Mariano Beguerisse-Diaz

Mariano Beguerisse-Diaz, Baojun Wang, Radhika Desikan et al.

Motivation: Estimating parameters from data is a key stage of the modelling process, particularly in biological systems where many parameters need to be estimated from sparse and noisy data sets. Over the years, a variety of heuristics have been proposed to solve this complex optimisation problem, with good results in some cases yet with limitations in the biological setting. Results: In this work, we develop an algorithm for model parameter fitting that combines ideas from evolutionary algorithms, sequential Monte Carlo and direct search optimisation. Our method performs well even when the order of magnitude and/or the range of the parameters is unknown. The method refines iteratively a sequence of parameter distributions through local optimisation combined with partial resampling from a historical prior defined over the support of all previous iterations. We exemplify our method with biological models using both simulated and real experimental data and estimate the parameters efficiently even in the absence of a priori knowledge about the parameters.

Jian Hu, Dimitrios Korkinof, Shaogang Gong et al.

We introduce ViSMap: Unsupervised Video Summarisation by Meta Prompting, a system to summarise hour long videos with no-supervision. Most existing video understanding models work well on short videos of pre-segmented events, yet they struggle to summarise longer videos where relevant events are sparsely distributed and not pre-segmented. Moreover, long-form video understanding often relies on supervised hierarchical training that needs extensive annotations which are costly, slow and prone to inconsistency. With ViSMaP we bridge the gap between short videos (where annotated data is plentiful) and long ones (where it's not). We rely on LLMs to create optimised pseudo-summaries of long videos using segment descriptions from short ones. These pseudo-summaries are used as training data for a model that generates long-form video summaries, bypassing the need for expensive annotations of long videos. Specifically, we adopt a meta-prompting strategy to iteratively generate and refine creating pseudo-summaries of long videos. The strategy leverages short clip descriptions obtained from a supervised short video model to guide the summary. Each iteration uses three LLMs working in sequence: one to generate the pseudo-summary from clip descriptions, another to evaluate it, and a third to optimise the prompt of the generator. This iteration is necessary because the quality of the pseudo-summaries is highly dependent on the generator prompt, and varies widely among videos. We evaluate our summaries extensively on multiple datasets; our results show that ViSMaP achieves performance comparable to fully supervised state-of-the-art models while generalising across domains without sacrificing performance. Code will be released upon publication.