Facial image anonymization is essential to enable privacy-preserving image data sharing. The core challenge lies in removing identity-revealing information without degrading the utility of the images, which is essential for, e.g., demographic analysis. However, existing techniques apply uniform pixel-level distortions or synthesize replacements using Generative Adversarial Networks (GANs), which do not retain the meaningful features necessary for downstream tasks. To address this issue, we introduce EDIALS (Explainability-Driven Image Anonymization in Latent Space), which selectively modifies identity-specific latent features identified via explainability techniques. By applying targeted, incremental distortions in the latent space of an adversarial autoencoder, EDIALS effectively anonymizes images while preserving their analytical utility much better than existing techniques. Empirical evaluations on a common dataset show that EDIALS achieves 0.42% re-identification risk (equivalent to random guessing) while maintaining high utility: 84.66% F1 for age, 97.91% for gender, and 82.61% for race classification. In contrast, DeepPrivacy2 —a state-of-the-art GAN-based approach— results in a re-identification risk as large as 16.27% and lower utility: 78.32% F1, 82.58%, and 73.32% for age, gender, and race classification, respectively.

Abstract Artificial Intelligence systems could find many important applications in the medical field, holding excellent potential for improving disease diagnosis, treatment identification and selection. These opportunities are often jeopardized by the lack of interpretability of such systems, slowing down AI adoption. To overcome the issue, we first introduce an analytical framework exploiting multimodal deep learning for the classification of prostate lesions using Magnetic Resonance Imaging (MRI) data and clinical information on the patients. Then, we propose a multimodal explainability approach based on visual explanations to interpret the proposed model decision-making process and identify how the different modalities contribute to each specific prediction. Our findings, based on the PI-CAI Grand Challenge dataset, demonstrate the potential of combining multimodal data with eXplainable AI (XAI) to enhance prostate cancer diagnosis, improving model predictive performance, interpretability and understanding in treatment decision-making.

We introduce a novel computational unit for neural networks that features multiple biases, challenging the traditional perceptron structure. This unit emphasizes the importance of preserving uncorrupted information as it is passed from one unit to the next, applying activation functions later in the process with specialized biases for each unit. Through both empirical and theoretical analyses, we show that by focusing on increasing biases rather than weights, there is potential for significant enhancement in a neural network model’s performance. This approach offers an alternative perspective on optimizing information flow within neural networks. See source code [5].

In the last years, DeepMind algorithm AlphaZero has become the state of the art to efficiently tackle perfect information two-player zero-sum games with a win/lose outcome. However, when the win/lose outcome is decided by a final score difference, AlphaZero may play score-suboptimal moves, because all winning final positions are equivalent from the win/lose outcome perspective. This can be an issue, for instance when used for teaching, or when trying to understand whether there is a better move. Moreover, there is the theoretical quest of the perfect game. A naive approach would be training a AlphaZero-like agent to predict score differences instead of win/lose outcomes. Since the game of Go is deterministic, this should as well produce outcome-optimal play. However, it is a folklore belief that "this does not work".In this paper we first provide empirical evidence to this belief. We then give a theoretical interpretation of this suboptimality in a general perfect information two-player zero-sum game where the complexity of a game like Go is replaced by randomness of the environment. We show that an outcome-optimal policy has a different preference for uncertainty when it is winning or losing. In particular, when in a losing state, an outcome-optimal agent chooses actions leading to a higher variance of the score. We then posit that when approximation is involved, a deterministic game behaves like a nondeterministic game, where the score variance is modeled by how uncertain the position is. We validate this hypothesis in a AlphaZero-like software with a human expert.

Explainable AI consists in developing mechanisms allowing for an interaction between decision systems and humans by making the decisions of the formers understandable. This is particularly important in sensitive contexts like in the medical domain. We propose a use case study, for skin lesion diagnosis, illustrating how it is possible to provide the practitioner with explanations on the decisions of a state of the art deep neural network classifier trained to characterize skin lesions from examples. Our framework consists of a trained classifier onto which an explanation module operates. The latter is able to offer the practitioner exemplars and counterexemplars for the classification diagnosis thus allowing the physician to interact with the automatic diagnosis system. The exemplars are generated via an adversarial autoencoder. We illustrate the behavior of the system on representative examples.

We develop a new framework for the game of Go to target a high score, and thus a perfect play. We integrate this framework into the Monte Carlo tree search - policy iteration learning pipeline introduced by Google DeepMind with AlphaGo. Training on 9×9 Go produces a superhuman Go player, thus proving that this framework is stable and robust. We show that this player can be used to effectively play with both positional and score handicap. We develop a family of agents that can target high scores against any opponent, recover from very severe disadvantage against weak opponents, and avoid suboptimal moves.

We propose a multiple-komi modification of the AlphaGo Zero/Leela Zero paradigm. The winrate as a function of the komi is modeled with a two-parameters sigmoid function, hence the winrate for all komi values is obtained, at the price of predicting just one more variable. A second novel feature is that training is based on self-play games that occasionaly branch -with changed komi- when the position is uneven. With this setting, reinforcement learning is shown to work on 7×7 Go, obtaining very strong playing agents. As a useful byproduct, the sigmoid parameters given by the network allow to estimate the score difference on the board, and to evaluate how much the game is decided. Finally, we introduce a family of agents which target winning moves with a higher score difference.