Auswahl der wissenschaftlichen Literatur zum Thema „Recurrent Problems“

Varicocele is one of the most common problems in modern reproductive medicine. The incidence of varicocele in the structure of the general male population is 15%, 40% of which have problems with fertility. Among the causes of male infertility, varicocele ranks second after idiopathic, thus being the most common curable cause of male infertility. While researching the pathophysiological mechanisms of infertility in varicocele, the question of the reasons for the varicocele relapses, both after surgical and endovascular methods of treatment remains open, as well as the tactics of managing such patients. The review aimed to systematize knowledge about the problem of recurrent varicocele, to analyze the frequency and etiology of relapses after various methods of primary treatment, as well as to select the optimal diagnostic and treatment option for varicocele recurrence.

Recurrent neural networks (RNNs) are an excellent fit for regression problems where sequential data are the norm since their recurrent internal structure can analyse and process data for long. However, RNNs are prone to the phenomenal vanishing gradient problem (VGP) that causes the network to stop learning and generate poor prediction accuracy, especially in long-term dependencies. Originally, gated units such as long short-term memory (LSTM) and gated recurrent unit (GRU) were created to address this problem. However, VGP was and still is an unsolved problem, even in gated units. This problem occurs during the backpropagation process when the recurrent network weights tend to vanishingly reduce and hinder the network from learning the correlation between temporally distant events (long-term dependencies), that results in slow or no network convergence. This study aims to provide an empirical analysis of LSTM networks with an emphasis on inefficiency in long-term dependencies convergence because of VGP. Case studies on NASA’s turbofan engine degradation are examined and empirically analysed.

We consider the method of Reduction of Dissipativity Domain to prove global Lyapunov stability of Discrete Time Recurrent Neural Networks. The standard and advanced criteria for Absolute Stability of these essentially nonlinear systems produce rather weak results. The method mentioned above is proved to be more powerful. It involves a multi-step procedure with maximization of special nonconvex functions over polytopes on every step. We derive conditions which guarantee an existence of at most one point of local maximum for such functions over every hyperplane. This nontrivial result is valid for wide range of neuron transfer functions.

Several studies show that African-Americans are less likely than whites to use alcohol or drugs. However, if African-Americans use drugs then they are more likely to become heavy and persistent users. African-Americans are also more likely to have a current substance abuse disorder. There is not much in the literature to explain this phenomenon. The purpose of this study is to examine the alcohol and drug abuse, use and dependence of blacks and whites in order to explain the differences in the course of the substance disorder, using data from the National Comorbidity Survey. There are many variables thought to contribute to the racial difference, such as socio-economic status (measured by income and education), religion, insurance, employment status, and marital status. The data in this literature indicate that the aforementioned variables do not explain the racial difference in substance disorders. However, after performing interaction analyses, it is clear that the effects of treatment are different for blacks and whites. Treatment is more effective for whites, and it may even cause the substance disorder to become worse for blacks. Several studies indicate that this may be the result of cultural differences between the treatment staff and the clients.
Master of Science

Research was undertaken based upon the suspicion that there were recurrent problems with the design and construction of hospital en-suites. An investigation was planned to confirm whether or not this suspicion was true and that there were, in fact, recurrent problems. Following on from this it would be to find out if they existed, what they were and whether and how they could be overcome. These objectives needed to be formulated, and then positioned within a framework that identified if the practitioners within the industry even knew about the problems and if they did, whether they chose not to resolve them or could not overcome the barriers that prevented the resolution. A review of the Literature identified that little research had been undertaken in the area of problems with the design and construction of hospital en-suites. The broader perspective of problems within the construction industry was covered by the theories supporting construction Process Improvement. This area was reviewed and it was found that the results of practical implementation of Process Improvement were rather less successful than had originally been anticipated. The Sub-processes that combined together to make up the whole processes were considered and reviewed also. Theories relating to the improvement of the Sub-process of Design; Management, Quality and Briefing were reviewed. These reviews of the Process and Design Sub-processes identified that there were significant barriers within organisation that made improvements difficult to implement. The opportunity arose to review the suggested existence of problems with the design and construction of hospital en-suites by way of case Study research at the £330million Derby PFI hospital redevelopment. At this project approximately 400 en-suites were going to be constructed. Action Research was proposed to implement practical interventions aimed at identifying the problems and their solutions. During the course of the research a second project, the similarly sized £330 million PFI redevelopment of the Mansfield hospital project, also with approximately 400 en-suites. This provided the basis for a second Case Study. The original research approach of Action Research within a Case Study was modified to two Case Studies each embracing Action Research contained within an overarching Action Research strategy. Planned interventions in the construction processes were undertaken to identify whether knowledge could become embedded in that construction process that would eliminate any problems identified in the design and construction of hospital en-suites. The techniques for information gathering included workshops, constructing Mock-ups, Simulation exercises, Structured interview based on carefully prepared Questionnaires to form agenda's and Site Visits. From the research emerged the 'Notion of Solute' as a theoretical model for problem resolution which overcomes the difficulty of Knowledge Transfer. From the investigations within the two Case Studies into hospital en-suites a solution to the problems required the transfer of the knowledge of three components: The nature of the problems, The implications of not remedying the problems. Recommendations for resolving the problems. These three components, which formulate the solutions, were suggested to be defined as a 'Solute'. This is a distinct and focussed 'package of knowledge' that can be readily transferred, both between individuals and also between projects. This 'Solute' is a theoreticz model that could be transferred to other problems within hospitals, other problems within construction projects and even to problems in other industries. Each 'Solute' is then populated with the specific parameters required to enable a problem to be understood and a solution to be implemented. Keywords: Knowledge, Design, Quality, Hospital, En-suite.

The purpose of my doctorate work has consisted in the exploration of the potentialities and of the effectiveness of different neural classifiers, by experimenting their application in the solution of classification problems occurring in the fields of interest typical of the research group of the “Laboratorio Circuiti” at the Department of Electronic Engineering in Tor Vergata. Moreover, though inspired by works already developed by other scholars, the adopted neural classifiers have been partially modified, in order to add to them interesting peculiarities not present in the original versions, as well as to adapt them to the applications of interest. These applications can be grouped in two great families. As regards the first application, the objects to be classified are identified by features of static nature, while as regards the second family, the objects to be classified are identified by features evolving in time. In relation to the research fields taken as reference, the ones that belong to the first family are the following: • classification, by means of fuzzy algorithms, of acoustic signals, with the aim of attributing them to the source that generated them (recognition of musical instruments) • exclusive classification of simple human motor acts for the purpose of a precocious diagnosis of nervous system diseases The second family of application has been represented by that research field that aims to the development of neural tools for the Automatic Tanscription of piano pieces. The first part of this thesis has been devoted to the detailed description of the adopted neural classification techniques, as well as of the modifications introduced in order to improve their behavior in relation to the particular applications. In the second part, the experiments by means of which I have estimated the before-mentioned neural classification techniques have been introduced. It exactly deals with experiments carried out in the chosen research fields. For every application, the II results achieved have been reported; in some cases, the further steps to perform have also been proposed. After a brief introduction to the biological neural model, a description follows about the model of the artificial neuron that has afterwards inspired all the other models: the one proposed by McCulloch and Pitts in 1943. Subsequently, the different typologies of architectures that characterize neural networks are shortly introduced, as regards the feed-forward networks as well as the recursive networks. Then, a description of some learning strategies (supervised and unsupervised), adopted in order to train neural networks, is also given; some criteria by means of which one can estimate the goodness of an opportunely trained neural network are also given (errors made vs. generalization capability). A great part of the adopted networks is based on adaptations of the Backpropagation algorithm; the other networks have been instead trained by means of algorithms based on statistical or geometric criteria. The Backpropagation algorithm has been improved by augmenting the degrees of freedom to the learning ability of a feed-forward neural network with the introduction of a spline adaptive activation function. A wide description has been given of the recurrent neural networks and particularly of the locally recurrent neural networks, networks for dynamic classification exploited in the automatic transcription of piano music. After a more or less rigorous definition of the concepts of classification and clustering, some paragraphs have been devoted to some statistical and geometric neural architectures, exploited in the implementation of static classifiers of common use and in particular in the application fields that have regarded my doctorate work. A separate paragraph has been devoted to the Simpson’s classifier and to the variants originated from my research work. They have revealed themselves to be static classifiers very simple to implement and at the same time very ductile and efficient, in many situations as well as regards the problem of musical source recognition. Two have been the choices in this case. In the first one, III these classifiers have been trained, by means of a pure supervised learning approach, while in the second the training algorithm, though keeping a substantially supervised nature, is prepared by a clustering phase, with the aim of improving, in terms of errors and generalization, the covering of the input space. Subsequently, the locally recurrent neural networks seen as dynamic classifiers are retrieved. However, their training has been rethought according to the effective reduction of the classification error instead of the classic mean-square error. The last three paragraphs have been devoted to a detailed description, in terms of specifications, implementative choices and final results, of the aforesaid fields of applications. The results obtained in all the three fields of application can be considered encouraging. Particularly, the recognition of musical instruments by means of the adopted neural networks has shown results tha can be considered out comparable if not better than those obtained by means of other techniques, but with considerably less complex structures. In case of the Automatic Transcription of piano pieces, the dynamic networks I adopted have given good results. Unfortunately, the required computational resources required by such networks cannot be considered negligible. As far as the medical applications, we are still in an incipent phase of the research. However, opinions expressed by those people who work in this field can be considered substantially eulogistic. The research activities my doctorate work is part of have been carried out in collaboration with the Department “INFOCOM” of the first University of Rome “La Sapienza”, as far as the recognition of musical instruments and the Automatic Transcription of piano pieces. The necessity to study the potentialities of neural classifiers in medical application has instead come from a profitable existing collaboration with the Istituto Superiore di Sanità in Rome.

Ce travail s'intéresse aux problèmes de décision pour lesquels on cherche une solution optimale ou quasi-optimale et dont il faut résoudre plusieurs instances successives (problèmes récurrents) lesquelles sont des variantes d'un même problème d'origine.On analyse la structure de tels problèmes afin de dégager les caractéristiques pouvant être exploitées efficacement et transférées d'une résolution à l'autre, afin d'améliorer incrémentalement la qualité de l'optimisation.On se place donc dans le cadre d'une interaction entre un processus d'apprentissage automatique (fouille de données d'optimisation) et un processus d'optimisation.D'une part, étant donné l'expérience de résolutions passées, on cherche à apprendre ce que l'on peut généraliser au problème courant.D'autre part, on cherche à utiliser ces connaissances au sein de l'algorithme d'optimisation afin de rendre son exécution plus efficace.En particulier, cette thèse présente trois contributions. La première introduit une méthode pour générer des sous-problèmes plus simple pour un instance d’un problème récurrent, en utilisant la classification multi-étiquette. Un sous-ensemble de variables décision est sélectionné et figé à une valeur de référence.La solution au sous-problème qui reste, même en étant pas garantie optimale pour le problème originale, peut être obtenu plus rapidement.La deuxième, emploie l’apprentissage supervisé, classification et régression, pour prédire et ajouter une contrainte additionnelle au problème récurrent modélisé par programmation mathématique. Au moment de résoudre une nouvelle instance, le modèle prédit en quelle mesure la solution au problème de référence est applicable, en permettant d’obtenir une résolution plus rapide.Dans la troisième, le contrôle dynamique de paramètres d’un algorithme évolutionnaire est encadré comme un problème d’apprentissage par renforcement. Les politiques de contrôle ainsi obtenues garantissent que l’algorithme d’optimisation atteint, en moyenne, la solution optimale dans le plus court délai
The interest is on those decision problems for which an optimal or quasi-optimal solution is sought, and for which it is necessary to solve successive instances (recurrent problems) that are variations of a common original problem.The structure of such problems is analysed to identify the characteristics that can be exploited and transferred from one resolution to another, to incrementally improve the quality of the optimization process. The research is characterized by the interaction between a process of statistical learning (from optimization data) and a process of optimization. The information extracted from past resolutions is generalized to the current problem and integrated into the optimization algorithm to make its execution more resource-efficient.In particular, this thesis presents three contributions.The first, introduces a method that generates a simpler sub-problem to an instance of a recurrent problem, using multi-label classification. A subset of decision variables is selected and set to a reference value. The solution to the remaining sub-problem, while not guaranteed to be optimal for the original problem, can be obtained faster.The second employs Supervised Learning, classification and regression, to predict an additional constraint to a reference recurrent problem modelled via Mathematical Programming. When a new instance is solved, the model predicts how much of the solution to the reference problem is still applicable, allowing for a more rapid resolution.In the third, the dynamic control of the parameters of Evolutionary Algorithms is framed as a Reinforcement Learning problem. The control policies obtained guarantee that the optimization algorithm reaches an optimal solution within the shortest, average time

Under recent years the gambling industry has been moving towards providing their customer the possibility to gamble online instead of visiting a physical location. Aggressive marketing, fast growth and a multitude of actors within the market have resulted in a spike of customers who have developed a gambling problem. Decision makers are trying to fight back by regulating markets in order to make the companies take responsibility and work towards preventing these problems. One method of working proactively in this regards is to identify vulnerable customers before they develop a destructive habit. In this work a novel method of predicting customers that have a higher risk in regards to gambling-related problems is explored. More concretely, a recurrent neural network with long short-term memory cells is created to process raw behaviour data that are aggregated on a daily basis to classify them as high-risk or not. Supervised training is used in order to learn from historical data, where the usage of permanent self-exclusions due to gambling related problems defines problem gamblers. The work consists of: obtain a local optimal configuration of the network which enhances the performance for identifying problem gam- blers who favour the casino section over sports section, and analyze the model to provide insights in the field. This project was carried out together with LeoVegas Mobile Gaming Group. The group offers both online casino games and sports booking in a number of countries in Europe. This collaboration made both data and expertise within the industry accessible to perform this work. The company currently have a model in production to perform these predictions, but want to explore other approaches. The model that has been developed showed a significant increase in performance compared to the one that is currently used at the company. Specifically, the precision and recall which are two metrics important for a two class classification model, increased by 37% and 21% respectively. Using raw time series data, instead of aggregated data increased the responsiveness regarding customers change in behaviour over time. The model also scaled better with more history compared to the current model, which could be a result of the nature of a recurrent network compared to the current model used.

O presente trabalho propõe resolver o clássico problema combinatorial conhecido como problema do caixeiro viajante. Foi usado no sistema de otimização de busca do menor caminho uma rede neural recorrente. A topologia de estrutura de ligação das realimentações da rede adotada aqui é conhecida por rede recorrente de Wang. Como regra de treinamento de seus pesos sinápticos foi adotada a técnica de perturbação simultânea com aproximação estocástica. Foi elaborado ainda uma minuciosa revisão bibliográfica sobre todos os temas abordados com detalhes sobre a otimização multivariável com perturbação simultânea. Comparar-se-á também os resultados obtidos aqui com outras diferentes técnicas aplicadas no problema do caixeiro viajante visando propósitos de validação.
This work proposes to solve the classic combinatorial optimization problem known as traveling salesman problem. A recurrent neural network was used in the system of optimization to search the shorter path. The structural topology linking the feedbacks of the network adopted here is known by Wang recurrent network. As learning rule to find the appropriate values of the weights was used the simultaneous perturbation with stochastic approximation. A detailed bibliographical revision on multivariable optimization with simultaneous perturbation is also described. Comparative results with other different techniques applied to the traveling salesman are still presented for validation purposes.

Causality and Predictability of Complex Systems pose fundamental challenges even under well-defined structural stochastic-dynamic conditions where the laws of motion and system symmetries are known. However, the edifice of complexity can be profoundly transformed by structural-functional coevolution and non-recurrent elusive mechanisms changing the very same invariants of motion that had been taken for granted. This leads to recurrence collapse and memory loss, precluding the ability of traditional stochastic-dynamic and information-theoretic metrics to provide reliable information about the non-recurrent emergence of fundamental new properties absent from the a priori kinematic geometric and statistical features. Unveiling causal mechanisms and eliciting system dynamic predictability under such challenging conditions is not only a fundamental problem in mathematical and statistical physics, but also one of critical importance to dynamic modelling, risk assessment and decision support e.g. regarding non-recurrent critical transitions and extreme events. In order to address these challenges, generalized metrics in non-ergodic information physics are hereby introduced for unveiling elusive dynamics, causality and predictability of complex dynamical systems undergoing far-from-equilibrium structural-functional coevolution. With these methodological developments at hand, hidden dynamic information is hereby brought out and explicitly quantified even beyond post-critical regime collapse, long after statistical information is lost. The added causal insights and operational predictive value are further highlighted by evaluating the new information metrics among statistically independent variables, where traditional techniques therefore find no information links. Notwithstanding the factorability of the distributions associated to the aforementioned independent variables, synergistic and redundant information are found to emerge from microphysical, event-scale codependencies in far-from-equilibrium nonlinear statistical mechanics. The findings are illustrated to shed light onto fundamental causal mechanisms and unveil elusive dynamic predictability of non-recurrent critical transitions and extreme events across multiscale hydro-climatic problems.

There is a reasonable likelihood that that the next global economic crisis could threaten the future of democracy. The economic system is a complex, adaptive system (CAS) subject to “tipping points” when underlying stresses lead to crisis. Moreover, the economic system is nested within a number of other CAS; political, environmental and public health among others. Looking forward, recurrent negative supply shocks imply a dangerous future of higher real interest rates and debt distress leading to either deflation (private debt distress) or higher inflation (sovereign debt distress). Such problems could threaten democratic political systems that are already showing signs of significant stress themselves. The paper finishes with some reflections on policy alternatives.

Se propone un método basado en homología persistente para la estimación del número de clusters en dos tipos de problemas de clasificación no supervisada recurrentes en el paradigma SON de gestión de redes de comunicaciones celulares.

Drivers’ noncompliance with the posted speed limit reductions on rural-to-urban freeway transition zones is a recurrent problem in Indiana and other states. Speed-reduction treatments, such as pavement markings and active signing, aim to solve this issue. Optical speed bars, longitudinal speed reduction markings, and speed feedback signs were identified as the most promising speed-reduction treatments for experimental implementation in Indiana. Probe-vehicle-based speed data for 1 year was acquired on selected freeway segments to measure the speed effect. A set of fixed-effects linear regression models were used to estimate the effect of speed-reduction treatments on three key speed behavior characteristics—average speed, 90th speed percentiles, and speed variability. Optical speed bars together with speed feedback signs are a promising combination of treatments. While this and other traditional speed-reduction treatments show promise, their influence area is limited to a portion of target road segments. Therefore, future speed management needs to consider other speed-reduction treatments, particularly automated and area-wide speed enforcement, to maintain reduced speeds inside urban areas.

Scrubbers and separators are used in natural gas pipelines to remove solid and liquid materials from the gas stream. Failure to remove the entrained materials from the gas can result in equipment damage, increased pressure drop due to liquid accumulation, flow measurement errors, and corrosion. The performance of separators is rarely tested after a separator is installed because there is a lack of test instrumentation and it is difficult to conduct tests at the high pressures. The only indicators of poor separator performance are recurring problems in downstream equipment or the detection of accumulated materials in downstream piping. Instrumentation is needed that can verify separator performance when the unit is installed and to periodically monitor separator performance. The report documents results of instrument tests. The objectives of the instrument evaluations were to verify that the instruments could be used to measure particles penetrating a separator, to provide a comparative evaluation of the instruments, and to identify any measurement problems that could be encountered in field testing. One important result was that the separator minimum removable drop size increased as the operating pressure increased. This trend is not generally known, since there is a lack of test results for pressures above atmospheric pressure. The separator performance test results are documented in this report. Two different particle measuring instruments were evaluated for documenting separator performance. The two instruments were the video imaging system with automatic image analysis and the laser-based phase Doppler particle measuring system. The instruments were evaluated in laboratory tests that were conducted on a commercially available vane-type separator. The objectives of the instrument evaluations were to verify that the instruments could be used to measure particles penetrating a separator, to provide a comparative evaluation of the two instruments, and to identify any measurement problems that could be encountered in field testing. The video imaging system has a number of attractive attributes, but it was not able to measure the small diameter drops at the separator exit. The primary limitation was that the optical system could not clearly image the small drops (in the range from 5 to 30 um). The phase Doppler particle measuring system was capable of measuring all of the parameters needed to document the separator performance. Based on the instrument evaluations, future efforts on developing measurement methods for documenting separator performance should focus on adapting the phase Doppler system to field testing.

Water-quality monitoring in National Park Service units of the Northern Colorado Plateau Network (NCPN) is made possible through partnerships between the National Park Service Inventory & Monitoring Division, individual park units, the U.S. Geological Survey, and the Utah Division of Water Quality. This report evaluates data from site visits at 62 different locations on streams, rivers, and reservoirs in or near ten NCPN park units between October 1, 2015 and September 30, 2018. Data are compared to state water-quality standards for the purpose of providing information to park managers about potential water-quality problems. The National Park Service does not determine the regulatory status of surface waters; state water quality agencies determine whether waters comply with the Clean Water Act. Evaluation of water-quality parameters relative to state water-quality standards indicated that 17,997 (96.8%) of the 18,583 total designated beneficial-use evaluations completed for the period covered in this report met state water-quality standards. The most common exceedances or indications of impairment, in order of abundance, were due to elevated nutrients, elevated bacteria (E. coli), elevated water temperature, elevated trace metals, elevated total dissolved solids (and sulfate), elevated pH, and low dissolved oxygen. While some exceedances were recurring and may have been caused by human activities in the watersheds, many were due to naturally occurring conditions characteristic of the geographic setting. This is most apparent with phosphorus, which can be introduced into surface water bodies at elevated levels by natural weathering of the geologic strata found throughout the Colorado Plateau. Higher phosphorus concentrations could also be attributed to anthropogenic activities that can accelerate erosion and transport of phosphorus. Some activities that can increase erosional processes include grazing, logging, mining, pasture irrigation, and off-highway vehicle (OHV) use. Exceedances for total phosphorus were common occurrences at nine out of ten NCPN park units, where at least one site in each of these parks had elevated phosphorus concentrations. At these sites, high levels of nutrients have not led to algal blooms or other signs of eutrophication. Sites monitored in Arches National Park (NP), Black Canyon of the Gunnison NP (BLCA), Bryce Canyon NP (BRCA), Capitol Reef NP (CARE), Curecanti National Recreation Area (CURE), Dinosaur National Monument (DINO), and Zion NP (ZION) all had E. coli ex-ceedances that could be addressed by management actions. While many of these sites already have management actions underway, some of the actions necessary to bring these waters into compliance are beyond the control of the National Park Service. Changes to agricultural practices to improve water quality involves voluntary participation by landowners and/or grazing permittees and their respective states. This could be the case with lands upstream of several parks with E. coli contamination issues, including Red Rock Canyon (BLCA); Sul-phur, Oak, and Pleasant creeks (CARE); Blue Creek and Cimarron River (CURE); Brush and Pot creeks (DINO); and North Fork Virgin River (ZION). Issues with E. coli contamination at Yellow Creek (BRCA) seemed to be resolved after the park boundary fence downstream of the site was repaired, keeping cattle out of the park. At North Fork Virgin River, E. coli exceedances have been less frequent since the State of Utah worked with landowners and grazing permittees to modify agricultural practices. Continued coordination between the National Park Service, state agencies, and local landowners will be necessary to further re-duce E. coli exceedances and, in turn, improve public health and safety in these streams. Selenium concentrations in Red Rock Canyon (BLCA) continued to exceed the state aquat-ic-life standard at both the upstream and downstream sites. Although selenium weathers naturally from bedrock and...

1.1 Resumen macroeconómico En el segundo trimestre la economía enfrentó varios choques, principalmente de oferta y de costos, la mayoría de los cuales no fueron anticipados, o los previstos fueron más persistentes de lo esperado, y que en conjunto interrumpieron la recuperación de la actividad económica observada a comienzos de año y llevaron la inflación total a niveles superiores a la meta. La inflación básica (sin alimentos ni regulados: SAR) aumentó, pero se mantuvo baja y acorde con lo esperado por el equipo técnico. A comienzos de abril se inició una tercera ola de pandemia, más acentuada y prolongada que la anterior, con un elevado costo en vidas humanas y algún impacto negativo en la recuperación económica. Entre mayo y mediados de junio los bloqueos de las carreteras y los problemas de orden público tuvieron un fuerte efecto negativo sobre la actividad económica y la inflación. Se estima que la magnitud de estos dos choques combinados habría generado una caída en niveles en el producto interno bruto (PIB) con respecto al primer trimestre del año. Adicionalmente, los bloqueos causaron un aumento significativo de los precios de los alimentos. A estos choques se sumaron los efectos acumulados de la disrupción global en algunas cadenas de valor y el incremento en los fletes internacionales que desde finales de 2020 vienen generando restricciones de oferta y aumentos de costos. Todos estos factores, que afectaron principalmente el índice de precios al consumidor (IPC) de bienes y de alimentos, explicaron la mayor parte del error de pronóstico del equipo técnico y el aumento de la inflación total a niveles superiores a la meta del 3 %. El incremento en la inflación básica y de los precios de los regulados fue acorde con lo esperado por el equipo técnico, y se explica principalmente por la eliminación de varios alivios de precios otorgados un año atrás. A todo esto se suma la mayor percepción de riesgo soberano y las presiones al alza que esto implica sobre el costo de financiamiento externo y la tasa de cambio. A pesar de los fuertes choques negativos, el crecimiento económico esperado para la primera mitad del año (9,1%), es significativamente mayor que lo proyectado en el informe de abril (7,1%), signo de una economía más dinámica que se recuperaría más rápido de lo previsto. Desde finales de 2020 las diferentes cifras de actividad económica han mostrado un crecimiento mayor que el esperado. Esto sugiere que los efectos negativos sobre el producto de las recurrentes olas de contagio estarían siendo cada vez menos fuertes y duraderos. No obstante, la tercera ola de contagio del Covid-19, y en mayor medida los bloqueos a las vías y los problemas de orden público, habrían generado una caída del PIB durante el segundo trimestre, frente al primero. Pese a lo anterior, los datos del índice de seguimiento a la economía (ISE) de abril y mayo han resultado mayores que lo esperado, y las nuevas cifras de actividad económica sectoriales sugieren que el impacto negativo de la pandemia sobre el producto se sigue moderando, en un entorno de menores restricciones a la movilidad y de mayor avance en el ritmo de vacunación. Los registros de transporte de carga (junio) y la demanda de energía no regulada (julio), entre otros, indican una recuperación importante después de los bloqueos en mayo. Con todo lo anterior, el incremento anual del PIB del segundo trimestre se habría situado alrededor del 17,3 % (antes 15,8 %), explicado en gran parte por una base baja de comparación. Para todo 2021 el equipo técnico incrementó su proyección de crecimiento desde un 6 % hasta el 7,5 %. Este pronóstico, que está rodeado de una incertidumbre inusualmente elevada, supone que no se presentarán problemas de orden público y que posibles nuevas olas de contagio del Covid-19 no tendrán efectos negativos adicionales sobre la actividad económica. Frente al pronóstico del informe pasado, la recuperación de la demanda externa, los niveles de precios de algunos bienes básicos que exporta el país y la dinámica de las remesas de trabajadores han sido mejores que las esperadas y seguirían impulsando la recuperación del ingreso nacional en lo que resta del año. A esto se sumaría la aún amplia liquidez internacional, la aceleración en el proceso de vacunación y las bajas tasas de interés, factores que continuarían favoreciendo la actividad económica. La mejor dinámica del primer semestre, que llevó a una revisión al alza en el crecimiento de todos los componentes del gasto, continuaría hacia adelante y, antes de lo esperado en abril, la economía recuperaría los niveles de producción de 2019 a finales de 2021. El pronóstico continúa incluyendo efectos de corto plazo sobre la demanda agregada de una reforma tributaria de magnitud similar a la proyectada por el Gobierno. Con todo eso, en el escenario central de este informe, el pronóstico de crecimiento para 2021 es del 7,5 % y para 2022 del 3,1 %. A pesar de esto, el nivel de la actividad económica seguiría siendo inferior a su potencial. La mejora en estas proyecciones, sin embargo, está rodeada de una alta incertidumbre. En junio la inflación anual (3,63 %) aumentó más de lo esperado debido al comportamiento del grupo de alimentos, mientras que la inflación básica (1,87 %) fue similar a la proyectada. En lo que resta del año el mayor nivel del IPC de alimentos persistiría y contribuiría a mantener la inflación por encima de la meta. A finales de 2022 la inflación total y básica retornarían a tasas cercanas al 3 %, en un entorno de desaceleración del IPC de alimentos y de menores excesos de capacidad productiva. En los meses recientes el aumento en los precios internacionales de los fletes y de los bienes agrícolas, y las mayores exportaciones de carne y el ciclo ganadero han ejercido presiones al alza sobre el precio de los alimentos, principalmente de los procesados. A estas fuerzas persistentes se sumaron los bloqueos de las vías nacionales y los problemas de orden público en varias ciudades registrados en mayo y parte de junio, los cuales se reflejaron en una fuerte restricción en la oferta y en un aumento anual no esperado del IPC de alimentos (8,52 %). El grupo de regulados (5,93 %) también se aceleró, debido a la baja base de comparación en los precios de la gasolina y a la disolución de parte de los alivios a las tarifas de servicios públicos otorgados en 2020. Como se proyectaba, la inflación SAR repuntó al 1,87 %, debido a la reactivación de los impuestos indirectos de algunos bienes y servicios eliminados un año atrás, y por las presiones al alza que ejercieron los alimentos sobre las comidas fuera del hogar (CFH), entre otros. En lo que resta del año se espera que el aumento en los alimentos perecederos se revierta, siempre y cuando no se registren nuevos bloqueos duraderos a las vías nacionales. El mayor nivel de precios de los alimentos procesados persistiría y contribuiría a mantener la inflación por encima de la meta a finales de año. La inflación SAR continuaría con una tendencia creciente, en la medida en que los excesos de capacidad productiva se sigan cerrando y registraría un aumento transitorio en marzo de 2022, debido principalmente al restablecimiento del impuesto al consumo en las CFH. Con todo esto, para finales de 2021 y 2022 se estima una inflación total del 4,1 % y 3,1 %, y una inflación básica del 2,6 % y 3,2 %, respectivamente. El comportamiento conjunto de los precios del IPC SAR, junto con continuas sorpresas al alza en la actividad económica, son interpretados por el equipo técnico como señales de amplios excesos de capacidad productiva de la economía. Estos persistirían en los siguientes dos años, al final de los cuales la brecha del producto se cerraría. El mayor crecimiento económico sugiere una brecha del producto menos negativa que la estimada hace un trimestre. Sin embargo, el comportamiento de la inflación básica, especialmente en servicios, indica que el PIB potencial se ha recuperado de forma sorpresiva y que los excesos de capacidad siguen siendo amplios, con una demanda agregada afectada de forma persistente. Esta interpretación encuentra soporte en el mercado laboral, en donde persiste un desempleo alto y la recuperación de los empleos perdidos se estancó. Adicionalmente, los aumentos en la inflación en buena medida están explicados por choques de oferta y de costos y por la disolución de algunos alivios de precios otorgados un año atrás. Los pronósticos de crecimiento y de inflación descritos son coherentes con una brecha del producto que se cierra más rápido y es menos negativa en todo el horizonte de pronóstico con respecto al informe de abril. No obstante, la incertidumbre sobre los excesos de capacidad es muy alta y es un riesgo sobre el pronóstico. Las perspectivas de las cuentas fiscales de Colombia se deterioraron, Standard & Poor’s Global Ratings (S&P) y Fitch Ratings (Fitch) redujeron su calificación crediticia, los bloqueos y problemas de orden público afectaron el producto y el país enfrentó una nueva ola de contagios de Covid-19 más acentuada y prolongada que las pasadas. Todo lo anterior se ha reflejado en un aumento de las primas de riesgo y en una depreciación del peso frente al dólar. Esto ha ocurrido en un entorno favorable de ingresos externos. Los precios internacionales del petróleo, del café y de otros bienes básicos que exporta el país aumentaron y han contribuido a la recuperación de los términos de intercambio y del ingreso nacional, y han mitigado las presiones al alza sobre las primas de riesgo y la tasa de cambio. En el presente informe se incrementó el precio esperado del petróleo para 2021 a USD 68 por barril (antes USD 61 bl) y para 2022 a USD 66 bl (antes USD 60 bl). Esta mayor senda presenta una convergencia hacia precios menores que los observados recientemente, como resultado de una mayor oferta mundial esperada de petróleo, la cual más que compensaría el incremento en la demanda de este bien básico. Por ende, se supone que el aumento reciente de los precios tiene un carácter transitorio. En el escenario macroeconómico actual se espera que las condiciones financieras internacionales sean algo menos favorables, a pesar de la mejora en los ingresos externos por cuenta de una mayor demanda y unos precios del petróleo y de otros productos de exportación más altos. Frente al informe de abril el crecimiento de la demanda externa fue mejor que el esperado, y las proyecciones para 2021 y 2022 aumentaron del 5,2 % al 6,0 % y del 3,4 % al 3,5 %, respectivamente. En lo corrido del año las cifras de actividad económica muestran una demanda externa más dinámica de la esperada. En los Estados Unidos y China la recuperación del producto ha sido más rápida que la registrada en los países de la región. En estos últimos la reactivación económica ha estado limitada por los rebrotes del Covid-19, las limitaciones en la oferta de vacunas y el poco espacio fiscal para enfrentar la pandemia, entre otros factores. La buena dinámica en el comercio externo de bienes se ha dado en un entorno de deterioro en las cadenas de valor y de un aumento importante en los precios de las materias primas y en el costo de los fletes. En los Estados Unidos la inflación sorprendió al alza y su valor observado y esperado se mantiene por encima de la meta, al tiempo que se incrementó la proyección de crecimiento económico. Con esto, el inicio de la normalización de la política monetaria en ese país se daría antes de lo proyectado. En este informe se estima que el primer incremento en la tasa de interés de la Reserva Federal de los Estados Unidos se dé a finales de 2022 (antes del primer trimestre de 2023). Para Colombia se supone una mayor prima de riesgo frente al informe de abril y se sigue esperando que presente una tendencia creciente, dada la acumulación de deuda pública y externa del país. Todo esto contribuiría a un incremento en el costo del financiamiento externo en el horizonte de pronóstico. La postura expansiva de la política monetaria sigue soportando unas condiciones financieras internas favorables. En el segundo trimestre la tasa de interés interbancaria y el índice bancario de referencia (IBR) se han mantenido acordes con la tasa de interés de política. Las tasas de interés promedio de captación y crédito continuaron históricamente bajas, a pesar de algunos incrementos observados a finales de junio. La cartera en moneda nacional detuvo su desaceleración anual y, entre marzo y junio, el crédito a los hogares se aceleró, principalmente para compra de vivienda. La recuperación de la cartera comercial y de los desembolsos a ese sector fue importante, y se alcanzó de nuevo el elevado saldo observado un año atrás, cuando las empresas requirieron niveles significativos de liquidez para enfrentar los efectos económicos de la pandemia. El riesgo de crédito aumentó, las provisiones se mantienes altas y algunos bancos han retirado de su balance una parte de su cartera vencida. No obstante, las utilidades del sistema financiero se han recuperado y sus niveles de liquidez y solvencia se mantienen por encima del mínimo regulatorio. A partir de este informe se implementará una nueva metodología para cuantificar y comunicar la incertidumbre que rodea los pronósticos del escenario macroeconómico central, en un entorno de política monetaria activa. Esta metodología se conoce como densidades predictivas (DP) y se explica en detalle en el Recuadro 1. Partiendo del balance de riesgos que contiene los principales factores que, de acuerdo con el juicio del equipo técnico, podrían afectar a la economía en el horizonte de pronóstico, la metodología DP produce distribuciones de probabilidad sobre el pronóstico de las principales variables (v. g.: crecimiento, inflación). Estas distribuciones reflejan el resultado de los posibles choques (a variables externas, precios y actividad económica) que podría recibir la economía y su transmisión, considerando la estructura económica y la respuesta de política monetaria en el futuro. En este sentido, permiten cuantificar la incertidumbre alrededor del pronóstico y su sesgo. El ejercicio DP muestra un sesgo a la baja en el crecimiento económico y en la brecha del producto, y al alza en la inflación. El balance de riesgos indica que las disyuntivas para la política monetaria serán potencialmente más complejas que lo contemplado en el pasado. Por el lado de las condiciones de financiamiento externo, se considera que el mayor riesgo es que se tornen un poco menos favorables, en un escenario en el cual la Reserva Federal de los Estados Unidos incremente con mayor prontitud su tasa de interés. Esto último, ante un crecimiento económico y del empleo mayor que el esperado en los Estados Unidos que genere presiones significativas sobre la inflación de ese país. A esto se suma la incertidumbre sobre el panorama fiscal en Colombia y sus efectos sobre la prima de riesgo y el costo del financiamiento externo. En el caso del crecimiento, la mayoría de los riesgos son a la baja, destacándose los efectos de la incertidumbre política y fiscal sobre las decisiones de consumo e inversión, la aparición de nuevas olas de contagio de la pandemia del Covid-19 y sus impactos sobre la actividad económica. En el caso de la inflación, se incorporó el riesgo de una mayor persistencia de los choques asociados con la disrupción de las cadenas de valor, mayores precios internacionales de las materias primas y de los alimentos, y una recuperación más lenta que la esperada de la cadena agrícola nacional afectada por los pasados bloqueos a las vías. Estos riesgos presionarían al alza principalmente los precios de los alimentos y de los bienes. Como principal riesgo a la baja se incluyó un alza de los arriendos menor que el esperado en el escenario central, explicada por una demanda débil y por una mayor oferta en 2022 dadas las altas ventas de vivienda observadas en el presente año. Con todo, el crecimiento económico presenta un sesgo a la baja y, con el 90 % de confianza, se encontraría entre un 6,1 % y 9,1 % para 2021 y entre el 0,5 % y 4,1 % para 2022. La brecha del producto tendría un sesgo a la baja, principalmente en 2022. El sesgo de la inflación es al alza, y se encontraría entre el 3,7 % y 4,9 % en 2021, y el 2,2 % y 4,7 % en 2022, con un 90 % de probabilidad. 1.2 Decisión de política monetaria En las reuniones de junio y julio la JDBR decidió mantener la tasa de política monetaria inalterada en 1,75 %.