Abstract The rapid development and widespread adoption of deep neural networks (DNNs) are attributed to their remarkable performance and practical efficacy across numerous applications. Nevertheless, the sophistication of the inference process often comes at the expense of explainability, becoming an increasingly critical concern. Explainable AI endeavors to shed light on the inner workings of neural networks, thereby enhancing trust in their predictive capabilities and facilitating more reliable and informed decision-making. Visual explanations for the behavior of DNNs contribute to understanding their functioning. However, selecting the right visualization method for a particular task and evaluating its performance remains a challenge. This paper offers a comprehensive analysis of post hoc methods for generating visual explanations, concentrating on class activation mapping (CAM) techniques. It introduces two novel approaches: high-resolution relevance propagation CAM (HiResRP-CAM) and EigenLayer-CAM. These methods are evaluated both qualitatively and quantitatively across various experimental setups within the context of convolutional neural networks (CNNs) for image classification tasks. The evaluation encompasses diverse network architectures, datasets, and metrics, including certain proposed enhancements. The paper discusses experimental results, including instances of model failure due to adversarial noise, and explores factors to deepen understanding of model behavior, characteristics of explainability methods and current strategies for assessing explanation quality. In this way, we provide a perspective on the challenges and limitations inherent in interpreting CNNs and on the relevant aspects for choosing a specific method.

Abstract The task of human action recognition (HAR) can be found in many computer vision practical applications. Various data modalities have been considered for solving this task, including joint-based skeletal representations which are suitable for real-time applications on platforms with limited computational resources. We propose a spatio-temporal neural network that uses handcrafted geometric features to classify human actions from video data. The proposed deep neural network architecture combines graph convolutional and temporal convolutional layers. The experiments performed on public HAR datasets show that our model obtains results similar to other state-of-the-art methods but has a lower inference time while offering the possibility to obtain an explanation for the classified action.

We propose a dual system for unsupervised object segmentation in video, which brings together two modules with complementary properties: a space-time graph that discovers objects in videos and a deep network that learns powerful object features. The system uses an iterative knowledge exchange policy. A novel spectral space-time clustering process on the graph produces unsupervised segmentation masks passed to the network as pseudo-labels. The net learns to segment in single frames what the graph discovers in video and passes back to the graph strong image-level features that improve its node-level features in the next iteration. Knowledge is exchanged for several cycles until convergence. The graph has one node per each video pixel, but the object discovery is fast. It uses a novel power iteration algorithm computing the main space-time cluster as the principal eigenvector of a special Feature-Motion matrix without actually computing the matrix. The thorough experimental analysis validates our theoretical claims and proves the effectiveness of the cyclical knowledge exchange. We also perform experiments on the supervised scenario, incorporating features pretrained with human supervision. We achieve state-of-the-art level on unsupervised and supervised scenarios on four challenging datasets: DAVIS, SegTrack, YouTube-Objects, and DAVSOD. We will make our code publicly available.

One of the most prominent approaches in the field of planning and scheduling is Constraint Satisfaction Problem (CSP), which is the process of finding a solution according to a set of constraints, the paper aims to discuss the possibility of using constraint satisfaction in scheduling people's daily activities and proposes an automatic scheduling algorithm that aids the users to plan their everyday activities, the scheduling process released using the concepts of Temporal Constraint Satisfaction Problem (TCSP). The paper suggests a scenario and explains the properties of the human being's daily activity to be absorbed in a TCSP algorithm, and proposes a set of special constraints the algorithm will tie itself to them to produce a final schedule that may satisfy the user. Searching for a solution will be done using backtracking and back-jumping techniques.

The increase in life expectancy and the slumping birth rates across the world result in lengthening the average age of the society. Therefore, we are in need of techniques that will assist the elderly in their daily life, while preventing their social isolation. The recent developments in Ambient Intelligence and Information and Communication Technologies have facilitated a technological revolution in the field of Ambient Assisted Living. At present, there are many technologies on the market that support the independent life of older adults, requiring less assistance from family and caregivers, yet most of them offer isolated services, such as health monitoring, reminders etc, moreover none of current solutions incorporates the integration of various functionalities and user preferences or are formally analyzed for their functionality and quality-of-service attributes, a much needed endeavor in order to ensure safe mitigations of potential critical scenarios. In this paper, we propose a novel architectural solution that integrates necessary functions of an AAL system seamlessly, based on user preferences. To enable the first level of the architecture's analysis, we model our system in Architecture Analysis and Design Language, and carry out its simulation for analyzing the end-to-end data-flow latency, resource budgets and system safety.

The paper presents a multi-agent system that tries to solve the problem of rational exploitation of natural renewable resources by self-interested agents. The particular problem instance is that of several fishing companies trying to catch fish in several fishing banks. The system comprises a set of cognitive agents that represent the companies, are able to build plans and will interact with other companies to ensure the ecological balance of the resources and the achievement of their own plans. The environment is represented by a particular agent with which the companies interact. The agents are using both a symbolic representation and a genetic representation to build plans, to model the evolution of the unpredictable world in which they live and to conduct negotiation. The genetic approach is based on cooperative coadapted species and is used to model the multi-agent world from the point of view of a particular agent.