Datasets:

SouthernCrossAI
/

Arxiv

Dataset card Data Studio Files Files and versions Community

Split (1)

train · 132k rows

Filename stringlengths 22 64	Paragraph stringlengths 8 5.57k
Processed_A_Faster_Tableau_for_CTL*.txt	is a sequence x of hues such that \|x \| = \|s \|, for each j < \|x \|, x j+1.
Processed_A_Faster_Tableau_for_CTL*.txt	Using the rules described above, using any applicable one at any stage, allows construction of tableaux. We know that the LG rule ensures that any successful ones which we build thus will guar- antee that f is satisﬁable. In the next section we consider whether we can build a successful tableau for any satisﬁable formula in the way.
Processed_A_Faster_Tableau_for_CTL*.txt	Thus, in a Hintikka tableau, the labels tell us exactly which formulas hold there. The completeness result in [17] shows the following, in terms of the concepts deﬁned in this paper: Lemma. If f ∈ L is satisﬁable then it has a Standard Hintikka Tableau. The proof of this lemma is a straightforward translation of the deﬁnitions from [17] but we need to specify how to deﬁne our current predecessor relation p and we also need to check that the tableau is ﬁnished.
Processed_A_Faster_Tableau_for_CTL*.txt	is not made explicit in the tableau structures of the earlier paper. Instead we require that the colour of a node t is related by a successor relation RX between colours to the colour of any successor t′. This means that for any hue in the colour of t′ there is a hue h in the colour of t such that h and h′ are related by a successor relation between hues. We can use such a hue h as the predecessor of h′ and so deﬁne p .
Processed_A_Faster_Tableau_for_CTL*.txt	To show that the tableau (T, s, h , p ) is ﬁnished, we just need to check all the rules of our tableau construction and make sure none require the tableau to be changed in any way. This needs to be done each rule at a time, and needs to be done carefully, although it is straightforward.
Processed_A_Faster_Tableau_for_CTL*.txt	The proof in [17] uses a ﬁnite model theorem for CTL* to obtain a branch boundedness result on the Hintikka tableau. We can guarantee existence of a such a tableau with a certain function of the length of the formula bounding the length of each branch (before an up-link). The bound is triple exponential in the length of the formula, so rather large.
Processed_A_Faster_Tableau_for_CTL*.txt	in a step by step manner only using the construction rules from section 7.1. Thus we make a sequence T 0, T 1, ... of tableaux each one using a construction step to get to the next.
Processed_A_Faster_Tableau_for_CTL*.txt	that T i = (T, s, h , p ) and T −¥ = (T ′, s′, h ′, p that wi(t, j) = (u, k) for u ∈ T ′. Then k < \|h phue of t with the kth phue of u.
Processed_A_Faster_Tableau_for_CTL*.txt	We also show that the constructed tableau does not fail at any stage if one of the check rules such as LG, NTP or the existence of direct contradictions in phues. This follows from the fact that the phues in its labels are subsets of the hues in the labels of the Hintikka tableau.
Processed_A_Faster_Tableau_for_CTL*.txt	If T is ﬁnished (leaﬂess), supported and all predecessors exist then we are done. If T is not supported then choose any formula a in any phue in the label of any node that is not supported. Depending on the form of a we apply one of the tableau rules to add some successor, or some phue and/or some formula(s) in a phue that will ensure that a is then supported. See [15] for details.
Processed_A_Faster_Tableau_for_CTL*.txt	There are only a ﬁnite number of formulas that can be added in hues in labels in a ﬁnite structure . This guarantees that the process will eventually terminate.
Processed_A_Faster_Tableau_for_CTL*.txt	In this deﬁnition of tableau we have guaranteed termination of any tableau construction algorithm by putting a simple but excessive bound on the length of branches. This allows us to conclude failure in a ﬁnite time and to also abbreviate the search for successful tableaux.
Processed_A_Faster_Tableau_for_CTL*.txt	In this paper we have only brieﬂy mentioned the limit on the length of branches as a way of guaranteeing that there are only ﬁnitely many tableau, and so that a search will terminate one way or another. The limit, based on a theoretical upper bound on the minimal CTL* model size, is very generous and hence this is an inefﬁcient way of cutting short tableau searches. Being so generous slows down both negative and positive satisﬁability reports.
Processed_A_Faster_Tableau_for_CTL*.txt	In order to make some sort of working implementation to demonstrate the practicality of this tableau it is necessary to have a better way of preventing the construction of wastefully long branches. For want of better terminology we will call such a facility, a “repetition checker”.
Processed_A_Faster_Tableau_for_CTL*.txt	The task of making a quick and more generally usable repetition checker will be left to be advanced and presented at a later date. In fact, eventually we hope to provide a useful set of criteria for earlier termination of construction of branches depending on the properties of the sequence of colours so far. A simple example of the sort of criterion is the repeated appearance of the same sequence of colours and hues along a non-branching path without being able to construct any up-links. Other more sophisticated ideas are easily suggested but we want to develop a more systematic set of tests before presenting this in future work.
Processed_A_Faster_Tableau_for_CTL*.txt	In [17], we present some basic repetition checking tests for the Hintikka style tableau. These can be used in order to allow some faster automated tableau construction. The tests can be modiﬁed to work with our sparser labels, and we will present full details in a future paper. There are many opportunities for more thorough repetition checks as well.
Processed_A_Faster_Tableau_for_CTL*.txt	Say that \|f \| = l. Thus f has ≤ l subformulas and cl f contains at most 2l formulas. Since each hue contains, for each a ≤ f at most one of a or ¬a , there are at most ≤ 2l hues. Thus there are less than 22l colours. It is straightforward to see that there is a triple exponential upper bound if the tableau search algorithm uses the double exponential bound on branch length [17] to curtail searches down long branches.
Processed_A_Faster_Tableau_for_CTL*.txt	A prototype implementation written by the author shows that for many interesting, albeit relatively small, formulas, the experimental performance of the system is relatively impressive. There are some preliminary results detailed in [15] which show a comparison of running times with the older Hintikka- style tableau technique of [17] and the state of the art game-based CTL* reasoner from [7]. In general the new reasoner is more than an order of magnitude quicker at deciding formulas from a range of basic and distinctive CTL* validities and their negations and a few other satisﬁable formulas. The implementation is available as Java code for public download [15]. Online reasoner coming soon.
Processed_A_Faster_Tableau_for_CTL*.txt	The implementation for the new technique that is used in these experiments, uses some basic repeti- tion checking derived from the checks given earlier in the Hintikka-style system [17]. The new, slightly modiﬁed versions of these mechanisms are not described in the current paper. Instead they will be de- scribed in a future paper.
Processed_A_Faster_Tableau_for_CTL*.txt	In [7], four series of formulas are suggested to examine asymptotic behaviour. Timing results for our system on these formulas are presented in Table 6. We compare the performance of our new tableau with the state of the art in game-based techniques for deciding CTL*. This is using published performance of the reasoner from [7] as reported in experiments in [11]. Consider the following series of formulas: i and b i+1 = AFAGb i. In table 6, we compare i+1 = AFGa a 1 = AFGq, b 1 = AFAGq and for each i ≥ 1, a the performance of the Hintikka-style tableau system from [17], the game-based reasoner from [7] and the new tableau system of this paper (using basic repetition checking) on the growing series built from these formulas. Although the running times, are on different computers, and so not directly comparable, we can see the difference in asymptotic performance. Running times greater than an hour or two are curtailed. From the results we see that we have achieved very noticeable and signiﬁcant improvements in performance on the satisﬁable examples.
Processed_A_Faster_Tableau_for_CTL*.txt	Pure tableau-style reasoning on unsatisﬁable formulas often involves exhaustive searches and the new technique is not immune to such problems. See the 400 series of examples in the asymptotic experiments. We will say more about these examples when proposing some new repetition mechanisms in the future. There are some, more theoretical descriptions of other game-based and automata-based techniques for model-checking CTL* in older papers such as [10], [2] and [9]. However, these do not seem di- rectly applicable to satisﬁability decisions and/or there do not seem to be any easily publicly available implemented tools based on these approaches.
Processed_A_Faster_Tableau_for_CTL*.txt	In this paper we have presented, albeit in a fairly high level sketch, a traditional tableau approach to reasoning with the important logic CTL*. Soundness and completeness results are proved and prototype implementation demonstrates the signiﬁcantly improved performance of the new approach on a range of test formulas.
Processed_A_Faster_Tableau_for_CTL*.txt	The next task in this direction is to build on the foundation here and present full details and proofs of the repetition checking mechanisms that can be used with the tableau construction. There are some basic repetition mechanisms available in the previous, Hintikka style tableau [17] but they need to be modiﬁed slightly. There are opportunities for additional techniques. It is also important to improve and document the rule-choice algorithms which have a bearing on running times.
Processed_A_Faster_Tableau_for_CTL*.txt	In the future, it will be useful to develop reasoning tools which combine the latest in tableaux, au- tomata and game-based approaches to CTL. Having tools working in parallel should allow faster de- cisions. It will also be useful to extend the work to logics of multi-agent systems such as ATL and strategy logic [12].
Processed_Solar_Line_Asymmetries:_Modelling_the_Effect_of_Gr.txt	In this way, the solar spectrum has become a valuable tool for probing the sun.
Processed_Solar_Line_Asymmetries:_Modelling_the_Effect_of_Gr.txt	are looked at. Asymmetries in the spectra of other stars are compared to solar lines.
Processed_Solar_Line_Asymmetries:_Modelling_the_Effect_of_Gr.txt	The asymmetry of a spectral line is often described by the bisector of the line.
Processed_Solar_Line_Asymmetries:_Modelling_the_Effect_of_Gr.txt	The bisector is the line half-way between points of equal intensity on the line profile.
Processed_Solar_Line_Asymmetries:_Modelling_the_Effect_of_Gr.txt	the strong line is broader, this is to be expected if the asymmetry is fairly uniform.
Processed_Solar_Line_Asymmetries:_Modelling_the_Effect_of_Gr.txt	the effects of velocities can be seen more readily, shows more asymmetry in the wings.
Processed_Solar_Line_Asymmetries:_Modelling_the_Effect_of_Gr.txt	The difference in the line cores, where the effect of damping is small, is much less.
Processed_Solar_Line_Asymmetries:_Modelling_the_Effect_of_Gr.txt	The temperature of a system in LTE can be defined as the temperature of the particles.
Processed_Solar_Line_Asymmetries:_Modelling_the_Effect_of_Gr.txt	if we consider the energy of the ground state of the lower ionisation state to be zero.
Processed_Solar_Line_Asymmetries:_Modelling_the_Effect_of_Gr.txt	to the formation of the solar spectrum, both for the continuum and for spectral lines.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	ABSTRACT A number of attempts have been made at estimating the amount of box ofﬁce revenue a ﬁlm will generate during its opening weekend. One such attempt makes extensive use of the number of views a ﬁlm’s Wikipedia page has attracted as a predictor of box ofﬁce success in the United States. In this paper we develop a similar method of approximating box of- ﬁce success. We test our method using 325 ﬁlms from the United States and then apply it to ﬁlms from four foreign markets: Japan (95 ﬁlms), Australia (118 ﬁlms), Germany (105 ﬁlms), and the United Kingdom (141 ﬁlms). We ﬁnd the technique to have inconsistent performance in these nations. While it makes relatively accurate predictions for the United States and Australia, its predictions in the remaining markets are not accurate enough to be useful.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	well they would do in theaters [1]. More recently, Mesty´an et al. successfully used data gathered from Wikipedia2, includ- ing page views, total edits, and number of editors, to predict the amount of money a ﬁlm (released in the United States) would bring in during its opening weekend [5].
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	It is a rather unexpected result that the number of views a ﬁlm’s Wikipedia page has is so highly correlated with the money earned by that ﬁlm during its opening weekend. A possible explanation for this phenomenon is as follows. In modern society potential moviegoers who are unsure which movie to see often turn to the internet to aid them in making their decision. It was recently suggested by Panaligan et al. that opening weekend box ofﬁce sales for a ﬁlm could be well approximated by the search volume for terms related to the movie. They were able to account for as much as 94% of the variation in opening weekend box ofﬁce performance using a combination of related search term volume, franchise status, and seasonality [6]. If we take this result into account along with the observation that upon searching for a ﬁlm (using Google or Bing), that ﬁlm’s corresponding Wikipedia page is among the top search results, it is reasonable to hypothesize that the Wikipedia page acts as a proxy for searches relating to the ﬁlm.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	INTRODUCTION The motion picture industry is a multi-billion dollar market. Therefore many attempts have been made at predicting how much money ﬁlms will bring in during their respective open- ing weekends in theaters [1, 5, 8]. There is also a body of research on the effects of various factors on box ofﬁce rev- enue which assists those attempting to predict it [3, 4, 9, 10].
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for proﬁt or commercial advantage and that copies bear this notice and the full citation on the ﬁrst page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior speciﬁc permission and/or a fee. WebSci’14, June 24 – June 26, 2014, Bloomington, Indiana. ACM 978-1-4503-1889-1.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	Our goals in this article are to recreate the results of Mesty´an et al. using a more recent set of ﬁlms released in the United States and then to test whether or not the technique achieves comparable results when applied to ﬁlms released in the United Kingdom, Australia, Japan, and Germany. First we discuss the methods we used to complete our study and then we present the results accompanied by a brief analysis.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	Once we had the URLs of the Wikipedia pages corresponding to the ﬁlms we were interested in, we used a website which records Wikipedia article trafﬁc statistics4 to download the daily view counts of each page (see the Regression Model Section). We then used a simple multivariate linear regression to attempt to predict each ﬁlm’s opening weekend box ofﬁce revenue based on its Wikipedia page views and screen count. Finally, we analyzed the accuracy of the regression model using leave-one-out cross-validation (in the Cross-Validation Section).
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	Finding Wikipedia URLs Wikipedia URLs unambiguously identify ﬁlms. Other sen- sors of public sentiment used for box ofﬁce prediction, such as mentions of titles in Twitter text [1] or Google search queries [6], are difﬁcult to work with for titles based on com- mon words (e.g. a tweet mentioning ”Frozen” may not be referencing the ﬁlm). Disambiguation becomes substantially more difﬁcult when long lists of titles in multiple foreign lan- guages are involved.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	Wikipedia articles seldom contain information about open- In order to conduct our research ing weekend box ofﬁce. we therefore need to construct a mapping between the open- ing weekend box ofﬁce data available from Box Ofﬁce Mojo and the associated language-speciﬁc Wikipedia URLs. Since Wikipedia URLs are not necessarily exact title names (e.g. de.wikipedia.org/wiki/42_(Film) ), this mapping must be constructed by hand (which is prohibitively time consuming) or through an automated fashion.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	We check lists built from DBpedia5 against results obtained via the Google Custom Search API6 to resolve spelling dif- ferences between titles on Box Ofﬁce Mojo and Wikipedia URLs. For each of the nations considered, we restrict the search API to to return links only from the national language’s Wikipedia (i.e. en.wikipedia.org, ja.wikipedia.org, de. wikipedia.org) and query the search API for Box Ofﬁce Mojo title names followed by the appropriate translation of the word “ﬁlm” (i.e. “ﬁlm” (en), “ﬁlm” (de), or “映画” (ja) ).
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	Despite the addition of “ﬁlm” to our searches, the highest ranked search results are not necessarily the Wikipedia pages for the queried ﬁlms.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	To ensure that the links retrieved from our searches indeed refer to the Wikipedia URLs for each ﬁlm, we use language- speciﬁc variants of DBpedia [2] to build lists of URLs which are categorized as ﬁlms and discard search results which do not appear in the lists.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	These queries provide us with lists of Wikipedia URLs for: 769 Japanese ﬁlms, 1026 German ﬁlms, and 3079 English ﬁlms.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	Opening weekend box ofﬁce data on Box Ofﬁce Mojo is available for: 104 Japanese ﬁlms, 166 German ﬁlms, 219 Australian ﬁlms and 225 UK ﬁlms.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	The technique provided here allows us to align Wikipedia URLs with box ofﬁce data for: 73 Japanese ﬁlms, 132 Ger- man ﬁlms, 118 Australian ﬁlms, and 141 UK ﬁlms. Some alignments were also performed manually.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	where yi, xi,1, xi,2, and εi are the opening weekend box of- ﬁce revenue, screen count, number of Wikipedia page views8, and the error in the prediction, for ﬁlm i, respectively. α1 and α2 are the regression coefﬁcients. Notice that xi,2 is a func- tion of time in the above equation. Since we have a range of dates for which we know how many views each ﬁlm’s Wikipedia page received, we have the ability to form a large number of regression models, one for each day leading up to the ﬁlm’s release. For example, we may wish to predict a movie’s box ofﬁce success using only data available one month before it premiers. Collecting the Wikipedia page hits over a long period of time allows us to discern how the ac- curacy of the regression model changes over time, as more information is introduced. One quantity we will use to mea- sure accuracy is the R2 Coefﬁcient of Determination. We also found it prudent to use a second tool to evaluate our method’s performance.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	where yi is the actual amount ﬁlm i earned during its open- ing weekend, and pi is the prediction generated by the mul- tivariate regression11. A signiﬁcant relative error indicates a discrepancy between a movie’s predicted box ofﬁce revenue and its actual revenue which is large relative to the movie’s actual box ofﬁce revenue.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	United States Box Ofﬁce As was noted in the Methods Section, once we had each US ﬁlm’s associated Wikipedia page view data along with its screen count, we formed a separate linear regression model on each day leading up to its premier. Thus we were able to visualize the accuracy of our predictions as a function of time as in Figure 1.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	As one would expect, the accuracy of the model improves as the movie premiers draw near and movie awareness swells. This is the time period when potential moviegoers are pre- sumably researching which movie to see during the coming weekend. Using this data set we obtain accuracy comparable to that reached by the model based only on Wikipedia page views used by Mesty´an et al.12[5]. This is the level of accu- racy we expected to obtain since the only major difference between the two models is that they were used on different lists of ﬁlms.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	we divide the difference between the actual box ofﬁce rev- enue and the prediction by the actual box ofﬁce revenue. For example, over predicting by one million dollars the box ofﬁce revenue for a ﬁlm which brought in ten million dollars dur- ing its opening weekend would result in a relative error of 0.1 while over predicting by the same amount the box ofﬁce rev- enue for a ﬁlm which only brought in one hundred thousand dollars would yield a much greater relative error of 9.0. A large relative error for a ﬁlm that did well during its opening weekend indicates a considerable gap between our prediction and its actual box ofﬁce revenue. We see in Figure 2 that the relative errors are reasonably small, as we would expect con- sidering the R2 value achieved predicting United States box ofﬁce results.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	Foreign Box Ofﬁce After recreating one of the results of Mesty´an et al. we ap- plied the same techniques to sets of ﬁlms released in the United Kingdom, Australia, Germany, and Japan.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	three countries speak a common language. As a result in- ternet users in all three nations use the same English branch of Wikipedia. This makes it impossible to discern which of the three countries a page view on a movie’s Wikipedia page came from. Combined with the different release dates for ﬁlms across the three markets, a large amount of noise is likely being introduced. For instance, if ﬁlm “A” premiers in the United States before Australia, the United States premier could create an inﬂux of Wikipedia page views earlier than normal, relative to Australia’s premier. This, in turn, could cause the model to overestimate the ﬁlm’s expected box ofﬁce revenue in the Australian market. Indeed there are over 90 ﬁlms in both the lists from Australia and the United Kingdom which have different release dates from their counterparts in the United States.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	Regressing on the number of Wikipedia page views and screen counts for each of the 141 UK ﬁlms at each day leading up to the ﬁlms’ premiers, we produce a plot of the accuracy of our model as a function of time (Figure 3). The maximum value the R2 coefﬁcient of determination attains here is about 0.34; much less than that for the United States predictions. The shape of the graph is also strikingly different from Fig- ure 1. The R2 coefﬁcient is at its maximum 50 days before the ﬁlm premiers and decreases as time progresses, eventually in- creasing marginally in the few days before the premiers. This strange behavior may be the result of differing movie release dates, as discussed above.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	The relative errors for the ﬁlms from the United Kingdom are given in Figure 4. There are a few outliers with very large relative errors, but the errors are larger on average, than for the US ﬁlms. This is intuitive given that the coefﬁcient of determination of the linear regression is much lower for the ﬁlms from the United Kingdom.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	to that in Figure 1; it increases as time does. The model at- tains a relatively high R2 value of 0.57 a day before the movie release dates. Figure 6 shows the relative error for 50 Aus- tralian ﬁlms. There are fewer ﬁlms with huge relative errors than we saw for the UK data set and the model produces lower relative errors, on average, for the Australian data set than the UK one.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	Japanese and German Speaking Box Ofﬁces Since Wikipedia has separate versions written entirely in Japanese and German, it is a reasonable assumption that each is accessed by primarily Japanese and German citi- zens, respectively. This resolves the stumbling block we faced with the English-speaking nations, namely determining which market each viewer belonged to. Hence one might ex- pect that predictions for Japanese and German opening week- end box ofﬁce revenue using Wikipedia page views would outperform those for the UK and Australia.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	compared to the almost 0.61 R2 coefﬁcient associated with the United States ﬁlms. Figure 8 shows the relative error for the German ﬁlms.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	The results are even worse when the model is used to predict opening weekend box ofﬁce revenues for 95 ﬁlms released in Japan. Figure 9 shows the evolution of the coefﬁcient of determination for this multivariate regression. The same pattern present in the United States’, Australian, and German R2 evolution plots does not appear here. The overall accuracy of the regression model is also markedly lower when it is applied to the Japanese market than the German market (maximum R2 ≈ 0.30). In Figure 10 we have the relative error for the Japanese ﬁlms. Using the regression model often predicts double the amount ﬁlms actually earn dur- ing their opening weekends. This is unsurprising, given the low coefﬁcient of determination attained with this set of ﬁlms.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	as often as people in the United States and Australia. Or, per- haps they use different means than the internet to aid their decisions altogether, or do not perform any research at all before going to the theaters. Another possible explanation for the disparity in accuracy between the predictions for the ﬁlms released in the United States and those released in the UK, Germany, and Japan is the number of movies consid- ered. Since there were two to three times as many ﬁlms in the United States data set as the other sets, it included numerous low-budget, lesser-known movies. It is possible that the box ofﬁce success of lesser-known movies is easier to predict. In- deed, if we remove the top 100 highest grossing ﬁlms from the list of US ﬁlms, our method obtains a maximum R2 of almost 0.72. However, this does not explain why we are able to predict box ofﬁce revenue in Australia with more certainty than in other foreign countries. In any case, it is clear that a technique to predict opening weekend box ofﬁce proﬁts in the United Kingdom, Germany, or Japan must rely on more than Wikipedia page views alone.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	ofﬁce revenues in the United States and Australia with rea- sonable accuracy, its performance drops signiﬁcantly when applied to various foreign markets. In particular, we con- structed a multivariate linear regression using the number of views the Wikipedia pages of various ﬁlms released in the United States received to predict their successes during their opening weekends in theaters. We automated the process of determining a ﬁlm’s corresponding Wikipedia page so that we could apply the same technique to foreign movies. However, when we used the model to predict the opening weekend box ofﬁce revenues generated by ﬁlms in British, Japanese, and German theaters, we found its accuracy to be far from satis- factory. Finally we gave brief discussions of possible causes of the discrepancies.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	While Wikipedia page views may be a strong predictor of box ofﬁce performance for ﬁlms in the United States, the same cannot necessarily be said for ﬁlms released in other nations. Before a model similar to that presented here is used to pre- dict box ofﬁce sales in a foreign market, it should be tested on backdata gathered from that market.
Processed_Prediction_of_Foreign_Box_Office_Revenues_Based_on.txt	ACKNOWLEDGMENTS The authors would like to thank Dean Shaw for his help in- terpreting Japanese ﬁlm titles and URLs.
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	Tharaka A. Lamahewa, Rodney A. Kennedy, Thushara D. Abhayapala and Terence Betlehem Research School of Information Sciences and Engineering, The Australian National University, ACT 0200, Australia.
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	Abstract— This paper presents an analytical model for the fading channel correlation in general scattering environments. In contrast to the existing correlation models, our new approach treats the scattering environment as non-separable and it is modeled using a bi-angular power distribution. The bi-angular power distribution is parameterized by the mean departure and arrival angles, angular spreads of the univariate angular power distributions at the transmitter and receiver apertures, and a third parameter, the covariance between transmit and receive angles which captures the statistical interdependency between angular power distributions at the transmitter and receiver apertures. When this third parameter is zero, this new model reduces to the well known “Kronecker” model. Using the proposed model, we show that Kronecker model is a good approximation to the actual channel when the scattering channel consists of a single scattering cluster. In the presence of multiple remote scattering clusters we show that Kronecker model over estimates the performance by artiﬁcially increasing the number of multipaths in the channel.
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	Wireless channel modelling has received much attention in recent years since space-time processing using multiple antennas is becoming one of the most promising areas for improvements in performance of mobile communication sys- tems [1, 2]. In channel modelling research, the effects of fading channel correlation due to insufﬁcient antenna spacing and sparse scattering environments are of primary concern as they impact the performance of multiple-input multiple-output (MIMO) communication systems.
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	which gives channel correlation for a general class of scat- tering environments. In our proposed model, fading channel correlation is parameterized by the antenna conﬁguration de- tails (spacing and the placement) both at the transmitter and the receiver arrays, and the joint bi-angular power distribution between transmitters and receivers, which models the scatter- ing environment surrounding the transmit and receive antenna apertures. The bi-angular power distribution is parameterized by the mean departure and arrival angles, angular spreads of the univariate angular power distributions at the transmitter and receiver apertures, and a third parameter, the covariance between transmit and receive angles which captures the sta- tistical interdependency between angular power distributions at the transmitter and receiver apertures. When this third parameter is zero, i.e., the power distribution at the transmitter is independent of the power distribution at the receiver, the proposed correlation model reduces to the Kronecker model. In order to model the scattering environment we propose several bi-angular power distributions and also ﬁnd the correlation coefﬁcients associated with these distributions in closed form. Using the proposed model, we show that Kronecker model is a good approximation to the actual channel when the scattering channel consists of a single scattering cluster. We also show that when the scattering channel consists of multiple remote scattering clusters, the Kronecker model over estimates the performance MIMO systems by artiﬁcially increasing the number of scattering clusters in the scattering channel.
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	A popular channel model that has been used in MIMO performance analysis is the “Kronecker” model [3–5]. In this model, the correlation properties of the MIMO channel are modeled at the transmitter and receiver separately, neglecting the statistical interdependency between scattering distributions at the transmitter and receiver antenna apertures. Measurement and analytical results presented in [6, 7] suggest the Kronecker model does not accurately model the underlying scattering channel, therefore it does not provide accurate performance results.
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	Notations: Throughout the paper, the following notations (upper) letters denote vectors will be used: Bold lower ]† denote the transpose, complex ]∗ and [ ]T , [ (matrices). [ · · · conjugate and conjugate transpose operations, respectively. The symbols δ( denote the Dirac delta function and ) and · Matrix Kronecker product, respectively. The notation E {·} denotes the mathematical expectation,vec(A) denotes the vectorization operator which stacks the columns of A, . ⌉ ⌈ denotes the ceiling operator and S1 denotes the unit circle.
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	R.A. Kennedy and T.D. Abhayapala are also with National ICT Australia, Locked Bag 8001, Canberra, ACT 2601, Australia. National ICT Australia is funded through the Australian Government’s Backing Australia’s Ability initiative, in part through the Australian Research Council.
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	positions yr, r = 1, 2, , nR relative to the receiver array yr k denote and rR ≥ origin. rT ≥ max the transmitter and the radius of spheres that contain all receiver antennas, respectively. We assume that scatterers are distributed in the far ﬁeld from the transmitter and receiver antennas and regions containing the transmit and receive antennas are distinct.
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	Jn(x) is the spatial-to-mode function (SMF) which where maps the antenna location to the n-th mode of the region. The form which the SMF takes is related to the shape of the scatterer-free antenna region. For a circular region in 2- dimensional space, the SMF is given by a Bessel function of the ﬁrst kind [8] and for a spherical region in 3-dimensional space, the SMF is given by a spherical Bessel function [9]. For a prism-shaped region, the SMF is given by a prolate spheroidal function [10].
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	complex scattering gain between the representing the 1)-th mode of the scatter-free transmit region (m MT − − and (ℓ 1)-th mode of the scatter-free receiver region, − where g(φ, ϕ) is the effective random complex scattering gain function for signals with angle-of-departure φ from the scatter-free transmitter region and angle-of-arrival ϕ at the scatter-free receiver region. The reader is referred to [8] for more information about the channel decomposition (1).
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	RS =          is the correlation between m-th and m′-th where RS,m,m′ columns of H S. A diagonal block RS,m,m gives the modal correlation matrix at the receiver region due to the m-th mode at the transmit region whereas off diagonal blocks RS,m,m′ give the cross correlation between two distinct modal pairs at the transmit and receiver apertures.
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	In this section, we outline several examples of bi-angular scattering distributions along with their modal correlation coefﬁcients (5), which give the entries of the modal correlation matrix RS.
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	where R1/2 is the positive deﬁnite matrix square root of R and W is a nR × nT matrix which has zero-mean independent and identically distributed complex Gaussian random entries with unit variance.
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	Fig. 1. Average mutual information of 3-transmit UCA and 3-receive UCA MIMO system in separable (Kronecker with ρ = 0) and non-separable (ρ = 0.8) scattering environments: bivariate truncated Gaussian azimuth ﬁeld with mean AOD = 90◦, mean AOA = 90◦, transmitter angular spread σt = 10◦ and receiver angular spreads σr = {30◦, 10◦}.
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	performance for all SNRs. When the angular spread at the receiver is small, e.g. σr = 10◦, we can observe that the Kronecker model gives slightly higher performance than the non-separable model for higher SNRs. However, the margin of capacity over estimation is insigniﬁcant in comparison with the i.i.d. channel capacity performance. Therefore, Kronecker model provides a good estimation to the actual scattering channel when the joint scattering distribution is uni-modal. Reasoning for this claim will be discussed in the next section.
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	A multi-modal azimuth power distribution arises when there are two or more strong multipaths exist in a fading channel. This may be the result of multiple remote macro- scopic scattering clusters, for instance. A multi-modal bivariate distribution can be constructed from a mixture of uni-modal bivariate distributions. Fig. 2 shows a multi-modal bivariate Gaussian distributed azimuth ﬁeld with 3 modes centered 40◦, 40◦), (0◦, around (φ0, ϕ0) = , each mode with angular spreads σr = σt = 5◦ and ρ = 0.8. Note that, in this case the effective angular spreads at the receiver and transmitter are larger than 5◦.
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	Fig. 3. Average mutual information of 3-transmit UCA and 3-receive UCA MIMO system for separable and non-separable scattering channel considered in Fig. 2.
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	is given by PRx(ϕ) are G(φ, ϕ) = the transmit and receive power distributions, generated by marginalizing G(φ, ϕ). Fig. 4 shows the Kronecker model e G(φ, ϕ) of the scattering channel considered in Fig. 2. PSD Comparing Fig. 4 with Fig. 2 we can observe that G(φ, ϕ) consist of six extra modes, corresponding to additional six scattering clusters. Therefore, Kronecker model introduces virtual scattering clusters located at the intersection of the actual scattering clusters. As a result, Kronecker model will increase the effective angular spread at the transmit and re- ceive apertures (lower modal correlation) and hence improved system performance. Therefore, the popular Kronecker model does not model the MIMO channel accurately when there exist multiple scattering clusters in the channel. These observations match the measurement results published in [6].
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	Now we consider the uni-modal PSD used in our ﬁrst simulation example. Fig. 5 shows the corresponding Kronecker G(φ, ϕ) for this channel, for σr = 10◦ and Model PSD σt = 10◦. In this case the Kronecker model does not introduce any additional virtual scattering clusters into the channel. As a result, no increase in the number of multipaths of the channel, hence both models give very similar performance.
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	Fig. 5. Kronecker model PSD G(φ, ϕ) = PT x(φ)PRx(ϕ) of the uni-modal non-separable scattering distribution used in the ﬁrst example to obtain the results in Fig. 1 for σr = 10◦.
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	We showed that 2-D Fourier series coefﬁcients of the bi- angular power distribution and transmit and receive antenna sampling points contribute to the entries of the correlation matrix. We proposed several bi-angular power distributions and their 2-D Fourier series coefﬁcients in closed form.
Processed_MIMO_Channel_Correlation_in_General_Scattering_Env.txt	Using the proposed model, we show that Kronecker model is a good approximation to the actual channel when the scattering channel consists of a single scattering cluster. In the presence of multiple remote scattering clusters we show that Kronecker model over estimates the performance of MIMO systems by introducing virtual scattering clusters into the channel. Therefore, in this case, Kronecker model cannot be used to represent the channel.
Processed_Dark-Matter_Content_of_Early-Type_Galaxies_with_Pl.txt	Dark Galaxies and Lost Baryons Proceedings IAU Symposium No. 244, 2007 J. I. Davies & M. J. Disney, eds.
Processed_Dark-Matter_Content_of_Early-Type_Galaxies_with_Pl.txt	Abstract. We examine the dark matter properties of nearby early-type galaxies using plane- tary nebulae (PNe) as mass probes. We have designed a specialised instrument, the Planetary Nebula Spectrograph (PN.S) operating at the William Herschel telescope, with the purpose of measuring PN velocities with best eﬃciency. The primary scientiﬁc objective of this custom- built instrument is the study of the PN kinematics in 12 ordinary round galaxies. Preliminary results showing a dearth of dark matter in ordinary galaxies (Romanowsky et al. 2003) are now conﬁrmed by the ﬁrst complete PN.S datasets. On the other hand early-type galaxies with a “regular” dark matter content are starting to be observed among the brighter PN.S target sample, thus conﬁrming a correlation between the global dark-to-luminous mass virial ratio (fDM = MDM/M⋆) and the galaxy luminosity and mass.
Processed_Dark-Matter_Content_of_Early-Type_Galaxies_with_Pl.txt	The currently accepted cosmological paradigm, the so-called concordance ΛCDM model, i.e. cold dark matter (CDM) plus a cosmological constant (Λ), has been overwhelmingly successful in describing the formation and evolution of structures in the Universe but there remain many observational discrepancies on galaxy and cluster scales that call for a critical veriﬁcation. While the ﬂat rotation curves observed since the 1970s in spiral galaxies have been one of the prime factors in support of the existence of dark mat- ter (Einasto et al.1974; Faber&Gallagher 1979), studies of late-type galaxies suggest, in some cases (e.g. low surface brightness galaxies), soft cores and relative low concentra- tion density proﬁles (Gentile et al. 2004; de Blok 2005) at odd with ΛCDM inner density cusps (Navarro et al. 1997, NFW hereafter; Moore et al. 1999). Given the long-standing dark matter puzzles in disk galaxies, it is obviously important to also examine the mass content of the second major galaxy type, the early-types (ETGs: ellipticals and S0). The mass content of these systems, which are generally free of cold gas and have their stars moving in random directions, is more diﬃcult to measure than in disk galaxies. The main diﬃculty is observational, since the only ubiquitous mass tracers in ETGs, i.e. stars, can be investigated through their integrated light (spectroscopy) which is hard to be mea- 2Reﬀ (Kronawitter et al. 2001, Samurovic sured with suﬃcient signal-to-noise outside ∼ & Danziger 2005). A suitable option is the X-ray emission from hot gas (Paolillo et al.
Processed_Dark-Matter_Content_of_Early-Type_Galaxies_with_Pl.txt	was more consistent with constant mass-to-light ratio systems rather than dark matter dominated systems for which a ﬂat VD proﬁle was expected.
Processed_Dark-Matter_Content_of_Early-Type_Galaxies_with_Pl.txt	This unexpected result has produced diﬀerent interpretations either in the ΛCDM framework (Dekel et al. 2005 = D+05) or in MOND theory (Milgrom & Sanders 2003). In particular, D+05 address very radial stellar orbits and projection eﬀects in order to explain declining VD proﬁles.
Processed_Dark-Matter_Content_of_Early-Type_Galaxies_with_Pl.txt	Napolitano et al. (2005, N+05 hereafter) made predictions of gradients of mass-to- light ratios in ETGs and have ascertained that “quasi-constant M/L” are indeed ex- pected within ΛCDM, although the R+03 sample shows M/L gradients which are too low and conﬂicting with acceptable star formation eﬃciency and baryon fraction. Low M/L gradients mirror a generalised trend of the global dark-to-luminous mass virial ratio (fDM = MDM/M⋆) with stellar mass/luminosity: brightest galaxies have a larger fraction of dark- to-luminous matter, fDM = with respect to fainter galaxies (MB > 20.5), the trend being possibly reversed toward dwarf galaxies. Here we characterise the combined VD proﬁles of the R+03 galaxy sample, i.e. NGC 821, NGC 3379 and NGC 4494, for which, after 4 years, we have now the ﬁnal PN datasets (see Table 1 for the ﬁnal number of PNe observed) and present very preliminary dynamical estimate of their dark matter content as measured through their fDM.
Processed_Dark-Matter_Content_of_Early-Type_Galaxies_with_Pl.txt	In Fig. 1 (left panel), with open symbols, we show the combined VD proﬁle of the three galaxies after having scaled (with respect to Reﬀ ) and normalised (with respect to the central velocity dispersion) the individual galaxy curves. These have been obtained as azimuthally averaged Vrms = √v2 + σ2 within radial bins following the galaxy isophotes distribution for all galaxies. It is easy to recognise the pseudo-Keplerian fall consistent with the one observed in the preliminary datasets as in R+03. In the same ﬁgure we also show a collection of long-slit data from the same galaxies as empty stars which demonstrate that the PN VD proﬁles lie on the extention of the stellar kinematics of the central regions, thus allowing to continuously map the galaxy kinematics out to many eﬀective radii. Beside their declining stellar VDs, these galaxies share other common properties like intermediate luminosities ( L∗, see Table 1) and the fact of being fast rotators (Cappellari et al. 2006), disky/cuspy systems (N+05).
Processed_Dark-Matter_Content_of_Early-Type_Galaxies_with_Pl.txt	In the same panel of Fig. 1, we also show for comparison two galaxies, NGC 4374 and NGC 5846 (full symbols), for which we have the PN.S observational program lately completed and ﬁnal PN catalogues done (see N. PNe in Table 1). Here the velocity dispersion proﬁles have been normalised to the average central VD value of the faint galaxies in order to mark the kinematical diﬀerences between the two samples; long-slit kinematics are also shown as full stars, accordingly. Diﬀerently from the R+03 sample, these galaxies show markedly ﬂat VD proﬁles which well contrasts with the pseudo- Keplerian fall of the former ones, suggesting the presence of a signiﬁcant dark halo. These two “dark-matter dominated” galaxies mark diﬀerences with respect to the R+03 sample for many other reasons: they are brighter (MB < 20.9), slow-rotators (Cappellari et al. 2006) and boxy/cored galaxies (N+05). This dichotomy on many structural galaxy properties is suggestive of some connection with galaxy formation history as discussed elsewhere (Capaccioli et al. 2002, N+05).
Processed_Dark-Matter_Content_of_Early-Type_Galaxies_with_Pl.txt	Figure 1. Left: Combined velocity dispersion proﬁle of 5 galaxies from the PN.S primary target list (see text). The pseudo-Keplerian fall of the faint sample (NGC 821, NGC 3379, NGC 4494, open symbols) is evident against ﬂat proﬁles of the brighter systems (NGC 4374 and NGC 5486, full symbols). Long-slit data from the inner integrated light are shown as star symbols (open and full accordingly). Right: the anisotropy proﬁle corresponding to our best ﬁt to the PN dispersion proﬁle for NGC 4494 (black solid line) is overplotted to Fig. 2 from Mamon and Lokas 2005 where they show with symbols results from diﬀerent N-body simulations and with curves analytical galaxy anisotropy proﬁles (long dashed matches the D+05 merger remnants).
Processed_Dark-Matter_Content_of_Early-Type_Galaxies_with_Pl.txt	As anticipated, NGC 5846 is the only galaxy showing a “regular” dark matter halo, Mvir relation expected having the concentration and virial mass consistent with the cvir from cosmological simulations (Bullock et al. 2001, N+05). The faint “ordinary” sample instead shows concentrations which are systematically lower than predicted from simula- tions. Another remarkable result is that the virial mass-to-luminous matter ratio, fDM is clearly an increasing function of the galaxy stellar mass, which conﬁrms a smooth trend shown in N+05 and discussed in Napolitano et al. (2006). This trend can be interpreted in terms of global star formation eﬃciency, ǫSF = M⋆/Mbar, i.e. the fraction of baryonic mass Mbar cooled in stars, assuming baryon conservation, such that ǫSF = 4.9/fd (see N+05 for further details). More massive systems are globally less eﬃcient in convert- ing gas to stars than galaxies around L∗, like the ones in the “ordinary” PN.S sample, which represent the most eﬃcient ones placing themselves in the maximum (minimum) of the ǫSF (fDM) run with M⋆. Indeed the trend is expected to be reversed moving to- ward the dwarf galaxies. The fact that in our analysis we are considering appropriate orbital anisotropy, consistent with predictions from cosmological simulations (see Fig. 1 right panel and Table 1), allows as to exclude that the trend of the dark matter fraction within the analysed galaxies can be the reﬂection of an anisotropy sequence with the mass.
Processed_Dark-Matter_Content_of_Early-Type_Galaxies_with_Pl.txt	proﬁles out to very large distance from the galaxy centres, while bright, slow-rotating, boxy/cored systems have ﬂat velocity dispersion proﬁles. This corresponds to a diﬀer- ent total dark-to-luminous matter virial ratio, which is larger for brighter systems with respect to fainter ones. The only galaxy for which we have a preliminary Jeans analysis in the PN.S target sample, NGC 5846, shows a DM distribution which fairly follows a “regular” NFW density proﬁle since the cvir and Mvir are consistent with the expectation of N-body simulations (Bullock et al. 2001, N+05). The faint sample, on the contrary, has cvir which are a factor of two lower than predicted from ΛCDM .
Processed_Dark-Matter_Content_of_Early-Type_Galaxies_with_Pl.txt	All these results seem to support the scenario discussed in Napolitano et al. (2005) where we interpreted the dichotomy in the global structure galaxy parameters as a cor- relation between dark matter content in galaxies and the mass. This is the consequence of a broader trend of the global eﬃciency of the star formation with the mass (Benson et al. 2000, Dekel & Birnboim 2006) which suggests a connection between the observed dark matter properties of early-type galaxies and their formation history.

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.