In the hunt for new physics phenomena, such as Axion-like particles (ALPs), it is crucial to compare experimental data to theoretical models. This involves inferring the most likely values of a model’s
parameters — such as particle masses and cross sections. However, traditional likelihood-based inference techniques are oftentimes not practically feasible without making significant simplifying assumptions, which decrease the reliability of the inference. This is especially the case for ALP-searches with gamma-ray telescopes such as the upcoming Cherenkov Telescope Array. Recently however, new likelihood-free inference (LFI) techniques based on machine learning have emerged to help overcome these limitations. In particular, “Neural Ratio Estimation” (NRE) stands out with its reported accuracy and efficiency. In this contribution, we have applied NRE to simulated CTA-data of the active galactic nucleus NGC1275 in the Perseus Cluster, in order to probe the viability of this technique for ALP-searches with cosmic gamma-rays. Our example-inferences provide encouraging evidence that NRE will be applicable to deriving sensitive and accurate limits. We also identify some challenges in the practical execution of such an analysis, as well as concrete next steps towards deriving formal and reliable limits on the ALP mass and coupling to photons.