AIhub monthly digest: November 2024 – dynamic faceted search, the kidney exchange problem, and AfriClimate AI

by Lucy Smith

29 November 2024

Harnessing AI for a climate-resilient Africa: An interview with Amal Nammouchi

This month, we had the pleasure of chatting to Amal Nammouchi, co-founder of AfriClimate AI, a grassroots community focused on using artificial intelligence to tackle climate challenges in Africa. Amal told us about the inspiration behind the initiative, some of their activities and projects, and plans for the future. Read the interview here.

Building trust in AI: Transparent models for better decisions

In this blog post, Danial Dervovic writes about work presented at IJCAI 2024 on improving the interpretability of logistic regression models. Danial and colleagues have proposed an augmentation to such logistic models, which makes decisions made by them more understandable.

The kidney exchange problem

In their work Parameterized Complexity of Kidney Exchange Revisited, presented at IJCAI 2024, Úrsula Hébert-Johnson, Daniel Lokshtanov, Chinmay Sonar and Vaishali Surianarayanan consider the issue of kidney donation and how to maximum the number of patients who receive a transplant. We hear from the authors about the kidney exchange problem and how they went about solving two of the open problems in this field.

Enhancing controlled query evaluation through epistemic policies

A significant data challenge concerns the sharing of information without compromising sensitive details. Gianluca Cima, Domenico Lembo, Lorenzo Marconi, Riccardo Rosati and Domenico Fabio Savo won a distinguished paper award at IJCAI 2024 for their work Enhancing Controlled Query Evaluation through Epistemic Policies, which considers the controlled query evaluation framework – an approach that safeguards confidentiality whilst still providing answers to queries. The team wrote about their research in this blog post.

Dynamic faceted search: from haystack to highlight

The number of scholarly articles is growing rapidly, and finding the most relevant information from this vast collection of data can be daunting. In their work A Neuro-symbolic Approach for Faceted Search in Digital Libraries, presented at ECAI-2024, Mutahira Khalid, Sören Auer and Markus Stocker utilise facet generation, an advanced search method that allows users to filter and refine search results. They outline three different dynamic methods that adapt and adjust facets in real-time, based on user input and the evolving nature of the dataset. You can read their blog post about this work here.

How to choose your loss function

In machine learning classification tasks, achieving high accuracy is only part of the goal. It is equally important to know how confident the models are in their predictions, a concept known as model calibration. A key factor influencing both the accuracy and calibration of a model is the choice of the loss function during training. In this blog post, Viacheslav Komisarenko explores how to choose a loss function to achieve good calibration. This is work that won him, and co-author Meelis Kull, an outstanding paper award at ECAI-2024.

The Turing Lectures: Can we trust AI? – with Abeba Birhane

The Turing Lectures series features influential figures from the world of data science and artificial intelligence. The latest lecture was given by Dr Abeba Birhane, in which she tackled the topic of biases in data and the downstream impact on AI systems, and showed how this can lead to unfair outcomes in our daily lives. You can watch the lecture here.

Everyday AI podcast back for season two

CSIRO’s Everyday AI podcast is back for a second season. Host Jon Whittle, and expert guests, delve into how AI is impacting various aspects of our lives, from healthcare and education to workplaces and beyond. You can listen to the series here.