A deep learning model via thermodynamics and condensed matter physics

Deep Learning architectures give brilliant results in a large variety of fields, but a comprehensive theoretical description of their inner functioning is still lacking. In this work, we try to understand the behavior of neural networks by modelling in the frameworks of Thermodynamics and Condensed Matter Physics. We approach neural networks as in a real laboratory and we measure the frequency spectrum and the entropy of the weights of the trained model. The stochasticity of the training occupies a central role in the dynamics of the weights and makes it difficult to assimilate neural networks to simple physical systems. However, the analogy with Thermodynamics and the introduction of a well defined temperature leads us to an interesting result: if we eliminate from a CNN the "hottest" filters, the performance of the model remains the same, whereas, if we eliminate the "coldest" ones, the performance gets drastically worst. This result could be exploited in the realization of a training loop which eliminates the filters that do not contribute to loss reduction. In this way, the computational cost of the training will be lightened and more importantly this would be done by following a physical model. In any case, beside important practical applications, our analysis proves that a new and improved modeling of Deep Learning systems can pave the way to new and more efficient algorithms.

A deep learning model for controlling the galvanizing process in a production line

In the industry of steelmaking, the process of galvanizing is a treatment which is applied to protect the steel from corrosion. The air knife effect (AKE) occurs when nozzles emit a steam of air on the surfaces of a steel strip to remove excess zinc from it. In our work we formalized the problem to control the AKE and we implemented, with the R&D dept.of MarcegagliaSPA, a DL model able to drive the AKE. We call it controller. It takes as input the tuple : a tuple of the physical conditions of the process line (t,h,s) with the target value of the zinc coating (c); and generates the expected tuple of (pres and dist) to drive the mechanical nozzles towards the (c). According to the requirements we designed the structure of the network. We collected and explored the data set of the historical data of the smart factory. Finally, we designed the loss function as sum of three components: the minimization between the coating addressed by the network and the target value we want to reach; and two weighted minimization components for both pressure and distance. In our solution we construct a second module, named coating net, to predict the coating of zinc resulting from the AKE when the conditions are applied to the prod. line. Its structure is made by a linear and a deep nonlinear “residual” component learned by empirical observations. The predictions made by the coating nets are used as ground truth in the loss function of the controller. By tuning the weights of the different components of the loss function, it is possible to train models with slightly different optimization purposes. In the tests we compared the regularization of different strategies with the standard one in condition of optimal estimation for both; the overall accuracy is ± 3 g/m^2 dal target for all of them. Lastly, we analyze how the controller modeled the current solutions with the new logic: the sub-optimal values of pres and dist can be optimize of 50% and 20%.

Progettazione di un kit di supporto ai corsi Project Based Learning: il ruolo del tutor nel modello Tech to Market.

L’istruzione superiore in Europa è stata oggetto di un significativo processo di riforma: è aumentato l’interesse per un modello di apprendimento intorno ai progetti, centrato sullo studente, che favorisse lo sviluppo di competenze trasversali – il project-based learning (PBL). Inserire il PBL nelle Università richiede un processo di innovazione didattica: il curriculum di un corso PBL e le competenze richieste all’insegnante si differenziano dall’apprendimento tradizionale. Senza un'adeguata attenzione ai metodi di supporto per insegnanti e studenti, questi approcci innovativi non saranno ampiamente adottati. L’obiettivo di questo studio è determinare in che modo sia possibile implementare un corso PBL non presenziato da figure esperte di PBL. Le domande della ricerca sono: è possibile implementare efficacemente un approccio PBL senza il coinvolgimento di esperti dei metodi di progettazione? come si declinano i ruoli della facilitazione secondo questa configurazione: come si definisce il ruolo di tutor d’aula? come rafforzare il supporto per l’implementazione del corso? Per rispondere alle domande di ricerca è stata utilizzata la metodologia AIM-R. Viene presentata la prima iterazione dell’implementazione di un corso di questo tipo, durante la quale sono state svolte attività di ricerca e raccolta dati. L’attività di facilitazione è affidata a tre figure diverse: docente, tutor d’aula e coach professionisti. Su questa base, sono stati definiti gli elementi costituenti un kit di materiale a supporto per l’implementazione di corsi PBL. Oltre a un set di documenti e strumenti condivisi, sono stati elaborati i vademecum per guidare studenti, tutor e docenti all’implementazione di questo tipo di corsi. Ricerche future dovranno essere volte a identificare fattori aggiuntivi che rendano applicabile il kit di supporto per corsi basati su un modello diverso dal Tech to Market o che utilizzino strumenti di progettazione diversi da quelli proposti durante la prima iterazione.

Sviluppo di una piattaforma per fornire contenuti formativi sfruttando la gamification: un caso di studio aziendale

Questo volume di tesi, dal titolo “Sviluppo di una piattaforma per fornire contenuti formativi sfruttando la gamification: un caso di studio aziendale”, tratta argomenti quali e-learning e game-based learning e come/quando questi possono essere applicati, presentando inoltre un esempio di prototipo di applicazione web che può fungere a questo scopo. Nello specifico, il primo capitolo si compone di tre sezioni principali: la prima introduce il concetto di e-learning e le molteplici declinazioni ad esso applicabili, oltre a presentare qualche cenno di carattere storico per individuare questo fenomeno nel tempo; la seconda tratta i campi d’applicazione e le tipologie di didattica inscrivibili nel termine “Game-based learning”. Nella terza sezione, “builder per esperienze gamificate”, infine, vengono presentate e analizzate due applicazioni web che possono concorrere alla creazione di un’esperienza di formazione gamificata in ambito scolastico e/o lavorativo. Il secondo e il terzo capitolo, rispettivamente con titoli “Tecnologie” e “Applicazione web: BKM – Learning Game”, sono fortemente correlati: vengono infatti presentate le tecnologie (nello specifico HTML, CSS, Javascript, NodeJs, VueJs e JSON) utilizzate per la creazione del progetto di tesi, poi viene descritto l’applicativo web risultante nel suo complesso. Il progetto è stato implementato durante il tirocinio in preparazione della prova finale, presso l’azienda Bookmark s.r.l.

A biophysical model for valence-based navigational learning: the role of D-Serine

Recent experiments have revealed the fundamental importance of neuromodulatory action on activity-dependent synaptic plasticity underlying behavioral learning and spatial memory formation. Neuromodulators affect synaptic plasticity through the modification of the dynamics of receptors on the synaptic membrane. However, chemical substances other than neuromodulators, such as receptors co-agonists, can influence the receptors' dynamics and thus participate in determining plasticity. Here we focus on D-serine, which has been observed to affect the activity thresholds of synaptic plasticity by co-activating NMDA receptors. We use a computational model for spatial value learning with plasticity between two place cell layers. The D-serine release is CB1R mediated and the model reproduces the impairment of spatial memory due to the astrocytic CB1R knockout for a mouse navigating in the Morris water maze. The addition of path-constraining obstacles shows how performance impairment depends on the environment's topology. The model can explain the experimental evidence and produce useful testable predictions to increase our understanding of the complex mechanisms underlying learning.

Structure learning of graphical models for task-oriented robot grasping

In the collective imaginaries a robot is a human like machine as any androids in science fiction. However the type of robots that you will encounter most frequently are machinery that do work that is too dangerous, boring or onerous. Most of the robots in the world are of this type. They can be found in auto, medical, manufacturing and space industries. Therefore a robot is a system that contains sensors, control systems, manipulators, power supplies and software all working together to perform a task. The development and use of such a system is an active area of research and one of the main problems is the development of interaction skills with the surrounding environment, which include the ability to grasp objects. To perform this task the robot needs to sense the environment and acquire the object informations, physical attributes that may influence a grasp. Humans can solve this grasping problem easily due to their past experiences, that is why many researchers are approaching it from a machine learning perspective finding grasp of an object using information of already known objects. But humans can select the best grasp amongst a vast repertoire not only considering the physical attributes of the object to grasp but even to obtain a certain effect. This is why in our case the study in the area of robot manipulation is focused on grasping and integrating symbolic tasks with data gained through sensors. The learning model is based on Bayesian Network to encode the statistical dependencies between the data collected by the sensors and the symbolic task. This data representation has several advantages. It allows to take into account the uncertainty of the real world, allowing to deal with sensor noise, encodes notion of causality and provides an unified network for learning. Since the network is actually implemented and based on the human expert knowledge, it is very interesting to implement an automated method to learn the structure as in the future more tasks and object features can be introduced and a complex network design based only on human expert knowledge can become unreliable. Since structure learning algorithms presents some weaknesses, the goal of this thesis is to analyze real data used in the network modeled by the human expert, implement a feasible structure learning approach and compare the results with the network designed by the expert in order to possibly enhance it.

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.

Analysis of Lead-Acid batteries with a third-order dynamical model

Electrical energy storage is a really important issue nowadays. As electricity is not easy to be directly stored, it can be stored in other forms and converted back to electricity when needed. As a consequence, storage technologies for electricity can be classified by the form of storage, and in particular we focus on electrochemical energy storage systems, better known as electrochemical batteries. Largely the more widespread batteries are the Lead-Acid ones, in the two main types known as flooded and valve-regulated. Batteries need to be present in many important applications such as in renewable energy systems and in motor vehicles. Consequently, in order to simulate these complex electrical systems, reliable battery models are needed. Although there exist some models developed by experts of chemistry, they are too complex and not expressed in terms of electrical networks. Thus, they are not convenient for a practical use by electrical engineers, who need to interface these models with other electrical systems models, usually described by means of electrical circuits. There are many techniques available in literature by which a battery can be modeled. Starting from the Thevenin based electrical model, it can be adapted to be more reliable for Lead-Acid battery type, with the addition of a parasitic reaction branch and a parallel network. The third-order formulation of this model can be chosen, being a trustworthy general-purpose model, characterized by a good ratio between accuracy and complexity. Considering the equivalent circuit network, all the useful equations describing the battery model are discussed, and then implemented one by one in Matlab/Simulink. The model has been finally validated, and then used to simulate the battery behaviour in different typical conditions.

Methods for monitoring Machine Learning models

The following thesis aims to investigate the issues concerning the maintenance of a Machine Learning model over time, both about the versioning of the model itself and the data on which it is trained and about data monitoring tools and their distribution. The themes of Data Drift and Concept Drift were then explored and the performance of some of the most popular techniques in the field of Anomaly detection, such as VAE, PCA, and Monte Carlo Dropout, were evaluated.

Cyber threat intelligence: identifying hardcoded secrets in GitHub

The usage of version control systems and the capabilities of storing the source code in public platforms such as GitHub increased the number of passwords, API Keys and tokens that can be found and used causing a massive security issue for people and companies. In this project, SAP's secret scanner Credential Digger is presented. How it can scan repositories to detect hardcoded secrets and how it manages to filter out the false positives between them. Moreover, how I have implemented the Credential Digger's pre-commit hook. A performance comparison between three different implementations of the hook based on how it interacts with the Machine Learning model is presented. This project also includes how it is possible to use already detected credentials to decrease the number false positive by leveraging the similarity between leaks by using the Bucket System.