5,193 research outputs found
Letter by Picichè and Versaci Regarding Article, "More Data Than Options for the 'No-Option' Refractory Angina Patient in the United States"
No full textApplications of Control Theory to Computer Systems Optimization
Full text linkComputer systems complexity is growing rapidly, thus making the efficient
use of their resources an increasingly challenging task. In many cases the
optimization of the management in such systems has been developed with
ad-hoc techniques and heuristics.
In this thesis, a more general and flexible approach is explored to resource
management, based on the powerful framework of stochastic optimal control.
This approach requires a careful modeling of the system of interest as a
dynamical system with appropriate cost functions and stochastic descriptions
of the inputs that are imposed by the external environment. Then, the
question of an optimal management policy that minimizes the expected cost
becomes mathematically well posed and can be systematically investigated.
Two cases studies illustrating the approach are developed, as summarized
below.
In Chapters 1–5, the classical replacement problem for memory hierarchies
is cast within the framework of optimal control theory. Memory references
are assumed to comply with the Least Recently Used Stack Model
(LRUSM); arbitrary stack-distance distributions are considered.
An optimal policy is derived to minimize the miss rate for an infinite trace
(a control over an infinite horizon). We call it a K-L policy where K(C)
and L(C) are parameters, whose value is a function of the buffer (cache)
size C, determined by the stack-distance distribution. Then, the concept of
Least Profit Rate (LPR) policy is introduced and it is shown that, for the
LRUSM model, LPR is an optimal policy over an infinite horizon which, in
steady state, coincides with the K-L policy. The LPR satisfies the inclusion
property whereby the content of a given buffer is also contained in all larger
buffers. This property is known to enable the efficient computation of the
number of misses for a given address trace, simultaneously for all buffer sizes.
Furthermore, the LPR formulation leads to a linear time computation of the
values K(C) and L(C) for all relevant values of C, improving on the cubic
bound that naturally arises within the K-L derivation. Furthermore, the
properties of LPR are exploited to derive an efficient algorithm to optimally
partition a buffer concurrently accessed by multiple processes. Finally, the
miss rate of LPR is compared with that of OPT, the well-known optimal
off-line policy, to investigate the difference between an exact and a statistical
knowledge of the future of the trace.
Obtaining a closed form characterization of the optimal replacement policy
over a finite horizon has proved to be rather more difficult than over the
infinite horizon. The problem has been solved for monotone stack-distance
distributions. Separate arguments establish the optimality of the Least
Recently Used (LRU) policy for all nonincreasing distributions and of the Most
Recently Used (MRU) policy for all nondecreasing distributions. Interestingly,
LRU and MRU are special cases of LPR, within the LRUSM model,
for nonincreasing and non decreasing distributions, respectively. The results
have been obtained by introducing a significant variant of the standard Bellman’s
equation, potentially useful for other control problems.
In Chapters 6–7 it is studied the problem of processors allocation for the
Galois System, a tool for automatically parallelizing, by means of speculative
execution, algorithms that present data amorphous parallelism. The Galois
System is modeled using graph-theoretic concepts and the optimization goal
is identified in trying to maximize the parallelism, while keeping a constant
conflict ratio. This is linked to a function for which we analytically derive
some properties that are then used to design an algorithm that controls the
number of processors in a quick and stable way. For this purpose an extension
to the well known Turán’s theorem is developed.La complessità dei sistemi informatici sta crescendo rapidamente, rendendo
l’uso efficiente delle loro risorse un compito sempre piú proibitivo. In molti
casi la gestione ottimizzata di questi sistemi è stata sviluppata con tecniche
ad-hoc ed euristiche.
In questa tesi viene esplorato un approccio piú generale e flessibile alla
gestione delle risorse, fondato sul potente quadro teorico del controllo ottimo
stocastico. Questo approccio richiede un’attenta modellizzazione del sistema
di interesse come sistema dinamico, con appropriate funzioni di costo e descrizioni
stocastiche degli ingressi imposti dall’ambiente esterno. A questo punto
la ricerca di una politica di gestione ottima che minimizzi il costo aspettato è
matematicamente ben posta e può essere risolta sistematicamente. Vengono
sviluppati due casi di studio, come riassunto di seguito.
Nei Capitoli 1–5 il classico problema di sostituzione (replacement) per le
gerarchie di memoria viene formulato nel quadro della teoria del controllo
ottimo. Si assume che i riferimenti di memoria rispettino il Least Recently
Used Stack Model (LRUSM) e vengono considerate distribuzioni arbitrarie
sullo stack.
Una politica ottima per minimizzare il tasso di miss (accessi fuori dal
buffer) per una traccia di lunghezza infinita viene derivata e chiamata K-L,
dove K(C) ed L(C) sono parametri il cui valore è un funzione, determinata
dalla distribuzione di accessi allo stack, della taglia C del buffer. In seguito
viene introdotto il concetto di politica a tasso di profitto minimo (Least Profit
Rate – LPR) e si dimostra che, nell’LRUSM, LPR è una politica ottima
su orizzonte infinito che, allo stato stazionario, coincide con la politica K-L.
LPR soddisfa la proprietà di inclusione: i contenuti di buffer di dimensione
minore sono inclusi in quelli maggiori. Questa proprietà consente di calcolare
efficientemente il numero di miss per una data traccia in contemporanea per
tutte le taglie del buffer. Inoltre le proprietà della LPR vengono sfruttate
per derivare un efficiente algoritmo di partizionamento per buffer condivisi
concorrentemente fra piú processi. Infine il tasso di miss di LPR viene confrontato
con quello di OPT, la nota politica ottima off-line, per indagare la
differenza fra una conoscenza esatta e una statistica del futuro della traccia.
Ottenere una forma chiusa per la politica di sostituzione su orizzonte
finito si è dimostrato un problema ben piú difficile che su orizzonte infinito,
ed è stato risolto nel caso di distribuzioni di accesso monotone. Argomenti
diversi dimostrano rispettivamente l’ottimalità della politica Least Recently
Used (LRU) per distribuzioni non crescenti e quella di Most Recently Used
(MRU) per distribuzioni non decrescenti. I risultati sono stati ottenuti grazie
all’introduzione di una significativa variante dell’usuale equazione di Bellman,
potenzialmente utile in altri problemi di controllo.
Nei Capitolo 6–7 viene studiato il problema dell’allocazione dei processori
nel sistema Galois, uno strumento per la parallelizzazione automatica,
per mezzo di esecuzione ottimistica (speculative execution), di algoritmi che
presentino un cosiddetto parallelismo amorfo sui dati (data amorphous parallelism).
Il sistema Galois viene modellizzato tramite concetti di teoria dei
grafi e l’obiettivo dell’ottimizzazione è identificato nella massimizzazione del
parallelismo col vincolo di mantenere basso il tasso di conflitti. Questo viene
collegato ad una funzione, per cui vengono analiticamente derivate alcune
proprietà che sono poi usate nella progettazione di un algoritmo capace di
controllare il numero di processori in maniera stabile e veloce. A tal fine
viene sviluppata un’estensione del noto teorema di Turán
Stent Oversizing or not-oversizing? An unresolved dilemma.
No full textPrimary PCI plus stent implantation is indicated for treatment of acute coronary syndrome.
Technically, due to the difference in lumen diameter between vessel portions and stent size in some
coronary arteries, post-expansion of the stent is usually necessary to match proximal reference
diameter and optimize stent apposition. Incomplete stent apposition has been associated with
increased risk of in stent restenosis (ISR) and stent thrombosis (1) despite new stent technologies
have largely reduced this risk (2-3). Although new antiplatelet drugs showed a more rapid onset of
action and a stronger effect (4-5), stent apposition remains an important risk factor for stent
thrombosis. As stent post-dilatation is commonly performed, typically with large over-expansion in
the setting of long tapering vessel segment. There is an increasing recognition that information with
regard to the different stent model designs can have a critical impact on result
Ventriculo-atrial gradient due to first degree atrio-ventricular block: a case report
No full textAbstract Background Isolated, asymptomatic first degree AV block with narrow QRS has not prognostic significance and is not usually treated with pacemaker implantation. In some cases, yet, loss of AV synchrony because of a marked prolongation of the PR interval may cause important hemodynamic alterations, with subsequent symptoms of heart failure. Indeed, AV synchrony is crucial when atrial systole, the "atrial kick", contributes in a major way to left ventricular filling, as in case of reduced left ventricular compliance because of aging or concomitant structural heart disease. Case presentation We performed a trans-septal left atrium catheterization aimed at evaluating the entity of a mitral valve stenosis in a 72-year-old woman with a marked first-degree AV block, a known moderate aortic stenosis and NYHA class III symptoms of functional deterioration. We occurred in a deep alteration in cardiac hemodynamics consisting in an end-diastolic ventriculo-atrial gradient without any evidence of mitral stenosis. The patient had a substantial improvement in echocardiographic parameters and in her symptoms of heart failure after permanent pacemaker implantation with physiological AV delay. Conclusion We conclude that if a marked first degree AV block is associated to instrumental signs or symptoms of heart failure, the restoration of an optimal AV synchrony, achieved with dual-chamber pacing, may represent a reasonable therapeutic option leading to a consequent clinical improvement.</p
Prevention of restenosis after stenting: the emerging role of inflammation
No full textRestenosis after stent implantation is mainly caused by neointimal proliferation through the stent struts. Experimental studies indicate a marked activation of inflammatory cells at the site of stent struts, which are likely to play a key role in the process of neointimal proliferation and restenosis. Coronary stenting is a strong inflammatory stimulus and the acute systemic inflammatory response to local inflammation produced by coronary stenting is highly individual and predicts restenosis and event-free survival. The benefit of anti-inflammatory therapy during the periprocedural period and long-term follow-up is dependent on the inflammatory status. Measurement of cytokine and acute phase proteins, such as C-reactive protein, may be important to identify high-risk subjects and develop specific treatment tailored to the individual patient. (C) 2004 Lippincott Williams Wilkins
Coronary stenting and inflammation
No full textRestenosis after stent implantation is mainly caused by neointimal proliferation through the stent struts. Experimental studies performed in the last decade indicate, that inflammatory mechanisms play a key role in the process of neointimal proliferation and restenosis. Coronary stenting is a strong inflammatory stimulus, and the acute local and systemic inflammatory responses to local inflammation produced by coronary stenting are highly individual and predictive of restenosis and event-free survival. The benefit of anti-inflammatory periprocedural therapy, such as with 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors (statins) and steroids, and long-term follow-up is dependent on the individual's inflammatory status. Measurement of acute-phase reactants, such as C-reactive protein plasma concentration, appears to be important for the identification of subjects at high risk and the development of specific treatment tailored to individual patients. (c) 2005 Elsevier Inc. All rights reserved
Neoangiogenesis connecting a left internal mammary artery proximal stump to an ischemic area of the heart after bypass occlusion
No full textInternal mammary arteries (IMAs), via angiogenesis/arteriogenesis, have great potential for developing collaterals. This generally occurs when ischemia involves the inferior limbs or heart. Although this phenomenon may be spontaneous, it seems to be promoted by iatrogenic or voluntary occlusion of the IMAs. We present a patient who underwent triple coronary artery bypass grafting with two saphenous vein grafts and a free-graft left IMA and suffered an acute myocardial infarction 6 weeks after surgery. Emergency coronary angiography revealed occlusion of the grafts and an amazing branch arising from the proximal stump of the surgically-cut and clipped IMA, leading to the anterior wall of the heart, functionally creating an effective natural bypass, probably stimulated both by anterior wall ischemia and interrupted flow within the proximal left IMA
Restauro e problemi di conservazione: la valle del Belice in Sicilia
No full textIn the 1980s, the Laboratories of Belice were a place of experimentation and validation of the compositional proposals of Italian architecture. Since then, the issues related to the approach followed have been very much discussed, especially those related to urban planning strategies established for the new town of Gibellina. However, the aspects related to the conservation and restoration project, on which this experience should have been founded, have not been treated yet, with the same care. Actually, followed the 1968 earthquake, the architects intervened in many ways on the surviving built heritage, stressing on formal and symbolical aspects, often forcing principles already consolidated in the Restoration Charters, or perhaps paying little attention to the architectural durability over time. For their different approach to memory and for the results in terms of ageing and material decay, the experiences carried out in Gibellina by Alberto Burri with the edification of the Cretto, by Ludovico Quaroni with the construction of the Chiesa Madre, by Francesco Venezia, Franco Purini and Laura Thermes, etc. need to be analyzed. In this sense, this paper aims to discuss the choices made in such examples of monumentality - almost contemporary archaeological sites - and their consequences in terms of functionality, structural behavior and maintenance needs, comparing them with other examples in which the conservation of the ruin coincides with the wish to preserve the memory of a natural (or intentional) destructive act
- …
