's inherent utilities for other agents as well as her distance from these agents on the topology graph. This model of topological distance games (TDGs) offers an appealing combination of important aspects of several prominent settings in coalition formation, including (additively separable) hedonic games, social distance games, and Schelling games. We study the existence and complexity of stable outcomes in TDGs—for instance, while a jump stable assignment may not exist in general, we show that the existence is guaranteed in several special cases. We also investigate the dynamics induced by performing beneficial jumps.

Topological Distance Games

Estimating direct and indirect causal effects from observational data is crucial to understanding the causal mechanisms and predicting the behaviour under different interventions. Causal mediation analysis is a method that is often used to reveal direct and indirect effects. Deep learning shows promise in mediation analysis, but the current methods only assume latent confounders that affect treatment, mediator and outcome simultaneously, and fail to identify different types of latent confounders (e.g., confounders that only affect the mediator or outcome). Furthermore, current methods are based on the sequential ignorability assumption, which is not feasible for dealing with multiple types of latent confounders. This work aims to circumvent the sequential ignorability assumption and applies the piecemeal deconfounding assumption as an alternative. We propose the Disentangled Mediation Analysis Variational AutoEncoder (DMAVAE), which disentangles the representations of latent confounders into three types to accurately estimate the natural direct effect, natural indirect effect and total effect. Experimental results show that the proposed method outperforms existing methods and has strong generalisation ability. We further apply the method to a real-world dataset to show its potential application.

Disentangled Representation for Causal Mediation Analysis

Ensuring the overall end-user experience in arbitrary style transfer (AST) is challenging as people often rate the stylization quality of the same stylized image very differently. Producing a satisfactory image requires iterating through several AST models or repeatedly adding randomness with diversified AST (DAST). This approach is exhaustive and computationally inefficient. We propose a novel solution that allows users to modulate AST behavior at run-time systematically. We begin with reformulating three prominent AST methods into a unified assign-and-mix problem and discover that the entropies of their assignment matrices exhibit a large variance. We then solve the unified problem in an optimal transport framework using the Sinkhorn-Knopp algorithm with a user input $\eps$ to control the said entropy and thus modulate stylization. Empirical results demonstrate the superiority of the proposed solution, with speed and stylization quality comparable to or better than existing AST and significantly more diverse than previous DAST works.

User-controllable Arbitrary Style Transfer via Entropy Regularization

Beyond achieving higher compression efficiency over classical image compression codecs, deep image compression is expected to be improved with additional side information, e.g., another image from a different perspective of the same scene.
To better utilize the side information under the distributed compression scenario, the existing method (Ayzik and Avidan 2020) only implements patch matching at the image domain to solve the parallax problem caused by the difference in viewing points. However, the patch matching at the image domain is not robust to the variance of scale, shape, and illumination caused by the different viewing angles, and can not make full use of the rich texture information of the side information image. To resolve this issue, we propose Multi-Scale Feature Domain Patch Matching (MSFDPM) to fully utilizes side information at the decoder of the distributed image compression model. Specifically, MSFDPM consists of a side information feature extractor, a multi-scale feature domain patch matching module, and a multi-scale feature fusion network. Furthermore, we reuse inter-patch correlation from the shallow layer to accelerate the patch matching of the deep layer. Finally, we find that our patch matching in a multi-scale feature domain further improves compression rate by about 20% compared with the patch matching method at image domain (Ayzik and Avidan 2020).

Learned Distributed Image Compression with Multi-Scale Patch Matching in Feature Domain

Unsupervised graph representation learning (UGRL) has drawn increasing research attention and achieved promising results in several graph analytic tasks. Relying on the homophily assumption, existing UGRL methods tend to smooth the learned node representations along all edges, ignoring the existence of heterophilic edges that connect nodes with distinct attributes. As a result, current methods are hard to generalize to heterophilic graphs where dissimilar nodes are widely connected, and also vulnerable to adversarial attacks. To address this issue, we propose a novel unsupervised Graph Representation learning method with Edge hEterophily discriminaTing (GREET) which learns representations by discriminating and leveraging homophilic edges and heterophilic edges. To distinguish two types of edges, we build an edge discriminator that infers edge homophily/heterophily from feature and structure information. We train the edge discriminator in an unsupervised way through minimizing the crafted pivot-anchored ranking loss, with randomly sampled node pairs acting as pivots. Node representations are learned through contrasting the dual-channel encodings obtained from the discriminated homophilic and heterophilic edges. With an effective interplaying scheme, edge discriminating and representation learning can mutually boost each other during the training phase. We conducted extensive experiments on 14 benchmark datasets and multiple learning scenarios to demonstrate the superiority of GREET.

Beyond Smoothing: Unsupervised Graph Representation Learning with Edge Heterophily Discriminating

Numeric planning is known to be undecidable even under severe restrictions. Prior work has investigated the decidability boundaries by restricting the expressiveness of the planning formalism in terms of the numeric functions allowed in conditions and effects. We study a well-known restricted form of Hoffmann's simple numeric planning, which is undecidable. We analyze the complexity by imposing restrictions on the causal structure, exploiting a novel method for bounding variable domain sizes. First, we show that plan existence for tasks where all numeric variables are root nodes in the causal graph is in PSPACE.
Second, we show that for tasks with only numeric leaf variables the problem is decidable, and that it is in PSPACE if the propositional state space has a fixed size. Our work lays a strong foundation for future investigations of structurally more complex tasks. From a practical perspective, our method allows to employ heuristics and methods that are geared towards finite variable domains (such as pattern database heuristics or decoupled search) to solve non-trivial families of numeric planning problems.

Structurally Restricted Fragments of Numeric Planning -- A Complexity Analysis

Despite improvements in safe water and sanitation services in low-income countries, a substantial proportion of the population in Africa still does not have access to these essential services. Up-to-date fine-scale maps of low-income settlements are urgently needed by authorities  to improve service provision. We aim to develop a cost-effective solution to generate fine-scale maps of these vulnerable populations using multi-source public information. The problem is challenging as ground-truth maps are available at only a limited number of cities, and the patterns are heterogeneous across cities. Recent attempts tackling the spatial heterogeneity issue focus on scenarios where true labels partially exist for each input region, which are unavailable for the present problem. We propose a dynamic point-to-region co-learning framework to learn heterogeneity patterns that cannot be reflected by point-level information and generalize deep learners to new areas with no labels. We also propose an attention-based correction layer to remove spurious signatures, and a region-gate to capture both region-invariant and variant patterns. Experiment results on real-world fine-scale data in three cities of Kenya show that the proposed approach can largely improve model performance on various base network architectures.

Planning and Learning for Non-Markovian Negative Side Effects Using Finite State Controllers

Downloads

Next from AAAI 2023

Reactive Synthesis of Dominant Strategies

Similar lecture

PLMmark: A Secure and Robust Black-box Watermarking Framework for Pre-trained Language Models

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES