Filename
stringlengths 22
64
| Paragraph
stringlengths 8
5.57k
|
|---|---|
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Another set of rule strings will be generated in the same way, some of which will replace previous strings based on roulette-wheel fitness selection. If stopping conditions are not met, the conditional probabilities for all nodes in the Bayesian network are updated again using the current set of rule strings. The algorithm thereby tries to explicitly identify and mix promising building blocks. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | It should be noted that for most scheduling problems, the structure of the network model is known and all variables are fully observed. In this case, the goal of learning is to find the rule values that maximize the likelihood of the training data. Thus, learning can amount to ‘counting’ in the case of multinomial distributions. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | The rest of this paper is organized as follows. Section 2 gives an overview on the nurse scheduling problem, and the following section 3 introduces the general concepts about graphical models and Bayesian networks. Section 4 discuses the proposed Bayesian optimization algorithm, describing the construction of a Bayesian network, learning based on the Bayesian network, and the four building rules in detail. Computational results using 52 data instances gathered from a UK hospital are presented in section 5. Concluding remarks are in section 6. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | of days worked is not usually the same as the number of nights. Therefore, it becomes important to schedule the ‘correct’ nurses onto days and nights respectively. The latter two characteristics make this problem challenging for any local search algorithm, because finding and maintaining feasible solutions is extremely difficult. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | The numbers of days or nights to be worked by each nurse defines the set of feasible weekly work patterns for that nurse. These will be referred to as shift patterns or shift pattern vectors in the following. For each nurse i and each shift pattern j all the information concerning the desirability of the pattern for this nurse is captured in a single numeric preference cost pij. These costs were determined in close consultation with the hospital and are a weighted sum of the following factors: basic shift- pattern cost, general day/night preferences, specific requests, continuity problems, number of successive working day, rotating nights/weekends and other working that violate mandatory history contractual requirements are marked as infeasible for a particular nurse and week by giving them a suitably high pij value. 2.2 Integer Programming The problem can be formulated as an integer linear program as follows. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Indices: i = 1...n nurse index; j = 1...m shift pattern index; k = 1...14 day and night index (1...7 are days and 8...14 are nights); s = 1...p grade index. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Constraint set (3) ensures that every nurse works exactly one shift pattern from his/her feasible set, and constraint set (4) ensures that the demand for nurses is covered for every grade on every day and night. Note that the definition of qis is such that higher graded nurses can substitute those at lower grades if necessary. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Typical problem dimensions are 30 nurses of three Integer grades and 400 shift patterns. Thus, Programming formulation has about 12000 binary variables and 100 constraints. This is a moderately sized problem. However, some problem cases remain unsolved after overnight computation using professional software. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | In this section, we introduce concepts from graphical models in general and Bayesian networks in particular. Section 4 will then explain how we applied these concepts to our nurse scheduling problem. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Graphical models are graphs in which nodes represent random variables, and the lack of edges represents conditional independence assumptions (Edwards 2000). They have important applications in many multivariate probabilistic systems in fields such as statistics, systems engineering, information theory and pattern recognition. In particular, they are playing an increasingly important role in the design and analysis of machine learning algorithms. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | theory provides the glue to combine the parts, ensuring that the whole system is consistent, and providing ways to interface models to data. The graph theory provides an intuitively appealing interface by which humans can model highly interacting sets of variables, and a data structure that leads itself naturally to the design of general-purpose algorithms. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | There are two main kinds of graphical models: undirected and directed. Undirected graphical models are more popular with the physics and vision communities. Directed graphical model, also called Bayesian networks, are more popular with the artificial intelligence and machine learning communities. Bayesian networks are often used to model multinomial data with both discrete and continuous variables by encoding the relationship between the variables contained in the modelled data, which represents the structure of a problem. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Moreover, Bayesian networks can be used to generate new instances of the variables with similar properties as those of given data. Each node in the network corresponds to one variable, and each variable corresponds to one position in the strings representing the solutions. The relationship between two variables is represented by a directed edge between the two corresponding nodes. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Any complete probabilistic model of a domain must represent the joint distribution, the probability of every possible event as defined by the values of all the variables. The number of such events is exponential. To achieve compactness, Bayesian networks joint distribution into local conditional distributions for each variable given its parents. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | a sequence of rules corresponding to nurses from the first one to the last one individually. We chose this approach, as the longer-term aim of our research is to model the explicit learning of a human scheduler. Human schedulers can provide high quality solutions, but the task is tedious and often requires a large amount of time. Typically, they construct schedules based on rules to human learnt during scheduling. Due limitations, these rules are typically simple. Hence, our rules will be relatively simple, too. Nevertheless, human generated schedules are of high quality due to the ability of the scheduler to switch between the rules, based on the state of the current solution. We envisage the Bayesian optimisation algorithm to perform this role. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | These probability values can be used to generate new rule strings, or new solutions. Since the first rule in a solution has no parents, it will be chosen from nodes N1j according to their probabilities. The next rule will be chosen from nodes Nij according to their probabilities conditioned on the previous nodes. This building process is repeated until the last node has been chosen from nodes Nmj, where m is number of the nurses. A link from nurse 1 to nurse m is thus created, representing a new possible solution. Since all the probability values are normalized, the roulette-wheel method is good strategy for rule selection. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | For clarity, consider the following toy example of scheduling five nurses with two rules (1: random allocation, 2: allocate nurse to low-cost shifts). In the beginning of the search, the probabilities of choosing rule 1 or 2 for each nurse is equal, i.e. 50%. After a few iterations, due to the selection pressure and reinforcement learning, we experience two solution pathways: Because pure low-cost or random allocation produces low quality solutions, either rule 1 is used for the first 2-3 nurses and rule 2 on remainder or vice versa. In essence, BOA learns ‘use rule 2 after 2-3x using rule 1’ or vice versa. 4.3 A Bayesian Optimization Algorithm Based on the estimation of conditional probabilities, this section introduces a Bayesian optimization algorithm for the nurse scheduling problem. It uses techniques from the field of modelling data by Bayesian networks to estimate the joint distribution of promising solutions. The nodes, or variables, in the Bayesian network correspond to the individual rules by which a schedule will be built step by step. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | The ‘Cover’ rule constructs solutions as follows. For each shift pattern in a nurse’s feasible set, calculate the total number of uncovered shifts and would be covered if the nurse worked that shift pattern. For simplicity, this calculation does not take into account how many nurses are still required in a particular shift. For instance, assume that a shift pattern covers Monday to Friday nights. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Further assume that the current requirements for the nights from Monday to Sunday are as follows: (-3, 0, +1, - 2, -1, -2, 0), where a negative number means undercover and a positive over cover. The Monday to Friday shift pattern hence has a cover value of 3, as the most negative value it covers is -3. In this example, a Tuesday to Saturday pattern would have a value of 2. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | In order to ensure that high-grade nurses are not ‘wasted’ covering unnecessarily for nurses of lower grades, for nurses of grade s, only the shifts requiring grade s nurses are counted as long as there is a single uncovered shift for this grade. If all these are covered, shifts of the next lower grade are considered and once these are filled those of the next lower grade. Due to the nature of this approach, nurses’ preference costs pij are not taken into account by this rule. However, they will influence decisions indirectly via the fitness function. Hence, the ‘Cover’ rule can be summarised as finding those shift patterns with corresponding largest amount of undercover. 4.4.4 Contribution Rule The fourth rule, called ‘Contribution’ rule, is biased towards solution quality but includes some aspects of feasibility by computing an overall score for each feasible pattern for the nurse currently being scheduled. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | The ‘Contribution’ rule is designed to take into account the nurses’ preferences. It therefore works with shift patterns rather than individual shifts. It also takes into account some of the covering constraints in which it gives preference to patterns that cover shifts that have not yet been allocated sufficient nurses to meet their total requirements. This is achieved by going through the entire set of feasible shift patterns for a nurse and assigning each one a score. The one with the highest (i.e. best) score is chosen. If there is more than one shift pattern with the best score, the first such shift pattern is chosen. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Maximum number of generations = 2000; • Penalty weight for each uncovered unit: wdemand =200; • For the ‘k-Cheapest’ rule, k = 5; • Weight set for the ‘Contribution’ rule: w ={8,2,1,1}; • Population size = 140; • Keep the best 40 solution in each generation; • The executing time of the algorithm is approx. 10-20 seconds per run and data instance on a Pentium 4 PC. N.B.: These fixed parameters are not necessarily the best for each instance. At this stage, there are based on our experience and intuition. We have kept them the same for consistency at this stage. When computing the mean a censored cost value of 255 has been used when an algorithm failed to find a feasible solution (N/A). |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Figures 2 and 3 show the results graphically. The bars above the y-axis represent solution quality. The black bars show the number of optimal, the grey near-optimal (within three units) solutions. The bars below the y-axis represent the number of times the algorithm failed to find a feasible solution. Hence, the shorter the bar is below the y-axis and the longer above, the better the algorithm’s performance. Note that ‘empty’ bars mean that feasible, but not optimal solutions were found. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Figure 2 shows that for the Bayesian algorithm 38 out of 52 data sets are solved to or near to optimality. Additionally, feasible solutions are always found for all data sets and hence nothing is plotted below the x-axis. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | For the GA in figure 3 the results are similar: 42 data sets are solved well, however many solutions are infeasible and for two instances not a single feasible solution had been identified. Both algorithms have difficulties solving the later data sets (nurse shortages), but BOA less so than the GA. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | The behaviour of an individual run of the Bayesian algorithm the improvement of the schedule for the 04 data instance. At the generation of 57, the optimal solution cost 17 has been achieved. Although the actual values may differ among various instances, the characteristic shapes of the curves are similar for all seeds and data instances. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Finally, Figure 5 compares performance of different GAs (Aickelin and Dowsland, 2000 and 2003) with the (Basic) Bayesian optimization algorithm presented here. The results are encouraging: with a fraction of the development time and simpler algorithm, the complex genetic algorithms are outperformed terms of feasibility, best and average results. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | includes an the Hill-climbing GA, which Only additional ‘best case’ local search, has a better performance. We believe that once this feature is added into the Bayesian optimization algorithm, we will see the best possible results. Our plan is to implement a post- processor that is similar to a human scheduler who ‘improves’ a finished schedule. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | A new scheduling algorithm based on Bayesian networks is presented in this paper. The approach is novel because it is the first time that Bayesian networks have been applied to the field of personnel scheduling. An effective method is proposed to solve the problem about how to implement explicit learning from past solutions. Unlike most existing approaches, the new approach has the ability to build schedules by using flexible, rather than fixed rules. Experimental results from real-world nurse scheduling problems have demonstrated the strength of the proposed Bayesian optimization algorithm. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | The proposed approach mimics human behaviour much more strongly than a standard GA based scheduling system. Although we have presented this work in terms of nurse scheduling, it is suggested that the main idea of the approach could be applied to many other scheduling problems where the schedules will be built systematically according to specific rules. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | It is also hoped that this research will give some preliminary answers about how to include human-like learning into scheduling algorithms and may therefore be of interest to practitioners and researchers in areas of scheduling and evolutionary computation. In future, we will try to extract the ‘explicit’ part of the learning process further, e.g. by keeping partial solutions and learnt rules from one data instances to the next. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | The work was funded by the UK Government’s major funding agency, Engineering and Physical Sciences Research Council (EPSRC), under grand GR/R92899/01. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Aickelin, U. and Dowsland, K. (2000) "Exploiting Problem Structure in a Genetic Algorithm Approach to a Nurse Rostering Problem," Journal of Scheduling 3(3): 139-153. Aickelin, U. and Dowsland, K. (2002) "Enhanced Direct and Indirect Genetic Algorithm Approaches for a Mall Layout and Tenant Selection Problem," Journal of Heuristics 8(5): 503-514. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Aickelin, U. and Dowsland, K. (2003) “An Indirect Genetic Algorithm for a Nurse Scheduling Problem,” Computers & Operations Research (in print). |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Dowsland, K.A. and Thompson, J.M. (2000) “Nurse Scheduling with Knapsacks, Networks and Tabu Search,” Journal of Operational Research Society: 825-833. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Heckerman, D. (1998) “A Tutorial on Learning with Bayesian Networks,” In Jordan, M. (Ed.), Learning in Graphical Models, MIT Press. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Holland, J.H. (1975) “Adaptation in Natural and Artificial Systems: an Introductory Analysis with Applications to Biology, Control, and Artificial Intelligence,” University of Michigan Press. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Li, J. and Kwan, R.S.K. (2001a) "A Fuzzy Simulated Evolution Algorithm in Proceedings of Congress on Evolutionary Computation (CEC2001), pp. 1115-1122, IEEE Press. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Li, J. and Kwan, R.S.K. (2001b) “A Fuzzy Theory Based Evolutionary Approach for Driver Scheduling,” in Spector, L. et al. (Eds.), Proceedings of the Genetic and Evolutionary Computation Conference (GECCO2001), pp. 1152-1158, Morgan Kaufmann Publishers. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Li, J. and Kwan, R.S.K. (2003) “A Fuzzy Genetic Algorithm for Driver Scheduling,” European Journal of Operational Research 147(2): 334-344. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Pelikan, M. and Goldberg, D. (2000) “Research on the Bayesian Optimization Algorithms,” IlliGAL Report No 200010, University of Illinois. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Pelikan, M., Goldberg, D. and Cantu-Paz, E. (1999) “BOA: The Bayesian Optimization Algorithm,” IlliGAL Report No 99003, University of Illinois. |
Processed_A_Bayesian_Optimisation_Algorithm_for_the_Nurse_Sc.txt | Sitompul, D. and Randhawa, S. (1990) “Nurse Scheduling Models: A State-of-the-art Review,” Journal of the Society of Health System 2: 62-67. |
Processed_Marked_spatial_point_processes:_current_state_and_.txt | proposed by Comas et al. (2011, 2013) for stationary point processes, and Ghorbani et al. |
Processed_Marked_spatial_point_processes:_current_state_and_.txt | Figure 2: Public street trees in Vancouver, Canada, in 2016. Left: different shapes show species of trees; right: colors show diameter at breast height in inches. |
Processed_Marked_spatial_point_processes:_current_state_and_.txt | process with intensity λ if ErN p ˘X X Aqs “ λ|A|, A Ă R2, where N is a count function. |
Processed_Marked_spatial_point_processes:_current_state_and_.txt | means that points of type j tend to maintain an interpoint distance from points of type j. |
Processed_Marked_spatial_point_processes:_current_state_and_.txt | neighbor, rather than considering all potential pairs of points within a distance r from x. |
Processed_Marked_spatial_point_processes:_current_state_and_.txt | tf “ 1 ´ p2{npn ´ 1qq i´1 j“1 mpxpjqq, for 2 ď i ď n. |
Processed_Marked_spatial_point_processes:_current_state_and_.txt | versions are defined for function-valued marks g1p¨q ‰ g2p¨q on F2 at locations x, y P X. |
Processed_Marked_spatial_point_processes:_current_state_and_.txt | point processes on linear networks, similarly to Section 3, we need some form of stationarity. |
Processed_Marked_spatial_point_processes:_current_state_and_.txt | Table 3: Test functions for point processes on linear networks with real-valued marks. The average and variance of all marks are denoted by µm,L and σ2 m,L, and µm,LprLq is the condi- tional mean of the marks for points with an interpoint distance rL. |
Processed_Marked_spatial_point_processes:_current_state_and_.txt | Figure 3: Examples of simulated data. From left to right: Models I, II, and III. Colors show mark values. |
Processed_Marked_spatial_point_processes:_current_state_and_.txt | mmprLq being evaluated over distances r P r0, 200s and rL P r0, 250s, respectively. |
Processed_Marked_spatial_point_processes:_current_state_and_.txt | Figure 4: simulated patterns from Models I, II, and III. From left to right: κmmprq, κL average. From top to bottom: Model I, Model II, and Model III. |
Processed_Marked_spatial_point_processes:_current_state_and_.txt | Figure 5: 95% envelopes for Stoyan’s mark correlation functions, under mark independence, based on 199 simulated patterns. The solid curves show Stoyan’s mark correlation functions for one point pattern per each model I, II, and III (left to right). |
Processed_Marked_spatial_point_processes:_current_state_and_.txt | Figure 6: Mark summary characteristics for butterfly data in Melbourne, Australia. From left to right, cross-type inhomogeneous K-function, cross-type inhomogeneous J-function, Stoyan’s mark correlation function for the Cabbage White type, and Stoyan’s mark correla- tion function for the Little Blue type. |
Processed_Marked_spatial_point_processes:_current_state_and_.txt | that for a pair of nearby trees, at least one of them has a large dbh. |
Processed_Marked_spatial_point_processes:_current_state_and_.txt | small interpoint distances, the dbh value for at least one of them is pretty small. |
Processed_Marked_spatial_point_processes:_current_state_and_.txt | Figure 7: Stoyan’s mark correlation function for the five species of trees alongside the street network of Vancouver, Canada; the real-valued mark is dbh. From left to right: Arnold, Populus, Involucrata, Aquifolium, Bignonioides. The top row considers the network; the bottom row ignores it. |
Processed_Marked_spatial_point_processes:_current_state_and_.txt | Penttinen, A. and D. Stoyan (1989). Statistical analysis for a class of line segment processes. |
Processed_Marked_spatial_point_processes:_current_state_and_.txt | Xie, Z. and J. Yan (2008). Kernel density estimation of traffic accidents in a network space. |
Processed_Flavour_data_constraints_on_new_physics_and_SuperI.txt | We discuss the implications of Bs → µ +µ − and B → K∗µ +µ − decays in the context of indirect searches for new physics, emphasising the new LHCb results. In particular, we derive the conse- quences of the MFV hypothesis and discuss the importance of the MFV predictions. The impact of the recent LHCb measurements in the context of the MSSM will also be addressed, and the SuperIso program will be briefly described. |
Processed_Flavour_data_constraints_on_new_physics_and_SuperI.txt | c(cid:13) Copyright owned by the author(s) under the terms of the Creative Commons Attribution-NonCommercial-ShareAlike Licence. |
Processed_Flavour_data_constraints_on_new_physics_and_SuperI.txt | The importance and deep implications of the information from flavour physics in the search for new physics (NP) is very well known. The flavour sector of the Standard Model (SM) provides tests of the quantum structure of the SM at loop level, of the Minimal Flavour Violation (MFV) hypothesis, and can probe sectors inaccessible to direct searches. In the context of specific NP scenarios such as supersymmetry (SUSY), indirect information from flavour physics is comple- mentary to the direct search information and very strong constraints on the SUSY parameter space can be obtained. |
Processed_Flavour_data_constraints_on_new_physics_and_SuperI.txt | While new physics particles and signatures are searched for very actively in ATLAS and CMS experiments, the LHCb experiment has also a rich BSM program through indirect searches. The key processes here are Bs → µ +µ − and B → K∗µ +µ −, in addition to the CP violating processes (not covered here). Given the impressive progress in the experimental accuracies, it is now crucial to have a precise and clear estimation of the SM predictions and errors in order to deduce solid constraints on the NP parameters. |
Processed_Flavour_data_constraints_on_new_physics_and_SuperI.txt | where Oi(µ) denote the operators and Ci(µ) their corresponding Wilson coefficients evaluated at the scale µ encoding the short distance physics. The primed operators are chirality flipped com- pared to the non-primed operators and are highly suppressed in the SM. In general, contributions from physics beyond the SM to the observables can be described by the modification of Wilson coefficients and/or by the addition of new operators. |
Processed_Flavour_data_constraints_on_new_physics_and_SuperI.txt | Observable 107GeV2 × (cid:104)dBR/dq2 (B → K∗µ +µ −)(cid:105)[1,6] (cid:104)AFB(B → K∗µ +µ −)(cid:105)[1,6] (cid:104)FL(B → K∗µ +µ −)(cid:105)[1,6] 0(B → K∗µ +µ −)/GeV2 q2 Table 1: SM predictions and experimental values of B → K∗µ +µ − observables [4]. |
Processed_Flavour_data_constraints_on_new_physics_and_SuperI.txt | The zero–crossing of the forward-backward asymmetry (q2 cancel out at leading order. Moreover, q2 its measurement allows to remove the sign ambiguity. |
Processed_Flavour_data_constraints_on_new_physics_and_SuperI.txt | Figure 1: Global MFV fit to the various NP coefficients δCi in the MFV effective theory with (upper panel) and without experimental data of LHCb (lower panel). |
Processed_Flavour_data_constraints_on_new_physics_and_SuperI.txt | to the known Yukawa couplings is said to be minimal flavour violating [9]. The hypothesis of MFV provides a model independent solution to the NP flavour. |
Processed_Flavour_data_constraints_on_new_physics_and_SuperI.txt | Compared to a model independent analysis (see for example Refs.[10, 11, 12]), the MFV hypothesis singles out a specific number of operators. Moreover, predictions based on the MFV benchmark have a specific meaning: Any measurement beyond the MFV bounds unambiguously indicates the existence of new flavour structures [9]. |
Processed_Flavour_data_constraints_on_new_physics_and_SuperI.txt | To derive the consequences of the MFV hypothesis we consider a set of ∆F = 1 observ- ables which includes in addition to the ones discussed in the previous section, the radiative decay ¯B → Xsγ, isospin asymmetry ∆0(B → K∗γ) and inclusive ¯B → Xsµ +µ − decay as described in [13]. To obtain constraints on the Wilson coefficients, we scan over δC7, δC8, δC9, δC10 and δCl 0 (where Cl 0 = 2CQ1 = −2CQ2 in MFV). For each point we compute the flavour observables using SuperIso and compare the predictions with the experimental results by calculating the χ 2. The SM predictions and experimental results can be found in Tables 3 and 4 of [13]. The global fit results are shown in Fig. 1. To see the impact of the recent LHCb results, the consequences of the MFV fit prior to the LHCb data are also provided. |
Processed_Flavour_data_constraints_on_new_physics_and_SuperI.txt | Figure 2: Global MFV fit with the latest data set excluding all LHCb measurements of B → K∗µ +µ − observables. |
Processed_Flavour_data_constraints_on_new_physics_and_SuperI.txt | We now turn to the implications in SUSY, focusing on the CMSSM scenario, which assumes SUSY breaking mediated by gravity and is characterised by the set of parameters [m0, m1/2, A0, tan β , sign(µ)] defined by unification boundary conditions at the GUT scale. |
Processed_Flavour_data_constraints_on_new_physics_and_SuperI.txt | we investigate the consequences in the CMSSM by varying all the parameters in random scans. In Fig. 3, we present the effect of this limit in the (M˜t1, tan β ) and (MH+, tan β ) parameter planes. We notice that Bs → µ +µ − excludes the region with tan β (cid:38) 50 independently of the stop and charged Higgs masses, while lower tan β values are only affected for light stop and charged Higgs masses [14]. |
Processed_Flavour_data_constraints_on_new_physics_and_SuperI.txt | Figure 3: Constraints from Bs → µ +µ − on the CMSSM, in the (m˜t1, tan β ) (left panel) and (MH+, tan β ) (right panel) parameter planes. The points compatible with the Bs → µ +µ − are displayed on top. |
Processed_Flavour_data_constraints_on_new_physics_and_SuperI.txt | Figure 4: SUSY spread of the averaged BR(B → K∗µ +µ −), AFB and zero crossing of AFB, in the low q2 region in function of the lightest stop mass, for tan β = 50 and A0 = 0. |
Processed_Flavour_data_constraints_on_new_physics_and_SuperI.txt | The SuperIso program [3] is a public C code dedicated to the computation of flavour physics observables. Various models are implemented, and in particular SM, THDM, MSSM and NMSSM. SuperIso provides a broad set of flavour physics observables, which includes the branching ratio of B → Xsγ, isospin asymmetry of B → K∗γ, branching ratios of Bs,d → µ +µ −, branching ratios of B → Xs(cid:96)+(cid:96)−, B → K∗µ +µ −, as well as several angular quantities in these decays, branching ratio of Bu → τντ , branching ratio of B → Dτντ , branching ratio of K → µνµ , branching ratio of D → µνµ , and the branching ratios of Ds → τντ and Ds → µνµ . SuperIso also provides the calculation of the anomalous magnetic moment of the muon. It uses a SUSY Les Houches Accord (SLHA) file [15] as input, which can be generated automatically by the program via a call to a spectrum generator or provided by the user. An extension of SuperIso which includes the calculation of relic density, SuperIso Relic, is also publicly available [16]. Finally, SuperIso can provide output in the Flavour Les Houches Accord (FLHA) format [17]. |
Processed_Flavour_data_constraints_on_new_physics_and_SuperI.txt | Over the past decades, flavour physics has established its role as an important player in the indirect search for new physics. This has now entered a new era with the start of the LHC and we presented here specific examples to demonstrate the implications of the recent LHCb results. We showed that the MFV hypothesis can be tested in a generic way. The MFV predictions can be checked in the near future and any measurement beyond the predictions would clearly indicate new flavour structure. |
Processed_Flavour_data_constraints_on_new_physics_and_SuperI.txt | The high priority LHCb processes, namely Bs → µ +µ − and B → K∗µ +µ − have impressive impacts both for model independent analysis and specific NP scenarios such as supersymmetry as discussed here. The theoretical uncertainties are under control for both decays and with more data becoming available these decays will play an even more important role in constraining NP scenarios in the near future. |
Processed_Extracting_high_fidelity_quantum_computer_hardware.txt | Abstract. An overview of current status and prospects of the development of quantum computer hardware based on inorganic crystals doped with rare earth ions is presented. Major parts of the experimental work in this area has been done in two places, Canberra, Australia and Lund, Sweden, and the present description follows more closely the Lund work. Techniques will be described that include optimal filtering of the initially inhomogeneously broadened profile down to well separated and narrow ensembles, as well as the use of advanced pulse-shaping in order to achieve robust arbitrary single-qubit operations with fidelities above 90%, as characterized by quantum state tomography. It is expected that full scalability of these systems will require the ability to determine the state of single rare-earth ions. It has been proposed that this can be done using special readout ions doped into the crystal and an update is given on the work to find and characterize such ions. Finally, a few aspects on possibilities for remote entanglement of ions in separate rare-earth-ion-doped crystals are considered. |
Processed_Extracting_high_fidelity_quantum_computer_hardware.txt | hardware based on inorganic crystals doped with rare earth ions. Such crystals are used extensively in current technology, e.g. as active laser materials or as scintillating materials in X-ray detectors. They are manufactured in single crystal growth processes where a dopant material is added during the growth such that the dopant ions replaces host ions in the regular crystal lattice at concentrations that may range between 0.01% and 10% depending on application. Although the average concentration is defined by the dopant concentration in the crystal melt, the relative positions of the dopant ions is largely random. However, randomness and less precise control of construction and growth parameters can be compensated for by using suitable quantum computing schemes and by developing robust pulses enabling high fidelity qubit operations. |
Processed_Extracting_high_fidelity_quantum_computer_hardware.txt | The next section (Section 2) will describe basic properties of rare-earth-ion-doped crystals relevant for quantum information. This includes coherence times which can be several seconds [7] and ion-ion interactions as a mechanism for gate operations. For a laser beam passing through a rare-earth-ion-doped crystal there may be 1015 ions within the laser focal region. However, qubits can still be created and prepared in well defined states and arbitrary qubit operations, characterized by full quantum state tomography, give fidelities, F , in the range 0.91 < F < 0.96 [8]. A brief description on the experimental setup, including the pulse-shaping system, will be given in Section 3, while Section 4 describes how to efficiently initialize the system into well defined qubits, which are inhomogeneous subgroups out of these 1015 ions. In Section 5 we show experimental data on qubits prepared in arbitrary superposition states and characterized by quantum state tomography. The high fidelity results are all based on qubit operations that are robust against individual differences between the ions [9, 10]. We will also present data from the development and use of optimal pulses sequences [11] for operation fidelities > 0.96. Section 6 starts with a description of how to, in contrast to the experiments mentioned above, create qubits in these crystals that each consist of just a single ion, which offers a better scalability. The single ion qubit scheme [12] is based on co-doping the crystals with ”readout ions”. Such readout ions should have a short upper state lifetime in order to enable repeated cycling. One such candidate is the 4f to 5d transition in Ce3+ and experimental data to characterize this transition is also presented in Section 6. The last Section before the summary contains a discussion on entanglement of remote qubits and further scalability. |
Processed_Extracting_high_fidelity_quantum_computer_hardware.txt | be increased, and there are also other rare-earth ions which have even longer coherence times, for example Er, where a T2 of 6.4 ms has been measured [14]. |
Processed_Extracting_high_fidelity_quantum_computer_hardware.txt | The qubit states are not represented by electronically excited states though, but rather by ground state hyperfine levels, shown in the level diagram of Pr3+ in Figure 1. For Pr3+:Y2SiO5 , the hyperfine lifetime is about 90 s and the coherence time about 500 µs. The hyperfine T2 can also be increased by magnetic fields, and up to 860 ms has In addition, dynamic decoupling been demonstrated [15]. sequences (bang-bang pulses) can also be applied to further increase the coherence time to more than 30 s [7]. |
Processed_Extracting_high_fidelity_quantum_computer_hardware.txt | The rare-earth ions are randomly doped into the host crystals, replacing one of the host ions in the lattice. This replacement causes changes to the crystal electric field which then causes each rare-earth ion to pick up a shift in its resonance frequency. This amounts to a large inhomogeneous broadening of the absorption line. For Pr3+, the inhomogeneous linewidth is about 5 GHz, which can be compared to the homogeneous linewidth of about 3 kHz, given by the optical T2, mentioned above. |
Processed_Extracting_high_fidelity_quantum_computer_hardware.txt | Figure 1: Relevant en- ergy level structure for the Pr3+ ion transition utilized in the present work. |
Processed_Extracting_high_fidelity_quantum_computer_hardware.txt | In the rare-earth quantum computing scheme, the inhomogeneous structure is of significant importance, as multiple qubits are addressed in the frequency domain, each qubit having a different resonance frequency. From above, the ratio of inhomogeneous over homogeneous linewidth is about 106 for Pr3+:Y2SiO5 , but since ions do not have only a single transition frequency, the number of frequency channels are in practice much smaller. Fig. 1 shows that the lowest and highest transition frequency for a given ion differ by 4.8 + 4.6 + 10.2 + 17.3 = 36.9 MHz. Thus, about 100 individual frequency channels are a more realistic value. The Pr3+ ion sits in a non-centrosymmetric site, and thus has a permanent electric dipole moment. The value of this dipole moment depends on the electron configuration and the value of the dipole moment of an electronically excited state, is therefore different from that of the ground state. Qubit-qubit gates can then be executed through dipole-dipole interactions. Since the dipole moment changes when the ion becomes excited, exciting an ion will shift the transition frequency of all near-lying ions. An ion belonging to one qubit can then be controlled by the excitation of a another ion, belonging to another qubit, if they are spatially sufficiently close to each other. This effect is used as the control mechanism for conditional gates, and the basis for this was demonstrated in [10]. |
Processed_Extracting_high_fidelity_quantum_computer_hardware.txt | doping concentration of 0.05%, which gives a maximum optical density of about 2 3. In order to get rid of phonon interactions, the crystal is submerged in liquid helium and cooled to 2 K. The available laser power is in the order of 100 mW before the cryostat, which translate into maximum Rabi frequencies of about 1-3 MHz, depending on which transition is targeted, given a laser focus of 100 µm. |
Processed_Extracting_high_fidelity_quantum_computer_hardware.txt | High fidelity gates requires a high-performance pulse-shaping system, as will be further discussed in Section 5. Two acousto-optic modulators (AOMs) are used to create the desired pulses. The incoming continuous wave (CW) photons are deflected by an RF acoustic wave in the AOMs. By varying the power of the acoustic wave, the amplitude of the deflected light is controlled. By virtue of energy and momentum conservation, the deflected photons must increase their frequency by an amount equal to that of the phonons. Thus, by varying the frequency of the acoustic wave we directly control the frequency of the light pulse, and furthermore, by changing the phase of the acoustic wave we also gain control over the optical phase of the light, which is necessary for controlling the coherent interaction with the ions. For these modulators, the power conversion from RF wave to light amplitude is quite non-linear. In order to achieve high accuracy of the pulse amplitude envelopes over as much as six orders of magnitude in dynamic range, an elaborate calibration system, that also includes the power dependence on RF frequency, was setup. The first AOM has a center frequency of 200 MHz and is aligned in a double pass configuration, in order to cancel out spatial movement of the deflected beam. After both passes the AOM has a total deflection range of 200 MHz. The second AOM has a 360 MHz center frequency, and is used only to split a pulse at a single frequency into two frequencies, such that an excitation pulse can be resonant with the transitions at the same time, as further discussed in 0 Section 5.1. |
Processed_Extracting_high_fidelity_quantum_computer_hardware.txt | Starting from the inhomogeneous distribution, we employ advanced optical hole-burning sequences in order to extract a useful qubit system. Initially, the ions occupy all three ground states, and the absorption at any given frequency within the absorption profile, contains contributions from all nine possible transitions (see Figure 1). The goal of the hole-burning is to initialize a narrow controlled ensemble of ions into the -state, well separated in frequency from any other ion. |
Processed_Extracting_high_fidelity_quantum_computer_hardware.txt | At first, a wide spectral hole, henceforth called a pit, is created. The hyperfine level splittings set the limit for the maximum frequency range of such a pit. Since the ions must be in one of the ground states, it can not be larger than 27.5 MHz, as can be seen in Figure 1. This maximum width is then further reduced by the excited state splitting down to 18.1 MHz. Scanned laser pulses target specific transitions inside this 18 MHz region, and are repeated in order to create a spectral pit where all ions have been removed within an interval of maximum 18 MHz. Such a simple pit is shown in Figure 2a. |
Processed_Extracting_high_fidelity_quantum_computer_hardware.txt | Figure 2: (color online) a) shows a simple pit created by repeated scanning across an interval less than 18 MHz. Panel b) displays an improved pit, where scans on multiple intervals have been iteratively repeated in order to shuffle all ions outside the pit as far out as possible, thereby creating an optimal pit. Also note that the scales in the two experimental figures are different. |
Processed_Extracting_high_fidelity_quantum_computer_hardware.txt | Eventually, qubit operation pulses will be used inside the pit structure, and it then becomes important to minimize the interactions between the tails of such pulses and the edges of the pit. The pit shown in Figure 2a displays non-optimal edges, that are a bit rounded and also go up to the maximum αL immediately. In order to improve on this, a more elaborate pit-burning scheme was introduced. In this scheme, pulses that target specific levels of Pr3+ outside the pit are used in order to burn them even further out, as illustrated by the top part of Figure 2b. For an ion having a particular are 1 frequency relative to the pit, as the ion illustrated in this figure, level | i aux targeted in an optical pumping process that eventually transfers them to their i | state. After this, ions of other frequency ranges might fall back down inside the pit, and the initial pit-burning pulses will have to be repeated iteratively together with the secondary pulses to keep the center of the pit empty. A similar procedure is then also repeated on the left side of the pit in order to achieve an optimal result, shown in the lower part of panel b. A more extensive description of the burning procedure can be found in [17] and a detailed list of the exact pulses used to burn such an optimal pit can be found in [18]. |
Processed_Extracting_high_fidelity_quantum_computer_hardware.txt | excited state, and from there down to one of the ground states in a controlled manner. 0 Figure 3a shows the system initialized to one of the ground states, in this case the state. |
Processed_Extracting_high_fidelity_quantum_computer_hardware.txt | Figure 4: (color online) a) shows the paths taken by three ions with different resonance frequencies, while being transferred to the excited state by a coherent sechyp pulse. Panel b) illustrates the dark and the bright states on the Bloch sphere, relevant for the gate operations, and panel c) shows a full qubit gate operation by means of two consecutive dark state pulses (two-color pulses). Note that the transfer in c) is going from and to the bright state. |
Processed_Extracting_high_fidelity_quantum_computer_hardware.txt | their final state, they circle around the top pole, and a higher Rabi frequency would just make them circle around the z-axis faster, not take them away from the desired state. These pulses were investigated more carefully in [9]. |
Processed_Extracting_high_fidelity_quantum_computer_hardware.txt | Figure 5: (color online) a) shows the projective measurements on the x-, y- and z-axis state and correspond respectively. The two left peaks are both transitions from the to a +1 measurement value, while the two right peaks similarly are transitions from the 1 value with the respect to the current axis. Panel b) 1 | shows the result of a quantum state tomography of the six states along the positive and negative axes. |
Processed_Extracting_high_fidelity_quantum_computer_hardware.txt | which is a matrix describing an arbitrary rotation around any vector on the equator of the Bloch sphere. By picking suitable phase angles, one can thus perform any qubit gate operation [9]. |
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.