Deep neural networks are often considered opaque systems, prompting the need for explainability methods to improve trust and accountability. Existing approaches typically attribute test-time predictions either to input features (e.g., pixels in an image) or to influential training examples. We argue that both perspectives should be studied jointly. This work explores training feature attribution, which links test predictions to specific regions of specific training images and thereby provides new insights into the inner workings of deep models. Our experiments on vision datasets show that training feature attribution yields fine-grained, test-specific explanations: it identifies harmful examples that drive misclassifica-tions and reveals spurious correlations, such as patch-based shortcuts, that conventional attribution methods fail to expose. Deep neural networks have achieved state-of-the-art performance across a wide range of domains, including image recognition, natural language processing, and multimodal reasoning (He et al., 2016; Devlin et al., 2019; Radford et al., 2021). However, this impressive performance comes at the cost of transparency: modern deep models operate as complex, highly-parameterized black boxes, where the reasoning behind individual predictions is often opaque (Lipton, 2018). This opacity can undermine user trust, hinder debugging, and conceal harmful biases or spurious correlations (Ar-jovsky et al., 2019; DeGrave et al., 2021).

We propose a fast and simple explainable AI (XAI) method for point cloud data. It computes pointwise importance with respect to a trained network downstream task. This allows better understanding of the network properties, which is imperative for safety-critical applications. In addition to debugging and visualization, our low computational complexity facilitates online feedback to the network at inference. This can be used to reduce uncertainty and to increase robustness. In this work, we introduce \emph{Feature Based Interpretability} (FBI), where we compute the features' norm, per point, before the bottleneck. We analyze the use of gradients and post- and pre-bottleneck strategies, showing pre-bottleneck is preferred, in terms of smoothness and ranking. We obtain at least three orders of magnitude speedup, compared to current XAI methods, thus, scalable for big point clouds or large-scale architectures. Our approach achieves SOTA results, in terms of classification explainability. We demonstrate how the proposed measure is helpful in analyzing and characterizing various aspects of 3D learning, such as rotation invariance, robustness to out-of-distribution (OOD) outliers or domain shift and dataset bias.

Distributed decision-making in multi-agent systems presents difficult challenges for interactive behavior learning in both cooperative and competitive systems. To mitigate this complexity, MAIDRL presents a semi-centralized Dense Reinforcement Learning algorithm enhanced by agent influence maps (AIMs), for learning effective multi-agent control on StarCraft Multi-Agent Challenge (SMAC) scenarios. In this paper, we extend the DenseNet in MAIDRL and introduce semi-centralized Multi-Agent Dense-CNN Reinforcement Learning, MAIDCRL, by incorporating convolutional layers into the deep model architecture, and evaluate the performance on both homogeneous and heterogeneous scenarios. The results show that the CNN-enabled MAIDCRL significantly improved the learning performance and achieved a faster learning rate compared to the existing MAIDRL, especially on more complicated heterogeneous SMAC scenarios. We further investigate the stability and robustness of our model. The statistics reflect that our model not only achieves higher winning rate in all the given scenarios but also boosts the agent's learning process in fine-grained decision-making.

Modern radio telescopes produce unprecedented amounts of data, which are passed through many processing pipelines before the delivery of scientific results. Hyperparameters of these pipelines need to be tuned by hand to produce optimal results. Because many thousands of observations are taken during a lifetime of a telescope and because each observation will have its unique settings, the fine tuning of pipelines is a tedious task. In order to automate this process of hyperparameter selection in data calibration pipelines, we introduce the use of reinforcement learning. We use a reinforcement learning technique called twin delayed deep deterministic policy gradient (TD3) to train an autonomous agent to perform this fine tuning. For the sake of generalization, we consider the pipeline to be a black-box system where only an interpreted state of the pipeline is used by the agent. The autonomous agent trained in this manner is able to determine optimal settings for diverse observations and is therefore able to perform 'smart' calibration, minimizing the need for human intervention.

Deep generative models such as Generative Adversarial Networks (GANs) and Variational Auto-Encoders (VAEs) are important tools to capture and investigate the properties of complex empirical data. However, the complexity of their inner elements makes their functioning challenging to assess and modify. In this respect, these architectures behave as black box models. In order to better understand the function of such networks, we analyze their modularity based on the counterfactual manipulation of their internal variables. Experiments with face images support that modularity between groups of channels is achieved to some degree within convolutional layers of vanilla VAE and GAN generators. This helps understand the functional organization of these systems and allows designing meaningful transformations of the generated images without further training.

Not only is this blunt-force approach extremely inefficient, it doesn't guarantee that none of the IT equipment will overheat. "You encounter hot spots even in an over-cooled data center," Rajat Gosh, CEO of AdeptDC, a startup whose software use machine learning to manage data center infrastructure, told Data Center Knowledge in an interview. One of the hardest problems to solve in data center cooling is pressure distribution, he said, and machine learning can be especially effective at solving it. Earlier this year, Joe Kava, the man in charge of data centers for Alphabet's Google, revealed to us that the company had been using machine learning algorithms to automatically tune its data center cooling systems, which enabled cooling energy savings of up to 30 percent. Google has considered turning the technology into a solution it can offer to other companies managing industrial facilities, and it may that sometime in the future, but you don't need to wait.

Influence Maps have been successfully used in controlling the navigation of multiple units. In this paper, we apply the idea to the problem of simulating a kiting behavior (also known as ¨attack and flee'¨) in the context of real-time strategy (RTS) games. We present our approach and evaluate it in the popular RTS game StarCraft, where we analyze the benefits that our approach brings to a StarCraft playing bot.

Shepherding with a dog presents an interesting challenge for artificial intelligence, with multiple intelligent systems assessing and interacting with each other in order to achieve a variety of goals. We present a solution to this problem, which consists of a dog AI making use of influence mapping, state machines and A* pathfinding to respond intelligently to real-life shepherding commands issued by a high-level shepherd AI steering the flock of sheep through waypoints on a variety of maps by using pathfinding and influence maps. The role of the AI shepherd can also be taken by a human player (using either a point and click or voice recognition interface) for matches against the artificial shepherd which proved to be a worthy opponent for human testers. The system was evaluated through user testing and provided a high degree of realism and engaging gameplay relying heavily on the workings of the presented AI components.

In Real-Time Strategy (RTS) video games, players (controlled by humans or computers) build structures and recruit armies, fight for space and resources in order to control strategic points, destroy the opposing force and ultimately win the game. Players need to predict where and how the opponents will strike in order to best defend themselves. Conversely, assessing how the opponents will defend themselves is crucial to mounting a successful attack while exploiting the vulnerabilities in the opponent's defence strategy. In this context, to be truly adaptable, computer-controlled players need to recognize their opponents' behaviour, their goals, and their plans to achieve those goals. In this paper we analyze the algorithmic challenges behind behaviour recognition in RTS games and discuss a generic RTS behaviour recognition system that we are developing to address those challenges. The application domain is that of RTS games, but many of the key points we discuss also apply to other video game genres such as multiplayer first person shooter (FPS) games.