Next generation machine learning explainability

In a world more and more influenced by Artificial Intelligence (AI), the ability to comprehend and trust the decisions made by AI is of utter importance. Enter the fascinating world of Explainable AI (XAI), where groundbreaking research is shedding light on the inner workings of these black-box models. This PhD project, investigates the possibilities of XAI with a particular focus on visualizations of latent space, aiming to demystify the decision-making processes of AI systems.

The latent space can be compared to an abstract, high-dimensional realm where AI models encode and process information. Imagine it as a hidden universe where data points are trans-formed into representations that guide the AI's decision-making. The challenge lies in making this latent space understandable to humans. Through innovative visualization techniques, this projects aims to create intuitive maps that render the latent space comprehensible. These visualizations serve as a bridge between the intricate computations occurring within the AI model and the human mind, providing a clear lens through which we can interpret, validate, and improve the decisions made by these systems – and maybe even learn from them.

Why does this matter? Well, AI is everywhere these days (and will be even more in the future) – from diagnosing diseases to self-driving cars and predicting stock market trends. As we journey into an era increasingly shaped by AI, this project strives to make these powerful technologies more transparent, accessible, and accountable. By unraveling the complexities of latent space through visualizations, this PhD project hopefully contributes to a future where AI decisions are not just powerful but also understandable and trustworthy.