1,721,245 research outputs found
Towards Real-World Data Streams for Deep Continual Learning
Full text linkContinual Learning deals with Artificial Intelligent agents striving to learn from an ever-ending
stream of data. Recently, Deep Continual Learning focused on the design of new strategies to
endow Artificial Neural Networks with the ability to learn continuously without forgetting previous
knowledge. In fact, the learning process of any Artificial Neural Network model is well-known to
lack the sufficient stability to preserve existing knowledge when learning new information. This
phenomenon, called catastrophic forgetting or simply forgetting, is considered one of the main
obstacles for the design of effective Continual Learning agents. However, existing strategies designed
to mitigate forgetting have been evaluated on a restricted set of Continual Learning scenarios. The
most used one is, by far, the Class-Incremental scenario applied on object detection tasks. Even
though it drove interest in Continual Learning, Class-Incremental scenarios strongly constraint the
properties of the data stream, thus limiting its ability to model real-world environments.
The core of this thesis concerns the introduction of three Continual Learning data streams, whose
design is centered around specific real-world environments properties. First, we propose the Class-
Incremental with Repetition scenario, which builds a data stream including both the introduction
of new concepts and the repetition of previous ones. Repetition is naturally present in many
environments and it constitutes an important source of information. Second, we formalize the
Continual Pre-Training scenario, which leverages a data stream of unstructured knowledge to keep
a pre-trained model updated over time. One important objective of this scenario is to study how to
continuously build general, robust representations that does not strongly depend on the specific task
to be solved. This is a fundamental property of real-world agents, which build cross-task knowledge
and then adapts it to specific needs. Third, we study Continual Learning scenarios where data
streams are composed by temporally-correlated data. Temporal correlation is ubiquitous and lies
at the foundation of most environments we, as humans, experience during our life. We leverage
Recurrent Neural Networks as our main model, due to their intrinsic ability to model temporal
correlations. We discovered that, when applied to recurrent models, Continual Learning strategies
behave in an unexpected manner. This highlights the limits of the current experimental validation,
mostly focused on Computer Vision tasks.
Ultimately, the introduction of new data streams contributed to deepen our understanding of
how Artificial Neural Networks learn continuously. We discover that forgetting strongly depends
on the properties of the data stream and we observed large changes from one data stream to
another. Moreover, when forgetting is mild, we were able to effectively mitigate it with simple
strategies, or even without any specific ones. Loosening the focus on forgetting allows us to turn our
attention to other interesting problems, outlined in this thesis, like (i) separation between continual
representation learning and quick adaptation to novel tasks, (ii) robustness to unbalanced data
streams and (iii) ability to continuously learn temporal correlations. These objectives currently
defy existing strategies and will likely represent the next challenge for Continual Learning research
Le praterie a Posidonia oceanica quali indicatori biologici per una corretta Valutazione di Impatto Ambientale (VIA) della fascia costiera. In: Bettini V., Zannin D. (Eds.) - Portoscuso. La gassificazione della V.I.A. Il gassificatore Ati-Sulcis, le problematiche ambientali connesse al progetto
No full textItalian Sociology,1945–2010. An Intellectual and Institutional Profile
No full textThis book provides a comprehensive profile of the development of sociology in Italy from the post-war period to the present day. The first English-language account of the history of Italian sociology, it focuses on the process of institutionalization of the discipline within the Italian university system and its changing relationships with extra-academic actors and institutions: political parties, unions, the Catholic Church, political and social movements, as well as local and national governments. Arranged chronologically across eight chapters, it presents all major steps in the development of the discipline in a theoretically-informed but accessible way. The authors explore the pioneering phase of the 1950s to the establishment of the first academic chairs in the 1960s, from the student revolts of 1968 to the creation of the first sociological association in the 1980s and up to the present day. It will appeal to social science and history scholars and students, as well as readers interested in the history of Contemporary Italy
In the field but not of the field: Clifford Geertz, Robert Bellah, and the practices of interdisciplinarity
No full textThe intellectual trajectories of social scientists Robert N. Bellah and Clifford Geertz are compared as a case study in the production of successful interdisciplinary work. Geertz and Bellah started from a similar position, in terms of scholarly habits, network cen- trality, and symbolic capital. However, while Geertz became an interdisciplinary star and left his mark in disciplines as diverse as history, sociology, and cultural studies, Bellah’s interdisciplinary appeal was more limited, while his ability to speak to the general public as a public intellectual was unmatched by Geertz. We thus review Bellah’s and Geertz’s parallel careers using a multidimensional analytical model intended to complete current field-based and performative-pragmatist models of intellectual success, arguing that interdisciplinary success can be accounted for by a combination of local ecological fac- tors, images of intellectual work, and texts showing a high degree of cross-disciplinary fluency
Lo Stato trofico dello stagno di Platamona
Full text linkSeasonal changes of nutrient concentrations and other chemical and biological parameters in Platamona's pond were investigated to asses its trophic level. The pond shows eutrophic. There is a large range of seasonal variations of totale alcalinity concentrations (3.3 - 6.9 meq l-1). The littoral and submersed macrophytes are very plentiful and the mean annual value of clorophylla(16 mg m-3) is high. Nevertheless the mean of concentrations of nitrate nitrogen (106 mg m-3), ammonia nitrogen (30 mg m-3) and reactive phosphorus (4 m-3) are low
Azathioprine-Induced Peripheral T Cell Apoptosis And Drug Response In Patients With Crohn’s Disease
Full text linkBackground and Aim: the long time interval for a trial of thioupurine therapy and the potential side effects in spite of the proven efficacy, do not encourage their use as early therapeutic option in Crohn’s Disease (CD). The development of tests predictive of responsiveness represents a major attempt in the clinical management of CD patients. Azathioprine (AZA) is able to induce apoptosis of T cells; therefore we analyzed the “in vitro” thiopurine-induced T cells apoptosis in a group of CD patients with known response to a previous treatment with AZA. Methods: peripheral CD4+ T cells from 16 CD patients were stimulated with antiCD3/CD28 mAbs in the presence or absence of AZA or 6-MP or 6-thioguanine; apoptosis was assessed using Annexin V staining. Results: Apoptosis stimulation index (% of apoptotic cells in the presence of thiopurine / % of apoptotic cells in their absence) was significantly lower in non responder when compared to responder patients (1.46 (0.97-1.8) vs. 2.19 (1.58-2.65) median (range), respectively; p=0.002 by Mann Whitney test). Conclusions: evaluation of apoptosis stimulation index of peripheral CD4+T cell induced by AZA might represent a parameter useful for a proper selection of CD patients candidate to thiopurine treatment
The Spider and the Fly. Authenticity, Dualism, and the Rolling Stones
Full text linkThis paper is dedicated to the historical and analytical deconstruction of the dualism between Jagger and Richards. We show how it slowly emerged from a more homogeneous vision of the Stones as an “organic” collective, and we claim that this internal differentiation has helped to anchor the Rolling Stones to a narrative of authenticity. Over the years, Keith Richards has become the embodiment of this authenticity, a process that started in the mid-1970s, parallel to a process of increasing individuation of the members of the band. The emergence and stabilization of this dualism have become essential to understand the contemporaneity of the Rolling Stones and the balance of independence and commitment that has marked their life in the past three decades
The high nasal flow therapy for pre-oxygenation: a new strategy for conventional procedures in intensive care settings?
No full textEndotracheal intubation is one of the most routinely performed invasive procedures, but also frequently associated with morbidity and mortality.
Severe complications can occur and hypoxemia is most often reported in case of difficult airway access.
In the last decades, the interest for High-Flow Nasal Cannula oxygen administration as an alternative technique for NIV or oxygenation has increased.
The new technique represents a real attractive topic and its application for preventing hypoxemia during intubation in critically ill patients is acquiring interest as demonstrated by the growing number of clinical trials on this argument.
As reported by the review of Ricard JD a growing body of evidence suggests that high flow oxygen therapy is an innovative and effective technique to improve oxygenation and prevent desaturation with different underlying diseases
- …
