Search

This chapter focuses on the inner rationale and consequences of four different archetypal positions regarding how ethical and political values are tackled in welfare economics. Welfare economics is standardly associated with the welfarist framework, for which social welfare is based on individual utility only. Beyond this, we distinguish the value-neutrality claim – for which ethical values should be and are out of the scope of welfare economics –, the value confinement ideal – for which ethical values are acceptable if they are minimal and consensual–, the transparency requirement – for which any ethical values may be acceptable in the welfare economics framework if explicit and formalized –, and the entanglement claim – which challenges the very possibility of demarcation between facts and values.

Index de les obres ressenyades: Ignacio FARÍAS ; José OSSANDÓN, Observando sistemas. Nuevas apropiaciones y usos de la teoría de Niklas Luhmann

We introduce you to TensorFlow input pipelines with the tf.data API, which enables you to build complex input pipelines from simple, reusable pieces. Input pipelines are the lifeblood of any deep learning experiment because learning models expect data in a TensorFlow consumable form. It is very easy to create high-performance pipelines with the tf.data.Dataset abstraction (a component of the tf.data API) because it represents a sequence of elements from a dataset in a simple format.

We introduce TensorFlow datasets (TFDS). We discuss many facets of a TFDS with code examples. We continue with a complete TFDS modeling example.

In the previous chapter, we demonstrated how to work with TFDS objects. In this chapter, we work through two end-to-end deep learning experiments with large and complex TFDS objects. The Fashion-MNIST and beans datasets are small with simple images.

Feedforward neural nets are generally great for classification and regression problems. CNNs are great for complex image classification. But activations for feedforward nets and CNNs flow only in one direction, from the input layers to the output layer. Since signals flow in only one direction, feedforward and convolutional nets are not ideal if patterns in data change over time. So we need a different network architecture to work with data impacted by time.

We introduce advanced transfer learning with code examples based on several transfer learning architectures. The code examples train learning models with these architectures.

We’ve already leveraged RNNs for NLP. In this chapter, we create experiments to forecast with time series data. We use the famous Weather dataset to demonstrate both a univariate and a multivariate example.

Retirement system sustainability is defined as the ability of plan boards and managers to be responsible investors, active stewards, and allocators of capital to economic activities with desirable social and environmental outcomes. In this paper, we examine the policy frameworks and important structural variables pertinent to private retirement systems in Australia, the UK, and the US. By analyzing various reports, interviewing experts, and using data from the Principles of Responsible Investment as well as national pension and retirement authorities, we identify key structural challenges within national retirement systems. These include market fragmentation, principal-agent conflicts in personal pensions, and the role of service providers. Our results provide insight into how, or whether, retirement systems can facilitate desirable economic, social, and environmental outcomes.

I study criminal street gangs using new data that describes the geospatial distribution of gang territory in Chicago and its evolution over a 15-year period. Using an event study design, I show that city blocks entered by gangs experience sharp increases in the number of reported batteries (6%), narcotics violations (18.5%), weapons violations (9.8%), incidents of prostitution (51.9%), and criminal trespassing (19.6%). I also find a sharp reduction in the number of reported robberies (-8%). The findings cannot be explained by pre-existing trends in crime, changes in police surveillance, crime displacement, exposure to public housing demolitions, reporting effects, or demographic trends. Taken together, the evidence suggests that gangs cause small increases in violence in highly localized areas as a result of conflict over illegal markets. I also find evidence that gangs cause reductions in median property values (-$8,436.9) and household income (-$1,866.8). Motivated by these findings, I explore the relationship between the industrial organization of the black market and the supply of criminal activity. I find that gangs that are more internally fractured or operate in more competitive environments tend to generate more crime. This finding is inconsistent with simple, market-based models of criminal behavior, suggesting an important role for behavioral factors and social interactions in the production of gang violence.

We guide you in the creation of augmented data experiments to increase the diversity of a training set by applying random (but realistic) transformations. Data augmentation is very useful for small datasets because deep learning models crave a lot of data to perform well.

Official foreign direct investment (FDI) statistics track FDI via direct investing nation rather than ultimate investing country (UIC). If a FDI project is sourced through an overseas subsidiary, normal FDI metrics will register that investment as being from the third country, not the home country of the multinational. There is a growing awareness that the large amount of FDI channeled through special purpose vehicles and offshore financial centers therefore makes official statistics unreliable: investment from certain nations may be significantly under or over counted. According to official statistics, the United States ranks 11th in terms of total amount of foreign direct investment (FDI) in Vietnam. As Vietnam joins the ranks of lower middle-income countries, and becomes more integrated into global value chains, a more representative methodology for calculating FDI would help Vietnam thoughtfully engage with foreign investors and support Vietnam's development objectives. This report, commissioned by the United States Agency for International Development (USAID) in Hanoi includes: 1) A literature review of FDI databases and FDI tracking methods, 2) a comparative analysis of the existing methods for tracking FDI via UIC, 3) a discussion on the feasibility of implementing those methods for Vietnam, 4) a novel method for estimating Vietnam’s inward FDI via UIC, 6) recommendations to USAID for further research and analysis.

In the previous chapter, we showed you how to work with a TFDS. But what if we want to work with another type of dataset? In this chapter, we show you how to work with other types of data with TensorFlow.

Between 1909 and 1922 a private deposit insurance company coexisted with the state-sponsored deposit insurance program in Kansas. This paper documents its development using primary sources. In addition, it examines if affiliation with the private deposit insurance (i) had an effect on risk-taking and the probability of failure; (ii) increased confidence among depositors, and (iii) was influenced by a neighboring bank’s membership in the state’s deposit insurance. We find that affiliation with the private deposit insurance did not affect a bank’s likelihood of failure, although smaller national bank members did increase risk-taking. The evidence does not support the hypothesis that the company enhanced depositor confidence. Lastly, we do find strong evidence that a bank’s decision to join the private deposit insurance was influenced by neighboring banks’ affiliation with the Kansas deposit insurance program.

GANs are effective at generating crisp synthetic images, but are limited in size to about 64 × 64 pixels. A Progressive Growing GAN is an extension of the GAN that enables training generator models to generate large high-quality images up to about 1024 × 1024 pixels (as of this writing). The approach has proven effective at generating high-quality synthetic faces that are startlingly realistic.

We study the reaction of investors to tail events across trading institutions. We conduct experiments in which investors bid on a financial asset that delivers a small positive reward in more than 99% of the cases and a large loss otherwise. The baseline treatment uses a repeated BDM mechanism whereas the market treatment replaces the uniform draw of the BDM mechanism by a uniform draw over the bids of the other participants. Our design is such that bids should not differ across treatments in normal times while allowing for potential differences to emerge after tail events have occurred. We find that markets tend to exacerbate the reaction of investors to tail losses and we attribute this effect to emotions.

Reinforcement learning (RL) is an area of machine learning that focuses on teaching intelligent agents how to take actions in an environment in order to maximize cumulative reward. Cumulative reward in RL is the sum of all rewards as a function of the number of training steps.

This paper studies the spillover effect of home mortgage delinquencies using a discrete-choice spatial network model. In our empirical study, a main challenge in estimating this model is that mortgage repayment decisions can only be observed for a sample of all the borrowers in the study region. We show that the nested pseudolikelihood (NPL) algorithm can be readily modified to accommodate this missing data issue. Monte Carlo simulations indicate that the proposed estimator works well in finite samples and ignoring this issue leads to a downward bias in the estimated spillover effect. We estimate the model using data on single-family residential mortgage delinquencies in Clark County of Nevada in 2010, and find strong evidence of spillover effects. We also conduct some counterfactual experiments to illustrate the policy relevance of the spillover effect.

Autoencoders don’t typically work well with images unless they are very small. But convolutional and variational autoencoders work much better than feedforward dense ones with large color images.

With feedforward neural networks, we achieved good training performance with MNIST and Fashion-MNIST datasets. But images in these datasets are simple and centered within the input space that contains them. That is, they are centered within the pixel matrix that holds them. Input space is all the possible inputs to a model.

We work through a complete deep learning example with Python’s TensorFlow 2.x library in the Google Colab cloud service. We also demonstrate how to link your Google Drive with the Colab cloud service.

Technology can help to reduce inefficiencies and inequalities in developing countries. Although whether this happens in practice often depends on the social and geographic context. Deployment of solar home systems (SHS) across Nepal provides one such example. We examine a policy to promote SHS through a subsidy for adoption and its downstream benefits - children’s education, time allocated to agricultural and household work (both unpaid), and working for a wage. We overcome the challenges of the non-random assignment of subsidies by using geographic regression discontinuity design using (cost) distance as the assignment variable. This is based on the assumption that households cannot perfectly control the assignment variable because migrating to a different region, simply to acquire an additional solar subsidy of maximum NRs4000, is challenging and unlikely in the case of Nepal. Our results show that subsidy-eligible households are about 43% more likely to adopt SHS. We find a positive effect of the SHS adoption on the grade for age and educational z-score of children’s educations with the effect being higher on girls than boys. We also find that females are 14.2% more likely to participate and spend more time (about 19 extra days in a year) in agricultural practices by reducing their engagement in family work. However, there is no effect on the participation rates across the labour market. Results also show that males increase their contribution to household work.

Transfer learning is the process of creating new learning models by fine-tuning previously trained neural networks. Instead of training a network from scratch, we download a pre-trained open source learning model and fine-tune it for our own purpose. A pre-trained model is one that is created by someone else to solve a similar problem. We can use one of these instead of building our own model. A big advantage is that a pre-trained model has been crafted by experts, so we can be confident that it performs at a high level (in most cases). Another advantage is that we don’t have to have a lot of data to use a pre-trained model.

Generative modeling is an unsupervised learning technique that involves automatically discovering and learning the regularities (or patterns) in input data so that a trained model can generate new examples that plausibly could have been drawn from the original dataset. A popular type of generative model is a generative adversarial network. Generative adversarial networks (GANs) are generative models that create new data instances that resemble the training data.

We introduce you to Tensor Processing Units with code examples. A Tensor Processing Unit (TPU) is an application-specific integrated circuit (ASIC) designed to accelerate ML workloads. The TPUs available in TensorFlow are custom-developed from the ground up by the Google Brain team based on its plethora of experience and leadership in the ML community. Google Brain is a deep learning artificial intelligence (AI) research team at Google who research ways to make machines intelligent for the improvement of people’s lives.

This essay is focused on the story of an old growth forest preserve in Cornwall, Connecticut known as the Cathedral Pines. The author, who grew up beside the forest, looks at the lifecycle of the forest as a case study in settler constructions of forests, conservation, and the American landscape. The author uses archival research and outreach to local indigenous and environmental activists to explore the social and ecological processes that have shaped the Cathedral Pines, within broader regional histories. The text looks at the political economic history and symbolism attached to white pine trees to explore the impacts of racial colonialism on the present ecology of forests within the Housatonic and Hudson River valley regions. The author argues that settler mythologies of wilderness continue to structure discourses of conservation. The concluding sections of the text draw on decolonial discourses of “reciprocity” to think about the potentialities embedded in efforts to align forest restoration efforts with indigenous-led environmental justice movements.

N/A

We present the Phase A Science Case for the Multi-conjugate Adaptive-optics Visible Imager-Spectrograph (MAVIS), planned for the Adaptive Optics Facility (AOF) of the Very Large Telescope (VLT). MAVIS is a general-purpose instrument for exploiting the highest possible angular resolution of any single optical telescope available in the next decade, either on Earth or in space, and with sensitivity comparable to (or better than) larger aperture facilities. MAVIS uses two deformable mirrors in addition to the deformable secondary mirror of the AOF, providing a mean V-band Strehl ratio of >10% (goal >15%) across a relatively large (30 arc second) science field. This equates to a resolution of, 141 pages, 79 figures, 10Mb. Modified version of the Phase A Science Case submitted to ESO in April 2020 as part of the MAVIS Phase A design study. Full resolution version available upon request

This paper provides invariance theorems that facilitate testing for the existence of an asset price bubble in a market where the price evolves as a Markov diffusion process. The test involves only the properties of the price process’ quadratic variation under the statistical probability. It does not require an estimate of either the equivalent local martingale measure or the asset’s drift. To augment its use, a new family of stochastic volatility price processes is also provided where the processes’ strict local martingale behavior can be characterized.