Giovani in un'ora - Ciclo di seminari - Quarta parte

Day - Time: 26 October 2023, h.11:00
Place: Area della Ricerca CNR di Pisa - Room: C-29

Fabio Carrara


Francesco Conti - "An automatic pipeline for topological machine learning"

Abstract: In this work, we develop a pipeline that associates Persistence Diagrams to digital data via the most appropriate filtration for the type of data considered. Using a grid search approach, this pipeline determines optimal representation methods and parameters. The development of such a topological pipeline for Machine Learning involves two crucial steps that strongly affect its performance: firstly, digital data must be represented as an algebraic object with a proper associated filtration in order to compute its topological summary, the Persistence Diagram. Secondly, the persistence diagram must be transformed with suitable representation methods in order to be introduced in a Machine Learning algorithm. We assess the performance of our pipeline, and in parallel, we compare the different representation methods on popular benchmark datasets. This work is a first step toward both an easy and ready-to-use pipeline for data classification using persistent homology and Machine Learning, and to understand the theoretical reasons why, given a dataset and a task to be performed, a pair (filtration, topological representation) is better than another.

Antonino Crivello - "Audiometry with pupil response"

Abstract: Human hearing is generally evaluated through the individual feedback of patients. In standard hearing tests, the feedback from a patient consists of pressing a button, raising a hand, or giving a verbal response to confirm or deny the detection of tones at varying amplitudes and frequencies. The measurement of cognitive resource allocation during hearing perception, or delivering listening efforts, provides valuable insight into the factors influencing auditory processing. This talk will overview the use of the pupillary dilation response (PDR), a short-latency component of the orienting response evoked by novel stimuli, as an indicator of sound detection. PDR can be considered a physiological signal and requires no voluntary reports. The main purpose is to show the feasibility of performing audiometric tests from a subjective to an objective examination.

Lucia Vadicamo - "Exploiting the nSimplex Projection for Permutation-Based Indexing "

Abstract: Approximate Nearest Neighbour (ANN) search is the prevalent paradigm for searching intrinsically high dimensional objects in large-scale data sets, especially when the "curse of dimensionality" prevents exact searches. Many ANN methods involve mapping the data objects into more manageable spaces for efficient retrieval. Among these, Permutation-Based Indexing (PBI) methods have attracted a lot of interest due to their versatility in handling general metric spaces. In PBI, database objects are represented as a sequence of identifiers (permutations) and then efficiently indexed using data structures like inverted files. While PBI methods excel at efficiently selecting a candidate set of results for a given query, achieving high recall often requires refining this candidate set using the original metric and data.

This seminar will provide an overview of a space transformation,  known as nSimplex projection, and its usage in PBI to create novel permutation-based representations or improving candidate result refinement without direct access to the original data. The nSimplex projection leverages distances to a set of reference objects to map metric objects into a finite-dimensional Euclidean space, where upper and lower bounds for the original distances can be calculated. It is applicable to metric spaces satisfying the n-point property, which provides geometric guarantees stronger than the triangle inequality. As such, its potential applications extend beyond the realm of PBI. Therefore, this seminar aims also to foster discussions among attending researchers on the possible applications of the nSimplex projection in other research domains.