Tuning automatico di parametri in un sistema di controllo semaforico auto-organizzante

L'elaborato di tesi tratta dei vantaggi ottenibili dall'uso di tecniche di automatic parameter tuning, applicando un'implementazione di iterated racing su di un innovativo sistema di controllo semaforico auto-organizzante ispirato da concetti di swarm intelligence.

Tuning remoto di filtering, pre-processing e buffering per scenari internet of things

La tesi propone una soluzione middleware per scenari in cui i sensori producono un numero elevato di dati che è necessario gestire ed elaborare attraverso operazioni di preprocessing, filtering e buffering al fine di migliorare l'efficienza di comunicazione e del consumo di banda nel rispetto di vincoli energetici e computazionali. E'possibile effettuare l'ottimizzazione di questi componenti attraverso operazioni di tuning remoto.

Sub-wavelength resonators as seismic Rayleigh waves shield

We explore the thesis that tall structures can be protected by means of seismic metamaterials. Seismic metamaterials can be built as some elements are created over soil layer with different shapes, dimensions, patterns and from different materials. Resonances in these elements are acting as locally resonant metamaterials for Rayleigh surface waves in the geophysics context. Analytically we proved that if we put infinite chain of SDOF resonator over the soil layer as an elastic, homogeneous and isotropic material, vertical component of Rayleigh wave, longitudinal resonance of oscillators will couple with each other, they would create a Rayleigh bandgap frequency, and wave will experience attenuation before it reaches the structure. As it is impossible to use infinite chain of resonators over soil layer, we considered finite number of resonators throughout our simulations. Analytical work is interpreted using finite element simulations that demonstrates the observed attenuation is due to bandgaps when oscillators are arranged at sub-wavelength scale with respect to the incident Rayleigh wave. For wavelength less than 5 meters, the resulting bandgaps are remarkably large and strongly attenuating when impedance of oscillators matches impedance of soil. Since longitudinal resonance of SDOF resonator are proportional to its length inversely, a formed array of resonators that attenuates Rayleigh waves at frequency ≤10 Hz could be designed starting from vertical pillars coupled to the ground. Optimum number of vertical pillars and their interval spacing called effective area of resonators are investigated. For 10 pillars with effective area of 1 meter and resonance frequency of 4.9 Hz, bandgap frequency causes attenuation and a sinusoidal impulsive force illustrate wave steering down phenomena. Simulation results proved analytical findings of this work.

Intermediate linguistic task fine-tuning on Multi-lingual models

State-of-the-art NLP systems are generally based on the assumption that the underlying models are provided with vast datasets to train on. However, especially when working in multi-lingual contexts, datasets are often scarce, thus more research should be carried out in this field. This thesis investigates the benefits of introducing an additional training step when fine-tuning NLP models, named Intermediate Training, which could be exploited to augment the data used for the training phase. The Intermediate Training step is applied by training models on NLP tasks that are not strictly related to the target task, aiming to verify if the models are able to leverage the learned knowledge of such tasks. Furthermore, in order to better analyze the synergies between different categories of NLP tasks, experimentations have been extended also to Multi-Task Training, in which the model is trained on multiple tasks at the same time.