'Bayesian Optimisation for Nurse Scheduling'

A Bayesian optimisation algorithm for a nurse scheduling problem is presented, which involves choosing a suitable scheduling rule from a set for each nurse's assignment. When a human scheduler works, he normally builds a schedule systematically following a set of rules. After much practice, the scheduler gradually masters the knowledge of which solution parts go well with others. He can identify good parts and is aware of the solution quality even if the scheduling process is not yet completed, thus having the ability to finish a schedule by using flexible, rather than fixed, rules. In this paper, we design a more human-like scheduling algorithm, by using a Bayesian optimisation algorithm to implement explicit learning from past solutions. A nurse scheduling problem from a UK hospital is used for testing. Unlike our previous work that used Genetic Algorithms to implement implicit learning [1], the learning in the proposed algorithm is explicit, i.e. we identify and mix building blocks directly. The Bayesian optimisation algorithm is applied to implement such explicit learning by building a Bayesian network of the joint distribution of solutions. The conditional probability of each variable in the network is computed according to an initial set of promising solutions. Subsequently, each new instance for each variable is generated by using the corresponding conditional probabilities, until all variables have been generated, i.e. in our case, new rule strings have been obtained. Sets of rule strings are generated in this way, some of which will replace previous strings based on fitness. If stopping conditions are not met, the conditional probabilities for all nodes in the Bayesian network are updated again using the current set of promising rule strings. 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, Bayesian network learns 'use rule 2 after 2-3x using rule 1' or vice versa. It should be noted that for our and most other 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. For our problem, we use our rules: Random, Cheapest Cost, Best Cover and Balance of Cost and Cover. In more detail, the steps of our Bayesian optimisation algorithm for nurse scheduling are: 1. Set t = 0, and generate an initial population P(0) at random; 2. Use roulette-wheel selection to choose a set of promising rule strings S(t) from P(t); 3. Compute conditional probabilities of each node according to this set of promising solutions; 4. Assign each nurse using roulette-wheel selection based on the rules' conditional probabilities. A set of new rule strings O(t) will be generated in this way; 5. Create a new population P(t+1) by replacing some rule strings from P(t) with O(t), and set t = t+1; 6. If the termination conditions are not met (we use 2000 generations), go to step 2. Computational results from 52 real data instances demonstrate the success of this approach. They also suggest that the learning mechanism in the proposed approach might be suitable for other scheduling problems. Another direction for further research is to see if there is a good constructing sequence for individual data instances, given a fixed nurse scheduling order. If so, the good patterns could be recognized and then extracted as new domain knowledge. Thus, by using this extracted knowledge, we can assign specific rules to the corresponding nurses beforehand, and only schedule the remaining nurses with all available rules, making it possible to reduce the solution space. Acknowledgements The work was funded by the UK Government's major funding agency, Engineering and Physical Sciences Research Council (EPSRC), under grand GR/R92899/01. References [1] Aickelin U, "An Indirect Genetic Algorithm for Set Covering Problems", Journal of the Operational Research Society, 53(10): 1118-1126,

Innovating a consumer awareness of sweetpotato: industry, research and consumer adoption issues

In Australia, sweetpotato production has grown remarkably (1700%) in the last 16 years. Growers currently market 75 000 t per annum, worth $80-90 million at farm gate. The orange-fleshed cultivars are the most familiar to consumers, but other cultivars with varying flesh colour and properties also have potential for the consumer market. Given that Australian sweetpotato growers desire alternative cultivars to promote market demand, it is important to articulate the characteristics of sweetpotatoes that are most and least desirable for consumers. Research indicates that consumer acceptability of the new cultivar 'Evangeline' may assist sweetpotato growers and marketers in understanding the impact of both sensory properties, such as colour and the importance of flavour and texture of sweetpotatoes, and an awareness of the potential health benefits of sweetpotato consumption. In addition, whilst consumer preferences (regarding size, colour, texture, skin tone) and nutritional knowledge of sweet potato (regarding glycaemic index) is increasing, there is limited research investigating consumers understanding of health messages of sweetpotato attributes. This industry and consumer research review highlights the potential for promoting innovative strategies to improve adoption of new cultivars in the marketplace.

New sweetpotato cultivars to meet market needs

The Australian Sweetpotato Growers Association partnered researchers from Agri-Science Queensland (with co-funding from Horticulture Australia Limited) to identify new, high performing sweetpotato cultivars with diverse colours and tastes. The project evaluated a mix of purple, red, orange and white skin and flesh, tailored for Australian growers and consumers. Australia's sweetpotato market currently relies on one gold cultivar for 90% of national production. Major retailers were requesting a reliable supply of quality sweetpotatoes in emerging categories such as red or white skin or purple flesh. To identify suitable cultivars, over 40 new sweetpotato cultivars were virus tested, and extensively evaluated in multiple experiments in Queensland and northern New South Wales. Larger-scale plantings by growers, using standard agronomy, provided additional performance feedback under commercial conditions. In partnership with growers and wholesalers, cultivars were evaluated in field and laboratory for desired characteristics such as shape, size range, skin and flesh colour, estimates of productivity and suitability for commercial production, cooking characteristics and taste. New high performing gold cultivars had better soil insect and nematode tolerance than the current cultivars. The new colours offered diverse health-related opportunities for consumers, more anthocyanins in purple-fleshed cultivars; higher beta carotene content in new gold fleshed cultivars; and potentially lower GI in white-fleshed cultivars. To enhance adoption, the industry/research partnership will tailor agronomic guidelines to maximise on-farm performance and identify niche marketing pathways for each of the cultivars. Increased consumption of this versatile vegetable will drive sweetpotato industry development and expansion into the future.

Optimising plant bed performance through adaptive research with the Australian sweetpotato industry

A key driver of Australian sweetpotato productivity improvements and consumer demand has been industry adoption of disease-free planting material systems. On a farm isolated from main Australian sweetpotato areas, virus-free germplasm is annually multiplied, with subsequent 'pathogen-tested' (PT) sweetpotato roots shipped to commercial Australian sweetpotato growers. They in turn plant their PT roots into specially designated plant beds, commencing in late winter. From these beds, they cut sprouts as the basis for their commercial fields. Along with other intense agronomic practices, this system enables Australian producers to achieve worldRSQUOs highest commercial yields (per hectare) of premium sweetpotatoes. Their industry organisation, ASPG (Australian Sweetpotato Growers Inc.), has identified productivity of mother plant beds as a key driver of crop performance. Growers and scientists are currently collaborating to investigate issues such as catastrophic plant beds losses; optimisation of irrigation and nutrient addition; rapidity and uniformity of initial plant bed harvests; optimal plant bed harvest techniques; virus re-infection of plant beds; and practical longevity of plant beds. A survey of 50 sweetpotato growers in Queensland and New South Wales identified a substantial diversity in current plant bed systems, apparently influenced by growing district, scale of operation, time of planting, and machinery/labour availability. Growers identified key areas for plant bed research as: optimising the size and grading specifications of PT roots supplied for the plant beds; change in sprout density, vigour and performance through sequential cuttings of the plant bed; optimal height above ground level to cut sprouts to maximise commercial crop and plant bed performance; and use of structures and soil amendments in plant bed systems. Our ongoing multi-disciplinary research program integrates detailed agronomic experiments, grower adaptive learning sites, product quality and consumer research, to enhance industry capacity for inspired innovation and commercial, sustainable practice change.

Ambient stored garlic is higher yielding than garlic stored at cooler temperatures

This experiment evaluated the effect of storage temperature on garlic emergence and yield. Six treatments, '5°C + heat', '5°C - heat', '15°C + heat', '15°C - heat', 'ambient + heat' and 'ambient - heat' were imposed. The heat treatment included a stepped heat treatment of between 30 and 38°C over a 42 day period. Highest emergence rate and yield was obtained for treatments stored under ambient conditions. Garlic stored at cool temperatures of either 5 or 15°C had poorer emergence and lower yields that were commercially unviable. Heat treatment had no observable effect on emergence or yield in ambient or 15°C treatments but in 5°C treatments, yield was significantly greater when the heat treatment was applied but nonetheless was still commercially unviable. The recommendations from this experiment are that in subtropical climates garlic seed-bulbs should be stored under ambient conditions. © ISHS.

Optimization of aeroponic technology for future integration in quality potato seed production systems in tropical environments

Virus and soil borne pathogens negatively impact on the production of potatoes in tropical highland and sub-tropical environments, limiting supply of an increasingly popular and important vegetable in these regions. It is common for latent disease infected seed tubers or field grown cuttings to be used as potato planting material. We utilised an International Potato Centre technique, using aeroponic technology, to produce low cost mini-tubers in tropical areas. The system has been optimised for increased effectiveness in tropical areas. High numbers of seed tubers of cultivar Sebago (630) and Nicola per m2 (>900) were obtained in the first generation, and the system is capable of producing five crops of standard cultivars in every two years. Initial results indicate that quality seed could be produced by nurseries and farmers, therefore contributing to the minimisation of soil borne diseases in an integrated management plan. This technology reduces seed production costs, benefiting seed and potato growers. © ISHS.

Influence of mucuna fallow crop on selected soil properties, weed suppression and taro yields in Taveuni, Fiji

Two field experiments were carried out in Taveuni, Fiji to study the effects of mucuna (Mucuna pruriens) and grass fallow systems at 6 and 12 month durations on changes in soil properties (Experiment 1) and taro yields (Experiment 2). Biomass accumulation of mucuna fallow crop was significantly higher (P<0.05) than grass fallow crop at both 6 and 12 month durations. The longer fallow duration resulted in higher (P<0.05) total soil organic carbon, total soil nitrogen and earthworm numbers regardless of fallow type. Weed suppression in taro grown under mucuna was significantly greater (P<0.05) than under natural grass fallow. Taro grown under mucuna fallow significantly outyielded taro grown under grass fallow (11.8 vs. 8.8 t ha-1). Also, the gross margin of taro grown under mucuna fallow was 52% higher than that of taro grown under grass fallow. © ISHS.

An Indirect Genetic Algorithm for Set Covering Problems

This paper presents a new type of genetic algorithm for the set covering problem. It differs from previous evolutionary approaches first because it is an indirect algorithm, i.e. the actual solutions are found by an external decoder function. The genetic algorithm itself provides this decoder with permutations of the solution variables and other parameters. Second, it will be shown that results can be further improved by adding another indirect optimisation layer. The decoder will not directly seek out low cost solutions but instead aims for good exploitable solutions. These are then post optimised by another hill-climbing algorithm. Although seemingly more complicated, we will show that this three-stage approach has advantages in terms of solution quality, speed and adaptability to new types of problems over more direct approaches. Extensive computational results are presented and compared to the latest evolutionary and other heuristic approaches to the same data instances.

'Bayesian Optimisation for Nurse Scheduling'

A Bayesian optimisation algorithm for a nurse scheduling problem is presented, which involves choosing a suitable scheduling rule from a set for each nurse's assignment. When a human scheduler works, he normally builds a schedule systematically following a set of rules. After much practice, the scheduler gradually masters the knowledge of which solution parts go well with others. He can identify good parts and is aware of the solution quality even if the scheduling process is not yet completed, thus having the ability to finish a schedule by using flexible, rather than fixed, rules. In this paper, we design a more human-like scheduling algorithm, by using a Bayesian optimisation algorithm to implement explicit learning from past solutions. A nurse scheduling problem from a UK hospital is used for testing. Unlike our previous work that used Genetic Algorithms to implement implicit learning [1], the learning in the proposed algorithm is explicit, i.e. we identify and mix building blocks directly. The Bayesian optimisation algorithm is applied to implement such explicit learning by building a Bayesian network of the joint distribution of solutions. The conditional probability of each variable in the network is computed according to an initial set of promising solutions. Subsequently, each new instance for each variable is generated by using the corresponding conditional probabilities, until all variables have been generated, i.e. in our case, new rule strings have been obtained. Sets of rule strings are generated in this way, some of which will replace previous strings based on fitness. If stopping conditions are not met, the conditional probabilities for all nodes in the Bayesian network are updated again using the current set of promising rule strings. 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, Bayesian network learns 'use rule 2 after 2-3x using rule 1' or vice versa. It should be noted that for our and most other 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. For our problem, we use our rules: Random, Cheapest Cost, Best Cover and Balance of Cost and Cover. In more detail, the steps of our Bayesian optimisation algorithm for nurse scheduling are: 1. Set t = 0, and generate an initial population P(0) at random; 2. Use roulette-wheel selection to choose a set of promising rule strings S(t) from P(t); 3. Compute conditional probabilities of each node according to this set of promising solutions; 4. Assign each nurse using roulette-wheel selection based on the rules' conditional probabilities. A set of new rule strings O(t) will be generated in this way; 5. Create a new population P(t+1) by replacing some rule strings from P(t) with O(t), and set t = t+1; 6. If the termination conditions are not met (we use 2000 generations), go to step 2. Computational results from 52 real data instances demonstrate the success of this approach. They also suggest that the learning mechanism in the proposed approach might be suitable for other scheduling problems. Another direction for further research is to see if there is a good constructing sequence for individual data instances, given a fixed nurse scheduling order. If so, the good patterns could be recognized and then extracted as new domain knowledge. Thus, by using this extracted knowledge, we can assign specific rules to the corresponding nurses beforehand, and only schedule the remaining nurses with all available rules, making it possible to reduce the solution space. Acknowledgements The work was funded by the UK Government's major funding agency, Engineering and Physical Sciences Research Council (EPSRC), under grand GR/R92899/01. References [1] Aickelin U, "An Indirect Genetic Algorithm for Set Covering Problems", Journal of the Operational Research Society, 53(10): 1118-1126,

'An Indirect Genetic Algorithm for Set Covering Problems'

This paper presents a new type of genetic algorithm for the set covering problem. It differs from previous evolutionary approaches first because it is an indirect algorithm, i.e. the actual solutions are found by an external decoder function. The genetic algorithm itself provides this decoder with permutations of the solution variables and other parameters. Second, it will be shown that results can be further improved by adding another indirect optimisation layer. The decoder will not directly seek out low cost solutions but instead aims for good exploitable solutions. These are then post optimised by another hill-climbing algorithm. Although seemingly more complicated, we will show that this three-stage approach has advantages in terms of solution quality, speed and adaptability to new types of problems over more direct approaches. Extensive computational results are presented and compared to the latest evolutionary and other heuristic approaches to the same data instances.

Historia de la Revolución de la República de Colombia en la América Meridional. Tomos I y II

Este es uno de los textos claves de la historiografía colombiana. Es una invitación a leer juiciosamente este libro, pues aunque no se compartan muchas de las ideas en él expuestas, ellas han tenido un peso decisivo en la conformación de la visión que se tiene hoy de la historia del país. Los personajes de la independencia en la obra del autor no son presentados en la forma idílica que una versión de la historia patria posterior ha difundido. Lejos de mostrar a estos héroes como prototipos de virtudes, amor por la patria, abnegación ante los infortunios, como modelos a seguir por las generaciones posteriores, en el, el autor señala las razones personales y de grupo que a su juicio los llevaron a actuar de una determinada manera. Así por ejemplo, la imagen que se da de la participación de Antonio Nariño en las luchas por el poder en el Estado de Cundinamarca en 1811 es la de un enemigo peligroso e insatisfecho por ocupar un cargo subalterno. La escritura de esta obra se enmarca dentro de la tradición de pensamiento racionalista del siglo XVIII.

MODELO PARA LA EVALUACIÓN FINANCIERA DE UNA INSTITUCIÓN PÚBLICA DE SERVICIO DE EDUCACIÓN SUPERIOR

El propósito fundamental del presente estudio es la elaboración de un modelo de evaluación cuantitativa de la dimensión financiera para ser utilizado en los procesos de evaluación y planificación de entidades públicas de servicio sin fines de lucro.La propuesta consiste en un instrumento analítico que coincide, en algunos aspectos, con el modelo tradicional de razones financieras, en razón de que el fenómeno en estudio es muy similar. Sin embargo, presenta diferencias importantes, las cuales ponen de manifiesto las características concretas y distantes de la organización de interés. En el caso de la universidad pública, por ejemplo, es indispensable tener presente las características en torno a sus objetivos (sobre todo la poca importancia en la ganancia) y a sus propietarios (especialmente por el fuerte compromiso con los contribuyentes del país).El modelo de evaluación financiera consta de cuatro dimensiones de análisis de las finanzas universitarias: la liquidez o la capacidad de la institución para hacer frente a las obligaciones de corto plazo; la administración de activos; la administración de las deudas; y la eficiencia y la eficacia con que opera la institución. Se pudo determinar que a pesar de la amplia validez del enfoque tradicional en dos de esos aspectos: la liquidez y la administración de deudas, fue necesaria una revisión cuidadosa de los indicadores tradicionales en vista de que las características concretas de la Universidad demostraron la conveniencia del diseño de otros indicadores para lograr una mejor evaluación.Las conclusiones más importantes del estudio son dos: que la evaluación financiera en los términos en que fue planteada no solo es pertinente sino que además es perfectamente factible. Al final se mencionan algunas recomendaciones que persiguen, con los propósitos del estudio, fortalecer el compromiso con el cumplimiento de los objetivos por los que la institución fue creada.

DNA 11

Combined media on photographic paper on steel. 39" x 29"

Development of super early genotypes for the dry bean (Phaseolus vulgaris) as affected by nitrogen management.

The super early genotypes (SEG) of dry bean (Phaseolus vulgaris L.) have a shorter life cycle (65-75 days) when compared with the season length of traditional cultivars (90-100 days). Timing of nitrogen top-dressing fertilization could be different because of this reduction in length of the SEG life cycle. This study aimed at characterizing, by using growth analysis and vegetation index, super early genotypes of dry bean development as affected by timing of nitrogen application. Field experiments were conducted in the 2014 and 2015 growing seasons in central Brazil with a randomized block experimental design with split plots scheme and four replicates. The plots comprised the dry bean genotypes (Colibri ? check cultivar, CNFC 15873, CNFC 15874, and CNFC 15875), and subplots comprised applications of N at different timings: 90 kg of N at sowing, 90 kg N at top-dressing; 45 kg of N at sowing plus 45 kg at top-dressing, with urea as the source of N. We also used a control treatment without N application. The CNFC 15874 super early genotype of dry bean had the higher grain yield (2776 kg ha-1) and differed from the CNFC 15873 genotype (2492 kg ha-1). Nitrogen fertilization allowed higher grain yield (2619 kg ha-1, when applied N at sowing, 2605 kg ha-1, when applied N at sowing and at top-dressing, and 2680 kg ha-1, when applied N at top-dressing) than the control, 2360 kg ha-1 (no N fertilization). The time of N fertilization in super early genotype of dry bean did not affect grain yield.