Browsing by Author "Guillot Suarez, Calvin"

Now showing 1 - 2 of 2

Human-in-the-loop Hyperparameter Tuning of Deep Nets to Improve Explainability of Classifications
(2022-05-16) Guillot Suarez, Calvin
Sähkötekniikan korkeakoulu | Master's thesis
Artificial Intelligence methods, especially the fields of deep-learning and other neural network based architectures have seen an increasing amount of development and deployment over the last decade. These architectures are especially suited to learning from large volumes of labelled data, and even though we know how they are constructed, they turn out to be equivalent to black boxes when it comes to understanding the basis upon which they produce predictions, especially as size of the network increases. Explainable AI (xAI) methods aim to disclose the key features and values that influence the prediction of black-box classifiers in a manner that is understandable to humans. In this project, the first steps are taken towards developing an interactive xAI system that places a human in the loop; here, a user’s ratings on the sensibility of explanations of individual classifications are used to iteratively find Hyperparameters of the neural net classifier (VGG-16), image segmentator (Felzenszwalb), and xAI (SHAP), to improve the sensibility of the explanations produced without affecting classification accuracy of the classifier in the training set. The users are asked to rate the sensibility of explanation from 1-10. The rating from the users is fed back to the Bayesian optimization algorithm that suggests new Hyperparameters values for the classifier, segmentator, and SHAP modules. The results of the user study suggests that the Hyperparameters which produced higher ratings on explanations tended to also improve the explainability of the images, thus generally improving the explainability for the image class. Improvement in the out-of-sample accuracy of the classifier (for the same class) was observed in some scenarios, but this still needs more comprehensive evaluation. More sensitive queries for the users, explore a variety of xAI methods, a variety of datasets, as well as conduct larger-scale experiments with users would be required to jointly improve explanations of multiple classes.
Sonorama: An exploration of sonic artificial life
(2024-12-31) Guillot Suarez, Calvin
School of Arts, Design and Architecture | Master's thesis
Acoustic ecology is an interdisciplinary field that aims to understand the relationship mediated through sound between human beings and their environment. The impact of industrialization and urbanization has led to a growing concern about noise pollution, its effects on wildlife, and how humans understand and perceive sounds in their increasingly loud habitats. Within this field, the niche hypothesis proposes that each species occupies a unique acoustic niche in the environment to minimize interference and maximize communication efficiency. Therefore, using these principles, this thesis aims to create a system that can simulate a natural sonic ecosystem that can react and adapt to natural and artificial sonic inputs. This work is part of a larger art project called R-Bus, where an autonomous driverless bus roams the streets of Helsinki. Urban soundscapes are captured by microphones deployed around the city and fed into an artificial life physarum simulation (Alife), where the system processes the incoming sound by transforming the signals into a spectrogram, then feeding this into the simulation, and finally producing sounds by using IFFT from the resulting simulated image. The agent-based simulation is controlled by a neural network (NN) guided by an evolutionary genetic algorithm (NEAT). This process determines the ecosystem’s behavior and final sonic expression. The system also implements autopoietic and sympoietic concepts that describe life as self-organizing and co-evolving systems to produce a naturalistic evolutionary process. The resulting audio signal is played for the audience inside the bus alongside the signals from the microphones and other soundscapes. The sounds produced by the ALife simulation vary widely, ranging from whale-like to machine-like and encompassing various insect-like noises and other mechanical sounds. Niche differentiation was achieved, though the agents’ lifetimes were relatively short, and convergence was not always observed. Approximately 1,200 people experienced the R-Bus installation. For some participants, the sounds were familiar, understandable, and even described as beautiful, while for others, the synthetic sounds were imperceptible. This project highlights the disruption that human-generated sounds cause to natural environments and offers a system that could be used to understand this disruption better, as well as the relationship between humans and their sonic environment.