Deep learning for computer vision: a comparison between convolutional neural networks and hierarchical temporal memories on object recognition tasks

In recent years, Deep Learning techniques have shown to perform well on a large variety of problems both in Computer Vision and Natural Language Processing, reaching and often surpassing the state of the art on many tasks. The rise of deep learning is also revolutionizing the entire field of Machine Learning and Pattern Recognition pushing forward the concepts of automatic feature extraction and unsupervised learning in general. However, despite the strong success both in science and business, deep learning has its own limitations. It is often questioned if such techniques are only some kind of brute-force statistical approaches and if they can only work in the context of High Performance Computing with tons of data. Another important question is whether they are really biologically inspired, as claimed in certain cases, and if they can scale well in terms of "intelligence". The dissertation is focused on trying to answer these key questions in the context of Computer Vision and, in particular, Object Recognition, a task that has been heavily revolutionized by recent advances in the field. Practically speaking, these answers are based on an exhaustive comparison between two, very different, deep learning techniques on the aforementioned task: Convolutional Neural Network (CNN) and Hierarchical Temporal memory (HTM). They stand for two different approaches and points of view within the big hat of deep learning and are the best choices to understand and point out strengths and weaknesses of each of them. CNN is considered one of the most classic and powerful supervised methods used today in machine learning and pattern recognition, especially in object recognition. CNNs are well received and accepted by the scientific community and are already deployed in large corporation like Google and Facebook for solving face recognition and image auto-tagging problems. HTM, on the other hand, is known as a new emerging paradigm and a new meanly-unsupervised method, that is more biologically inspired. It tries to gain more insights from the computational neuroscience community in order to incorporate concepts like time, context and attention during the learning process which are typical of the human brain. In the end, the thesis is supposed to prove that in certain cases, with a lower quantity of data, HTM can outperform CNN.

Geometric versus Model Predictive Control based guidance algorithms for fixed-wing UAVs in the presence of very strong wind fields.

The recent years have witnessed increased development of small, autonomous fixed-wing Unmanned Aerial Vehicles (UAVs). In order to unlock widespread applicability of these platforms, they need to be capable of operating under a variety of environmental conditions. Due to their small size, low weight, and low speeds, they require the capability of coping with wind speeds that are approaching or even faster than the nominal airspeed. In this thesis, a nonlinear-geometric guidance strategy is presented, addressing this problem. More broadly, a methodology is proposed for the high-level control of non-holonomic unicycle-like vehicles in the presence of strong flowfields (e.g. winds, underwater currents) which may outreach the maximum vehicle speed. The proposed strategy guarantees convergence to a safe and stable vehicle configuration with respect to the flowfield, while preserving some tracking performance with respect to the target path. As an alternative approach, an algorithm based on Model Predictive Control (MPC) is developed, and a comparison between advantages and disadvantages of both approaches is drawn. Evaluations in simulations and a challenging real-world flight experiment in very windy conditions confirm the feasibility of the proposed guidance approach.

A comparison between vertical mixing parameterizations in the Levantine Sea

The goal of this thesis was the study of an optimal vertical mixing parameterization scheme in a mesoscale dominated field characterized from a strong vorticity and the presence of a layer of colder, less saline water at about 100 m depth (Atlantic Waters); in these conditions we compared six different experiments, that differ by the turbulent closure schemes, the presence or not of an enhanced diffusion parameterization and the presence or not of a double diffusion mixing parameterization. To evaluate the performance of the experiments and the model we compared the simulations with the ARGO observations of temperature and salinity available in our domain, in our period of interest. The conclusions were the following: • the increase of the resolution gives better results in terms of temperature in all the considered cases, and in terms of salinity. • The comparisons between the Pacanovski-Philander and the TKE turbulent closure schemes don’t show significant differences when the simulations are compared to the observations. • The removing of the enhanced diffusion parameterization in presence of the TKE turbulent closure submodel doesn’t give positive results, and show limitations in the resolving of gravitational instabilities near the surface • The k-ϵ turbulent closure model utilized in all the GLS experiments, is the best performing closure model among the three considered, with positive results in all the salinity comparison with the in situ observation and in most of the temperature comparisons. • The double mixing parameterization utilized in the k-ϵ closure submodel improves the results of the experiments improving both the temperature and salinity in comparison with the ARGO data.