1,721,010 research outputs found
Communication and Learning in Dynamic Networks
Get PDFAl giorno d'oggi, i sistemi di telecomunicazione sono essenziali in innumerevoli settori e permettono l’utilizzo di tecnologie rivoluzionarie come l’industria 4.0, l’Internet delle cose, la realtà virtuale e il cloud computing. D'altra parte, la maggior parte delle reti di telecomunicazioni presenta ancora un'architettura piuttosto rigida che mal si presta a gestire scenari altamente dinamici con caratteristiche eterogenee. Per far fronte ai requisiti delle tecnologie future, è necessario ideare nuove soluzioni che consentano di riconfigurare servizi e funzioni di rete senza un intervento esterno. La comunità scientifica sta facendo un grande sforzo per realizzare questa visione, studiando strategie per fornire infrastrutture mobili, virtualizzare le risorse disponibili e implementare algoritmi cognitivi che gestiscano autonomamente le funzioni di rete.
In questa tesi affronteremo la sfida di definire un'architettura di rete dinamica, avente tutte le caratteristiche descritte e, pertanto, in grado di adattarsi a diversi scenari di lavoro. Il punto centrale della nostra idea è distribuire le funzioni di rete tra più unità intelligenti che interagiscono tra loro per individuare la configurazione ottima del sistema, massimizzando le prestazioni e garantendo al contempo la resilienza a guasti locali. Per raggiungere questo obiettivo, organizzeremo la tesi in tre parti, ciascuna delle quali indaga un diverso aspetto della gestione dinamica delle reti di telecomunicazioni. Nelle prima parte, definiremo un sistema di tracciamento, basato sull'unscented Kalman filter, per mappare un gruppo di nodi autonomi che trasmettono il loro stato locale mentre operano in diversi scenari operativi. In particolare, studieremo come adattare la strategia di comunicazione di ciascun nodo di rete alla stima dello condizioni del sistema nel suo complesso. Dopodiché, sfrutteremo il paradigma di reinforcement learning e implementeremo un'architettura di apprendimento distribuito che consenta alle singole unità di rete di cooperare e raggiungere un obiettivo comune. In questo contesto, dimostreremo che l'adattamento reciproco delle strategie di comunicazione e controllo porta a prestazioni superiori rispetto a uno scenario di base in cui la comunicazione è predeterminata.
Nell’ultima parte della tesi, studieremo nuove tecniche per gestire le risorse computazionali e di comunicazione nello scenario descritto, sfruttando i paradigmi di network slicing e di software-defined networking. A tal fine, distribuiremo la gestione delle risorse di rete tra più unità intelligenti, che andranno a costituire un’unica architettura adattabile, capace di operare in contesti diversi senza essere riconfigurata da zero. I nostri risultati mostrano che gli algoritmi di apprendimento sono fondamentali per ottenere un'architettura di rete completamente flessibile, che possa rilevare i cambiamenti nell'ambiente circostante e modificare le proprie funzioni di conseguenza. Allo stesso tempo, l'utilizzo dell’intelligenza artificiale richiede di dedicare parte delle risorse di rete disponibili per l’allenamento degli agenti di apprendimento che gestiscono la rete stessa. Questa competizione tra prestazioni immediate e capacità di innovazione di sistema determina un significativo trade-off, a cui dedicheremo l'ultimo capitolo di questa tesi.Nowadays, telecommunication systems are essential to numerous fields, enabling revolutionary applications such as industry 4.0, the Internet of Things, virtual reality, and cloud computing. On the other hand, most telecommunication networks still present a rigid architecture that does not lend itself to managing highly dynamic scenarios with heterogeneous characteristics. In order to address the requirements of future technologies, it is necessary to design new solutions that enable the continuous configuration of services and functions without external intervention. The scientific community is putting strong effort toward such a vision, investigating strategies to provide mobile infrastructures, enable the softwarization of network resources, and deploy cognitive algorithms for orchestrating system operations.
In this thesis, we address the challenge of defining a dynamic network architecture with all these characteristics and, thus, capable of adapting to various working scenarios. The central point of our idea is to distribute network functions among multiple intelligent units that interact with each other to find the optimal system configuration, maximizing performance while ensuring resilience to local failures. To this end, we organize the thesis into three parts, each investigating a different aspect of dynamic network management. Firstly, we define a tracking framework based on the unscented Kalman filter to map a group of autonomous nodes broadcasting their local status while operating in different application scenarios. Hence, we study how to adapt the communication policy of each node in accordance with the estimate of the overall network status. Then, we exploit the reinforcement learning paradigm and implement a distributed learning architecture to allow the single nodes to cooperate and reach a common goal. In this context, we show that adapting the communication and control policy to each other leads to higher performance than the basic scenario in which communication is pre-determined.
Lastly, we investigate new techniques to dynamically manage communication and computational resources in the target scenario, taking advantage of the network slicing and software-defined networking paradigms. To this goal, we distribute the resource allocation tasks among multiple agents, providing a highly flexible architecture that can operate in different contexts without retraining. Our results show that learning algorithms are fundamental for implementing a fully flexible network architecture, capable of detecting changes in the surrounding environment and adapting its functions consequently. At the same time, artificial intelligence requires network designers to devote part of the available resources to support the training of the learning agents managing the network. This competition between immediate performance and innovation capability determines a significant trade-off, to which we devote the last chapter of our research
Design and application of a stratified continuous rumen fermentation system
Get PDFIn the first half of the twentieth century several ruminal in vitro systems were developed. These systems can be classified as “batch culture” or “continuous culture”.
The continuous culture systems (CC) had a quite complex evolution process, and those nowadays used are classified as RUSITEC, Dual Flow Continuous Culture (DFCC) or Single Flow Continuous Culture (SFCC). The diverse CC are characterized by having different design and different working settings, such as: volume and shape of the fermentation vessel, stirring device, artificial saliva, solid and liquid flow and dilution rate, volume and origin of the rumen inoculum, feeding practice and device.
On the bases of the characteristics of the already existing ruminal continuous cultures, a stratified continuous culture has been developed at the Laboratory of Nutrition of the Department of Agricultural and Environmental Sciences of University of Udine. This system has the peculiarity of a simple design and of being built by using cheap and readily available materials. The CC consists of 8 x 2 L glass bottles, warmed at 39°C and placed on a water proof magnetic stirrer. The bottles are closed with a rubber stopper, provided with an insert for the inflow of artificial saliva, and have the outflow at the base.
A first trial has been made to test the stratified continuous culture. In particular, the capability of maintaining protozoa populations and reduce the accumulation of undigested material was investigated in two separated experiments. In Experiment 1 was evaluated the effect of two stirring frequencies (continuous vs intermittent) and two dilution rates (D, 1.29 vs 1.04) on the pH values and on the counts of the protozoa in the fermentation fluid. Experiment 2 aimed to test the effects of varying the daily diet amount (15, 20, 25, 30 g DM/d of a diet with 50:50 forage: concentrate ratio) on the organic matter digestibility and on the fermentative patterns. In 2 fermentation runs of 8 d each, bottles were inoculated with 450 ml of rumen fluid and 1050 ml of artificial saliva, which was then pumped at a rate of 78 ml/h. The tested CC reached stable conditions of fermentation after some days of adaptation and maintained an acceptable concentrations of the protozoa population till the end of the trial. The results of the second experiment suggest that the amount of feed administered to the fermenters has to be properly chosen to limit the accumulation of undigested materials in the glass bottles and maintain the fermentation patterns within acceptable ranges.
A second trial was organized with the aim to understand the effect of urea administration on the fermentation parameters; the trial consisted in two experiments. In the first experiment, a specified dose of urea was used to simulate the regular release of nitrogen in the rumen in a continuous fermenter environment. The aim was to study the rumen metabolism at controlled defined conditions of release of nitrogen, to identify the conditions of the optimal matching between nitrogen release and the rumen bacterial degradation of the carbohydrate fraction. The experiment consisted of 2 runs of 7 days each (4 adaptation days and 3 measuring days). The fermenters were supplied with 12 g DM/d of a diet containing corn silage (30% DM), hay (10% DM), dry corn (50% DM) and a mineral vitamin supplement (10% DM), divided in two equal portions at 09:00 and 15:00. The urea supplement was administered in 1, 2, 3 or 4 doses at 1 hour intervals. The second experiment, performed at the Ruminant Nutrition Laboratory of Utah State University, investigated the effects of supplementing slow-release urea (SRU; Optigen, Alltech, Nicholasville, KY) in 3 lactation dairy diets with different forage-to-concentrate ratios (F:C) on in vitro ruminal fermentation characteristics. The experiment was performed in 4 independent runs of continuous cultures. Each fermenter was offered a diet of 20 g DM/d in 2 equal portions at 08:00 and 20:00 h. In both experiments variations of pH values were detected with the administration of urea at doses usually utilized in ruminant diets and in the range of rumen acidity. In experiment 1, high concentrations of rumen ammonia increased propionate with a tendency of reducing the acetate:propionate ratio. Moreover, in experiment 2, the substitution of fiber from forages with that of byproducts in diets determined a reduction of the ammonia accumulation with a reduction in buffering the rumen; a reduction of the acetate concentrations and an increase in propionate concentrations and a reduction of the methane yield.
The third trial aimed to test the effects of increasing the dose of an active dry Saccharomyces cerevisiae on rumen metabolism in a rumen continuous culture system. The fermenters were supplied with 12 g DM/d of a diet containing corn silage (20% DM), hay (10% DM), alfalfa hay pellet (20% DM), and a compound feed (50% DM), divided in 2 equal portions at 9:00 and 16:00. At each feeding time, all the fermenters received a supplement (0.025 g per feeding) consisting of a commercial yeast extract and/or soybean meal in differing ratios Y:S (0:500, Y0; 5:495, Y5; 50:450, Y50; 500:0, Y500). In the present trial, for several measurements, such as pH, ammonia, and some protozoa numbers, a significant difference between control and yeast inclusion treatments were detected, but there was no dose-treatment effect, even though the doses of yeast followed an exponential rate of increase. A possible explanation of the lack of differences between yeast doses could be because of the favorable in vitro conditions for their survival and multiplication of the yeast (e.g. high pH values, reducing conditions, etc.) which could have favored the colonization of the fermentation bottles to a similar level, independently of the initial dose of yeast.
In general, the stratified continuous fermenter was able to reproduce the ruminal environment in a way to detect even small differences in the parameters measured as a consequence of the diets tested. In the experiments were measured mainly the basic fermentative parameters (pH, VFA, NH3, digestibility), with the future aim to integrate these analysis with technics of molecular biology, to better study the ruminal microbial population
Using Deep Reinforcement Learning to Enhance Channel Sampling Patterns in Integrated Sensing and Communication
Get PDFIn Integrated Sensing and Communication (ISAC) systems, estimating the micro-Doppler (mD) spectrogram of a target requires combining channel estimates retrieved from communication with ad-hoc sensing packets, which cope with the sparsity of the communication traffic. Hence, the mD quality depends on the transmission strategy of the sensing packets, which is still a challenging problem with no known solutions. In this letter, we design a deep Reinforcement Learning (RL) framework that fragments such a problem into a sequence of simpler decisions and takes advantage of the mD temporal evolution for maximizing the reconstruction performance. Our method is the first that learns sampling patterns to directly optimize the mD quality, enabling the adaptation of ISAC systems to variable communication traffic. We validate the proposed approach on a dataset of real channel measurements, reaching up to 40% higher mD reconstruction accuracy and several times lower computational complexity than state-of-the-art methods
Testing a stratified continuous rumen fermenter system
No full textThe capability of maintaining protozoa populations and reduce the accumulation of undigested material was investigated in two experiments using a stratified continuous-culture rumen fermenter (CCF) at various setting.
The CCF consists of 8 x 2 L glass bottles, warmed at 39°C and placed on a water proof magnetic stirrer. Bottles are closed with a rubber stopper, with an insert for the inflow of artificial saliva and have the outflow at the base.
In Experiment 1 the effect of two stirring frequencies (continuous vs intermittent, S1 and S2) and two dilution rates (D, 1.29 vs 1.04) on the pH values and on the counts of the protozoa in the fermentation fluid was evaluated. The pH reached a steady state after 4 d of fermentation and S1 bottles had lower pH than those S2 (6.50 vs 6.57, P<0.001). The protozoa in the fermentation fluid (2.92*105/ml at the beginning) declined sharply in the first 2 d of fermentation and stabilized at about 13*103/ml in the S2 flasks, while the S1 flasks had lower protozoa concentrations (about 0.8*103/ml, P<0.001). The D had no effects on pH and protozoa.
Experiment 2 aimed to test effects of varying the daily diet (F) amount (15, 20, 25, 30 g DM/d of a diet with 50:50 forage: concentrate ratio) on the organic matter (OM) digestibility (OMD) and on the fermentative patterns. In 2 fermentation runs of 8 d each, bottles were inoculated with 450 ml of rumen fluid and 1050 ml of artificial saliva, which was then pumped at a rate of 78 ml/h.
Lowering F reduced the OM accumulated inside the vessels (from 5.06 to 0.68 g/d; P<0.001) but did not affected the OMD for any of the dietary treatments (range between 0.460 and 0.510). The reduction of F led to a linear decrease (P<0.001) in volatile fatty acids concentration (from 65.8 to 45.3 mM), to an increase of acetate:propionate ratio (from 2.91 to 3.42) and to a reduction of pH (from 6.4-6.5 to 6.0-6.1). Lowering F tended to increase the number of protozoa (from 72 to 198*103/ml, P=0.082).
The tested CCF reaches stable conditions of fermentation after some days of adaptation, allows the survival of protozoa population and has a limited accumulation of undigested materials in the glass bottles
Effect of diets differing in protein concentration (low vs medium) and nitrogen source (urea vs soybean meal) on in vitro rumen fermentation and on performance of finishing Italian Simmental bulls
No full textThis research evaluated the effect of two dietary intestinal digestible protein (PDI) levels and a partial substitution of soybean meal (SBM) with urea (U) on the performance of finishing Italian Simmental bulls (Exp. 1) and on in vitro rumen fermentation characteristics (Exp. 2). In Exp. 1, eighty Italian Simmental bulls (495±58 kg of body weight (BW), 14-months old) were fed diets differing in terms of PDI concentration (85 and 72 g/kg dry matter (DM), medium (M) and low (L) diets, respectively) and nitrogen source (only SBM or SBM partly replaced by 0.5% DM of U) during the last 120 d of fattening. Animals were slaughtered at BW of 656±56 kg at 18 months of age. The average daily gain (ADG) was satisfactory (1.32 kg/d) and the medium PDI level tended (P<0.10) to increase the ADG (1.37 vs 1.28 kg/d). Apparent total tract digestibility did not differ between treatments and the feed efficiency tended to be more favourable for the M diets (0.147 vs 0.137 kg BW gain/ kg feed DM, P<0.10). Slaughter traits were unaffected by dietary treatments. In Exp. 2, continuous culture fermenters were inoculated with rumen fluid from bulls of Exp. 1 and were given the same diets. The fermentation fluid was sampled at feeding time and 1, 2 and 3 h after (T0, T1, T2 and T3, respectively) and the medium PDI level determined higher ammonia concentrations (P<0.01 at T0, T1, and T2; P<0.05 at T3). The drop of pH at T1 and T2 was less intense (P<0.05) for diets containing U, presumably due to the buffering capacity of urea. In vitro DM digestibility tended to be higher for diets containing U (P<0.10) and was unaffected by the dietary level of PDI. Differences in volatile fatty acids concentrations were limited to butyrate, which was higher for M diets (T0:13.6 vs 12.5 mol/100 mol, P<0.10; T2: 13.8 vs 12.5 mol/100 mol, P<0.01) and for diets containing U (T0: 14.3 vs 11.9 mol/100 mol; T2: 14.3 vs 12.1 mol/100 mol, P<0.01). In conclusion, L diets tended to decrease weight gain in finishing Italian Simmental bulls, but both PDI levels and SBM substitution with U had no detrimental effects on slaughter and meat quality traits. Further research efforts are required to explain the increase of butyrate in fermentation fluid of fermenters fed medium PDI diets or U diets. © 2016 Elsevier B.V
Going Beyond Counting First Authors in Author Co-citation Analysis
Get PDFThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
Variations on the Author
Get PDF“Variations on the Author” discusses two of Eduardo Coutinho’s recent films (Um Dia na Vida, from 2010, and Últimas Conversas, posthumously released in 2015) and their contribution to the general question of documentary authorship. The director’s filmography is characterized by a consistent yet self-effacing form of authorial self-inscription: Coutinho often features as an interviewer that rather than express opinions propels discourses; an interviewer that is good at listening. This mode of self-inscription characterizes him as an author who is not expressive but who is nonetheless markedly present on the screen. In Um Dia na Vida, however, Coutinho is completely absent form the image, while Últimas Conversas, on the contrary, includes a confessional prologue that moves the director from the margins to the center of his films. This article examines the ways in which these works stand out in the filmography of a director who offers new insights into the notion of cinematic authorship
Appropriate Similarity Measures for Author Cocitation Analysis
Get PDFWe provide a number of new insights into the methodological discussion about author cocitation analysis. We first argue that the use of the Pearson correlation for measuring the similarity between authors’ cocitation profiles is not very satisfactory. We then discuss what kind of similarity measures may be used as an alternative to the Pearson correlation. We consider three similarity measures in particular. One is the well-known cosine. The other two similarity measures have not been used before in the bibliometric literature. Finally, we show by means of an example that our findings have a high practical relevance.information science;Pearson correlation;cosine;similarity measure;author cocitation analysis
- …
