Homo economicus

From Wikipedia, the free encyclopedia

https://en.wikipedia.org/wiki/Homo_economicus

The term Homo economicus, or economic man, is the portrayal of humans as agents who are consistently rational and narrowly self-interested, and who pursue their subjectively defined ends optimally. It is a word play on Homo sapiens, used in some economic theories and in pedagogy.

In game theory, Homo economicus is often modelled through the assumption of perfect rationality. It assumes that agents always act in a way that maximize utility as a consumer and profit as a producer, and are capable of arbitrarily complex deductions towards that end. They will always be capable of thinking through all possible outcomes and choosing that course of action which will result in the best possible result.

The rationality implied in Homo economicus does not restrict what sort of preferences are admissible. Only naive applications of the Homo economicus model assume that agents know what is best for their long-term physical and mental health. For example, an agent's utility function could be linked to the perceived utility of other agents (such as one's husband or children), making Homo economicus compatible with other models such as Homo reciprocans, which emphasizes human cooperation.

As a theory on human conduct, it contrasts to the concepts of behavioral economics, which examines cognitive biases and other irrationalities, and to bounded rationality, which assumes that practical elements such as cognitive and time limitations restrict the rationality of agents.

History of the term

The term "economic man" was used for the first time in the late nineteenth century by critics of John Stuart Mill's work on political economy. Below is a passage from Mill's work that critics referred to:

[Political economy] does not treat the whole of man's nature as modified by the social state, nor of the whole conduct of man in society. It is concerned with him solely as a being who desires to possess wealth, and who is capable of judging the comparative efficacy of means for obtaining that end.

Later in the same work, Mill stated that he was proposing "an arbitrary definition of man, as a being who inevitably does that by which he may obtain the greatest amount of necessaries, conveniences, and luxuries, with the smallest quantity of labour and physical self-denial with which they can be obtained."

Adam Smith, in The Theory of Moral Sentiments, had claimed that individuals have sympathy for the well-being of others. On the other hand, in The Wealth of Nations, Smith wrote:

It is not from the benevolence of the butcher, the brewer, or the baker that we expect our dinner, but from their regard to their own interest.

This comment is perfectly in line with the notion of Homo economicus and the idea, propounded by Smith in The Wealth of Nations and, in the 20th century, by the likes of Ayn Rand (in The Virtue of Selfishness, for example), that pursuing narrow, individual self-interest promotes social well-being. In Book V, Chapter I, Smith argues, "The man whose whole life is spent in performing a few simple operations, of which the effects are perhaps always the same, or very nearly the same, has no occasion to exert his understanding or to exercise his invention in finding out expedients for removing difficulties which never occur. He naturally loses, therefore, the habit of such exertion, and generally becomes as stupid and ignorant as it is possible for a human creature to become." This could be seen as prefiguring one part of Marx's theory of alienation of labor; and also as a pro-worker argument against the division of labor and the restrictions it places upon freedom of occupation. But even so, taken in the context of the work as a whole, Smith clearly intends it in a pro-capitalism, pro-bourgeoisie, way: "removing difficulties", such as reducing the time needed for travel and trade, through "expedients", such as steam-engine ships, here means the typical argument that capitalism brings freedom of entrepreneurship and innovation, which then bring prosperity. Thus, Smith is not unreasonably called "The Father of Capitalism"; early on, he theorized many of today's most widespread and deep-seated pro-capitalism arguments.

The early role of Homo Economicus within neoclassical theory was summarised to include a general objective of discovering laws and principles to accelerate further growth within the national economy and the welfare of ordinary citizens. These laws and principles were determined by two governing factors, natural and social. It had been found to be the foundation of neoclassical theory of the firm which assumed that individual agents would act rationally amongst other rational individuals. In which Adam Smith explains that the actions of those that are rational and self-interested under homo economicus promotes the general good overall which was understood as the efficient allocation of material wealth. However, social scientists had doubted the actual importance of income and wealth to overall happiness in societies.

The term 'Homo economicus' was initially critiqued for its portrayal of the economic agent as a narrowly defined, money-making animal, a characterization heavily influenced by the works of Adam Smith and John Stuart Mill. Authors from the English Historical School of Economics sought to demote this model from its broad classification under the 'genus homo', arguing that it insufficiently captured the complex ethical and behavioral dimensions of human decision-making. Their critique emphasized the need for a more nuanced understanding of human agency beyond the mere pursuit of economic rationality.

Economists in the late 19th century—such as Francis Edgeworth, William Stanley Jevons, Léon Walras, and Vilfredo Pareto—built mathematical models on these economic assumptions. In the 20th century, the rational choice theory of Lionel Robbins came to dominate mainstream economics. The term "economic man" then took on a more specific meaning: a person who acted rationally on complete knowledge out of self-interest and the desire for wealth.

Model

Homo economicus is a term used for an approximation or model of Homo sapiens that acts to obtain the highest possible well-being for themself given available information about opportunities and other constraints, both natural and institutional, on their ability to achieve their predetermined goals. This approach has been formalized in certain social sciences models, particularly in economics.

The model of the homo economicus is currently the most widespread model of human behaviour in economics. There are still limitations in developing this model through the years of social development. Human nature is complex and full of contradictions. People can do deeds that are full of justice or they can do deeds that are annoying. In the vast majority of cases people are selfish and unselfish, depending on various natures. The Homo economicus model is usually based on pure self-interest in order to activate cooperation and thus contribute to society. Within economic social systems, humans are usually rational and selfish maximising personal preferences, and this form of model is also known as Homo economicus. within the homo economicus model, the principle of rationality and selfishness is well chosen. Given the same conditions, people only want to get more, not less. This is why managers can increase productivity through incentive policies.

Homo economicus is usually seen as "rational" in the sense that well-being as defined by the utility function is optimized given perceived opportunities. That is, the individual seeks to attain very specific and predetermined goals to the greatest extent with the least possible cost. Note that this kind of "rationality" does not say that the individual's actual goals are "rational" in some larger ethical, social, or human sense, only that they try to attain them at minimal cost. Only naïve applications of the Homo economicus model assume that this hypothetical individual knows what is best for their long-term physical and mental health and can be relied upon to always make the right decision for themself. See rational choice theory and rational expectations for further discussion; the article on rationality widens the discussion.

As in social science, these assumptions are at best approximations. The term is often used derogatorily in academic literature, perhaps most commonly by sociologists, many of whom tend to prefer structural explanations to ones based on rational action by individuals.

The use of the Latin form Homo economicus is certainly long established; Persky traces it back to Pareto (1906) but notes that it may be older. The English term economic man can be found even earlier, in John Kells Ingram's A History of Political Economy (1888). The Oxford English Dictionary (O.E.D.) cites the use of Homo oeconomicus by C. S. Devas in his 1883 work The Groundwork of Economics in reference to Mill's writings, as one of a number of phrases that imitate the scientific name for the human species:

Mill has only examined the Homo oeconomicus, or dollar-hunting animal.

According to the OED, the human genus name Homo is

Used with L. or mock-L. adjs. in names imitating Homo sapiens, etc., and intended to personify some aspect of human life or behaviour (indicated by the adj.). Homo faber ("feIb@(r)) [H. Bergson L'Evolution Créatrice (1907) ii. 151], a term used to designate man as a maker of tools.) Variants are often comic: Homo insipiens; Homo turisticus.

Note that such forms should logically keep the capital for the "genus" name—i.e., Homo economicus rather than homo economicus. Actual usage is inconsistent.

Amartya Sen has argued there are grave pitfalls in assuming that rationality is limited to selfish rationality. Economics should build into its assumptions the notion that people can give credible commitments to a course of conduct. He demonstrates the absurdity with the narrowness of the assumptions by some economists with the following example of two strangers meeting on a street.

"Where is the railway station?" he asks me. "There," I say, pointing at the post office, "and would you please post this letter for me on the way?" "Yes," he says, determined to open the envelope and check whether it contains something valuable.

Criticisms

Homo economicus bases its choices on a consideration of its own personal "utility function".

In recent times, few concepts have been as recognisable and accepted as the concept of the homo economicus. The system established by this concept has therefore almost become the basis for the concepts currently used in economics. As society develops and the modern economy evolves, will people follow the concept of the homo economicus."Self-interest is the main motivation of human beings in their transactions" is a theoretical structure in the concept of homo economicus. Over the years, economists have studied and discussed institutional economics, behavioural economics, political economy, economic anthropology and ecological economics. The economic man solution is considered to be inadequate and flawed.

Consequently, the Homo economicus assumptions have been criticized not only by economists on the basis of logical arguments, but also on empirical grounds by cross-cultural comparison. Economic anthropologists such as Marshall Sahlins, Karl Polanyi, Marcel Mauss and Maurice Godelier have demonstrated that in traditional societies, choices people make regarding production and exchange of goods follow patterns of reciprocity which differ sharply from what the Homo economicus model postulates. Such systems have been termed gift economy rather than market economy. Criticisms of the Homo economicus model put forward from the standpoint of ethics usually refer to this traditional ethic of kinship-based reciprocity that held together traditional societies. Philosophers Amartya Sen and Axel Honneth are noted for their criticisms of the normative assumptions made by the self-interested utility function.

Economists Thorstein Veblen, John Maynard Keynes, Herbert A. Simon, and many of the Austrian School criticise Homo economicus as an actor with too great an understanding of macroeconomics and economic forecasting in his decision making. They stress uncertainty and bounded rationality in the making of economic decisions, rather than relying on the rational man who is fully informed of all circumstances impinging on his decisions. They argue that perfect knowledge never exists, which means that all economic activity implies risk. Austrian economists rather prefer to use as a model tool the Homo agens.

Empirical studies by Amos Tversky questioned the assumption that investors are rational. In 1995, Tversky demonstrated the tendency of investors to make risk-averse choices in gains, and risk-seeking choices in losses. The investors appeared as very risk-averse for small losses but indifferent for a small chance of a very large loss. This violates economic rationality as usually understood. Further research on this subject, showing other deviations from conventionally defined economic rationality, is being done in the growing field of experimental or behavioral economics. Some of the broader issues involved in this criticism are studied in decision theory, of which rational choice theory is only a subset.

Behavioral economists Richard Thaler and Daniel Kahneman have criticized the notion of economic agents possessing stable and well-defined preferences that they consistently act upon in a self-interested manner. Using insights from psychological experiments found explanations for anomalies in economic decision-making that seemed to violate rational choice theory. Writing a column in the Journal of Economic Perspectives under the title Anomalies, Thaler wrote features on the many ways observed economic behavior in markets deviated from theory. One such anomaly was the endowment effect by which individual preferences are framed based on reference positions (Kahneman et al., 1990). In an experiment in which one group was given a mug and the other was asked how much they were willing to pay (WTP) for the mug, it was found that the price that those endowed with the mug where willingness to accept (WTA) greatly exceeded that of the WTP. This was seen as falsifying the Coase theorem in which for every person the WTA equals the WTP that is the basis of the efficient-market hypothesis. From this they argued the endowment effect acts on us by making it painful for us to give up the endowment. Kahneman also argued against the rational-agent model in which agents make decisions with all of the relevant context including weighing all possible future opportunities and risks. Evidence supports the claim that decisions are often made by "narrow framing" with investors making portfolio decisions in isolation from their entire portfolio (Nicholas Barberis et al., 2003). Shlomo Benartzi and Thaler found that investors also tended to use unreasonable time periods in evaluating their investments.

In Kahneman-Tversky’s criticism of the Homo Economicus model, many mainstream economists had utilised deductive logic to further progress the Homo Economicus idea as opposed to Daniel Kahneman and Amos Tversky in which they had applied inductive logic. Further findings of their experiments that opposed Homo Economicus had found that individuals will constantly adjust their choices according to changes in their income and market prices. Furthermore, Kahneman and Tversky had conducted experiments exploring prospect theory where results from several experiments concluded that individuals will generally put higher importance on avoiding loss over making a gain.

Other critics of the Homo economicus model of humanity, such as Bruno Frey, point to the excessive emphasis on extrinsic motivation (rewards and punishments from the social environment) as opposed to intrinsic motivation. For example, it is difficult if not impossible to understand how Homo economicus would be a hero in war or would get inherent pleasure from craftsmanship. Frey and others argue that too much emphasis on rewards and punishments can "crowd out" (discourage) intrinsic motivation: paying a boy for doing household tasks may push him from doing those tasks "to help the family" to doing them simply for the reward.

Another weakness is highlighted by economic sociologists and anthropologists, who argue that Homo economicus ignores an extremely important question, i.e. the origins of tastes and the parameters of the utility function by social influences, training, education, and the like. The exogeneity of tastes (preferences) in this model is the major distinction from Homo sociologicus, in which tastes are taken as partially or even totally determined by the societal environment (see below).

Further critics, learning from the broadly defined psychoanalytic tradition, criticize the Homo economicus model as ignoring the inner conflicts that real-world individuals suffer, as between short-term and long-term goals (e.g., eating chocolate cake and losing weight) or between individual goals and societal values. Such conflicts may lead to "irrational" behavior involving inconsistency, psychological paralysis, neurosis, and psychic pain. Further irrational human behaviour can occur as a result of habit, laziness, mimicry and simple obedience.

The emerging science of "neuroeconomics" suggests that there are serious shortcomings in the conventional theories of economic rationality. Rational economic decision making has been shown to produce high levels of cortisol, epinephrine and corticosteroids, associated with elevated levels of stress. It seems that the dopaminic system is only activated upon achieving the reward, and otherwise the "pain" receptors, particularly in the prefrontal cortex of the left hemisphere of the brain show a high level of activation. Serotonin and oxytocin levels are minimised, and the general immune system shows a level of suppression. Such a pattern is associated with a generalised reduction in the levels of trust. Unsolicited "gift giving", considered irrational from the point of view of Homo economicus, by comparison, shows an elevated stimulation of the pleasure circuits of the whole brain, reduction in the levels of stress, optimal functioning of the immune system, reduction in cortico-steroids and epinephrine and cortisol, activation of the substantia nigra, the striatum and the nucleus accumbens (associated with the placebo effect), all associated with the building of social trust. Mirror neurons result in a win-win positive sum game in which the person giving the gift receives a pleasure equivalent to the person receiving it. This confirms the findings of anthropology which suggest that a "gift economy" preceded the more recent market systems where win-lose or risk-avoidance lose-lose calculations apply.

Responses

Some economists disagree with these critiques, arguing that it may be relevant to analyze the consequences of enlightened egoism just as it may be worthwhile to consider altruistic or social behavior. Others argue that we need to understand the consequences of such narrow-minded greed even if only a small percentage of the population embraces such motives. Free riders, for example, would have a major negative impact on the provision of public goods. However, economists' supply and demand predictions might obtain even if only a significant minority of market participants act like Homo economicus. In this view, the assumption of Homo economicus can and should be simply a preliminary step on the road to a more sophisticated model.

Others argue that Homo economicus is a reasonable approximation for behavior within market institutions, since the individualized nature of human action in such social settings encourages individualistic behavior. Not only do market settings encourage the application of a simple cost-benefit calculus by individuals, but they reward and thus attract the more individualistic people. It can be difficult to apply social values (as opposed to following self-interest) in an extremely competitive market; a company that refuses to pollute, for example, may find itself bankrupt.

Defenders of the Homo economicus model see many critics of the dominant school as using a straw man technique. For example, it is common for critics to argue that real people do not have cost-less access to infinite information and an innate ability to instantly process it. However, in advanced-level theoretical economics, scholars have found ways of addressing these problems, modifying models enough to more realistically depict real-life decision-making. For example, models of individual behavior under bounded rationality and of people suffering from envy can be found in the literature. It is primarily when targeting the limiting assumptions made in constructing undergraduate models that the criticisms listed above are valid. These criticisms are especially valid to the extent that the professor asserts that the simplifying assumptions are true or uses them in a propagandistic way.

Perspectives

According to Sergio Caruso, when talking of Homo economicus, one should distinguish between the purely "methodological" versions, aimed at practical use in the economic sphere (e.g. economic calculus), and the" anthropological" versions, more ambitiously aimed at depicting a certain type of man (supposed to be actually existing), or even human nature in general. The former, traditionally founded on a merely speculative psychology, have proved unrealistic and frankly wrong as descriptive models of economic behaviour (therefore not applicable for normative purposes either); however, they are liable to be corrected resorting to the new empirically based economic psychology, which turns quite other than the philosophers' psychology that economists have used until yesterday. Among the latter (i.e. the anthropological versions), one can make a further distinction between the weak versions, more plausible, and the strong ones, irreparably ideological. Depicting different types of "economic man" (each depending on the social context) is in fact possible with the help of cultural anthropology, and social psychology (a branch of psychology economists have strangely ignored), if only those types are contrived as socially and/or historically determined abstractions (such as Weber's, Korsch's, and Fromm's concepts of Idealtypus, "historical specification", and "social character"). Even a Marxist theoretician such as Gramsci—reminds Caruso—admitted of the Homo economicus as a useful abstraction on the ground of economic theory, provided that we grant there be as many homines oeconomici as the modes of production. On the contrary, when one concept of Homo economicus claims to grasp the eternal essence of what is human, at the same time putting aside all other aspects of human nature (such as Homo faber, Homo loquens, Homo ludens, Homo reciprocans, and so on), then the concept leaves the field of good philosophy, not to speak of social science, and is ready to enter a political doctrine as the most dangerous of its ideological ingredients.

Homo sociologicus

Comparisons between economics and sociology have resulted in a corresponding term Homo sociologicus (introduced by German sociologist Ralf Dahrendorf in 1958), to parody the image of human nature given in some sociological models that attempt to limit the social forces that determine individual tastes and social values. (The alternative or additional source of these would be biology.) Hirsch et al. say that Homo sociologicus is largely a tabula rasa upon which societies and cultures write values and goals; unlike economicus, sociologicus acts not to pursue selfish interests but to fulfill social roles (though the fulfillment of social roles may have a selfish rationale—e.g. politicians or socialites). This "individual" may appear to be all society and no individual.

AutoAI

From Wikipedia, the free encyclopedia

https://en.wikipedia.org/wiki/AutoAI 

This article contains content that is written like an advertisement. Please help improve it by removing promotional content and inappropriate external links, and by adding encyclopedic content written from a neutral point of view. (January 2023) (Learn how and when to remove this template message)

Automated Artificial Intelligence (AutoAI) is a variation of the automated machine learning or AutoML technology, which extends the automation of model building towards automation of the full life cycle of a machine learning model. It applies intelligent automation to the task of building predictive machine learning models by preparing data for training and identifying the best type of model for the given data. then choosing the features or columns of data that best support the problem the model is solving. Finally, automation evaluates a variety of tuning options to reach the best result as it generates, then ranks, model-candidate pipelines. The best performing pipelines can be put into production to process new data, and deliver predictions based on the model training. Automated artificial intelligence can also be applied to making sure the model doesn't have inherent bias and automating the tasks for continuous improvement of the model. Managing an AutoAI model requires frequent monitoring and updating, managed by a process known as model operations or ModelOps.

The Automated Machine Learning and Data Science (AMLDS) is a small team within IBM Research, which was formed to apply techniques from artificial intelligence (AI), machine learning (ML) and data management to accelerate and optimize the creation of machine learning and data science workflows. AMLDS gets credit of driving the development of AutoAI.

Use case

A typical use case for AutoAI would be training a model to predict how customers might respond to a sales incentive. The model first gets training with actual data on how customers responded to the promotion. When the trained model presented with new data, can provide a prediction of how a new customer might respond, with a confidence score for the prediction. Prior to AutoML, data scientists had to build these predictive models by hand, testing various combinations of algorithms, then testing to see how predictions compared to actual results, whereas AutoML automated the processes of preparing the data for training, applying algorithms to process the data, and then further optimizing the results. Hence, AutoAI provides greater intelligent automation that allows for testing significantly more combinations of factors to generate model candidate pipelines that reflect and address the problem more accurately. Once built, the model evaluated for bias and updated to improve performance.

The AutoAI process

The user initiates the process by providing a set of training data and identifying the prediction column, which sets up the problem to solve. For example, the prediction column might contain values of yes or no in response to an offered incentive. In the data pre-processing stage, AutoAI applies various algorithms, or estimators, to analyze, clean (for example, remove redundant information or impute missing data), and prepare structured raw data for machine learning (ML).

The next is automated model selection that matches the data with a model type, such as classification or regression. For example, if there are only two types of data in a prediction column, AutoAI prepares to build a binary classification model. If there is an unknowable set of answers, AutoAI prepares a regression model, which employs a distinct set of algorithms, or problem-solving transformations. AutoAI ranks after testing candidate algorithms against small sub-sets of the information, increasing the size of the subset gradually for the algorithms that turns most promising to reach at the best match. This process of iterative and incremental machine learning is what sets AutoAI apart from earlier versions of AutoML.

Feature engineering transforms the raw data into the combination that represents the problem to arrive at the best accurate prediction. Part of this process is to evaluate how data in the training data source can best support an accurate prediction using algorithms, it weights few data more important than others to achieve the desired result. AutoAI automates the consideration of various features construction options in a non-exhaustive, structured manner, meanwhile progressively maximizing the accuracy of model using reinforcement learning. This results from an optimized sequence of information and data transformations that matches the best algorithms of the step involving model selection.

Finally, AutoAI applies the hyperparameter optimization step to refine and advance the best performing model pipelines. Pipelines are model candidates, evaluated and ranked by metrics like accuracy, precision. At the end of the process, the user can review the pipelines and choose the pipeline(s) to put into production to deliver predictions on new data.

History

In August 2017, AMLDS announced that they were researching the use of automated feature engineering to eliminate guesswork in data science. AMDLS members Udayan Khurana, Horst Samulowitz, Gregory Bramble, Deepak Toraga, and Peter Kirchner, along with Fatemeh Nargesian of the University of Toronto and Elias Khalil of Georgia Tech, presented their preliminary research at IJCAI that same year.

Called “Learning-based Feature Engineering,” their method learned the correlations between feature distributions, target distributions, and transformations, built meta-models that used past observations to predict viable transformations, and generalized thousands of data sets spanning different domains. To address feature vectors of varied sizes, it used Quantile Sketch Array to capture the essential character of a feature.

In 2018, IBM Research announced Deep Learning as a Service, which opened popular deep learning libraries such as Caffe, Torch and TensorFlow, to developers in the cloud. Jean-Francois Puget, PhD, a distinguished engineer specializing machine learning (ML) and optimization at IBM, entered the competition. He found out and decided to be ready for IBM AI and data science platforms like IBM Watson. In December 2018, IBM Research announced NeuNetS, a new capability that automated neural network model synthesis as part of automated AI model development and deployment.

In 2020, Liu et al. proposed a method for AutoML that used the alternating direction method of multipliers (ADMM) to configure multiple stages of an ML pipeline, such as transformations, feature engineering and selection, and predictive modeling. This was the first recorded time that IBM Research publicly applied the term “Auto” to machine-learning.

AutoAI: The evolution of AutoML

2019 was the year that AutoML became more widely discussed as a concept. “The Forrester New Wave™: Automation-Focused Machine Learning Solutions, Q2 2019,” evaluated AutoML solutions and found that the more powerful versions offered feature engineering. A Gartner Technical Professional Advice report from August 2019 reported that, based on their research, AutoML could augment data science and machine learning. They described AutoML as the automation of data preparation, feature engineering and model engineering tasks.

AutoAI is the evolution of AutoML. One of AutoAI's principal inventors, Jean-Francois Puget, PhD, describes it as automatically performing data preparation, feature engineering, machine learning algorithm selection, and hyper-parameter optimization to find the best possible machine learning model. The hyper-parameter optimization algorithm used in AutoAI differs from the hyper-parameter tuning of AutoML. The algorithm, optimized for cost function evaluations such as model training and scoring which are typical in machine learning, enabling rapid convergence to a satisfactory solution despite evaluation times of each iteration being of long duration.

Research scientists at IBM Research published a paper "Towards Automating the AI Operations Lifecycle", which describes the advantages and available technologies for automating more of the process, with the goal of limiting the human involvement required to build, test, and maintain a machine learning application. However, some HCI researchers argue that the machine learning application and its recommendations are inevitably taken by human decision makers, thus it is impossible to eliminate human involvement in the process. Rather, a more transparent and interpretable AutoAI design is the key to gain trust from human users, but such design itself is quite a challenge.

Awards for AutoAI

Winner, Best Innovation in Intelligent Automation Award at the AIconics AI Summit (2019), San Francisco.

Winner, iF Design Guide award for Communication in a Software Application (2020)

Automated machine learning

From Wikipedia, the free encyclopedia
https://en.wikipedia.org/wiki/Automated_machine_learning

Automated machine learning (AutoML) is the process of automating the tasks of applying machine learning to real-world problems.

AutoML potentially includes every stage from beginning with a raw dataset to building a machine learning model ready for deployment. AutoML was proposed as an artificial intelligence-based solution to the growing challenge of applying machine learning. The high degree of automation in AutoML aims to allow non-experts to make use of machine learning models and techniques without requiring them to become experts in machine learning. Automating the process of applying machine learning end-to-end additionally offers the advantages of producing simpler solutions, faster creation of those solutions, and models that often outperform hand-designed models.

Common techniques used in AutoML include hyperparameter optimization, meta-learning and neural architecture search.

Comparison to the standard approach

In a typical machine learning application, practitioners have a set of input data points to be used for training. The raw data may not be in a form that all algorithms can be applied to. To make the data amenable for machine learning, an expert may have to apply appropriate data pre-processing, feature engineering, feature extraction, and feature selection methods. After these steps, practitioners must then perform algorithm selection and hyperparameter optimization to maximize the predictive performance of their model. If deep learning is used, the architecture of the neural network must also be chosen by the machine learning expert.

Each of these steps may be challenging, resulting in significant hurdles to using machine learning. AutoML aims to simplify these steps for non-experts, and to make it easier for them to use machine learning techniques correctly and effectively.

AutoML plays an important role within the broader approach of automating data science, which also includes challenging tasks such as data engineering, data exploration and model interpretation and prediction.

Targets of automation

Automated machine learning can target various stages of the machine learning process. Steps to automate are:

• Data preparation and ingestion (from raw data and miscellaneous formats)
  • Column type detection; e.g., boolean, discrete numerical, continuous numerical, or text
  • Column intent detection; e.g., target/label, stratification field, numerical feature, categorical text feature, or free text feature
  • Task detection; e.g., binary classification, regression, clustering, or ranking
• Feature engineering
  • Feature selection
  • Feature extraction
  • Meta-learning and transfer learning
  • Detection and handling of skewed data and/or missing values
• Model selection - choosing which machine learning algorithm to use, often including multiple competing software implementations
• Ensembling - a form of consensus where using multiple models often gives better results than any single model
• Hyperparameter optimization of the learning algorithm and featurization
• Pipeline selection under time, memory, and complexity constraints
• Selection of evaluation metrics and validation procedures
• Problem checking
  • Leakage detection
  • Misconfiguration detection
• Analysis of obtained results
• Creating user interfaces and visualizations

Challenges and Limitations

There are a number of key challenges being tackled around automated machine learning. A big issue surrounding the field is referred to as "development as a cottage industry". This phrase refers to the issue in machine learning where development relies on manual decisions and biases of experts. This is contrasted to the goal of machine learning which is to create systems that can learn and improve from their own usage and analysis of the data. Basically, it's the struggle between how much experts should get involved in the learning of the systems versus how much freedom they should be giving the machines. However, experts and developers must help create and guide these machines to prepare them for their own learning. To create this system, it requires labor intensive work with knowledge of machine learning algorithms and system design.

Additionally, some other challenges include meta-learning challenges and computational resource allocation.

Development of the nervous system in humans

From Wikipedia, the free encyclopedia

https://en.wikipedia.org/wiki/Development_of_the_nervous_system_in_humans

The development of the nervous system in humans, or neural development or neurodevelopment involves the studies of embryology, developmental biology, and neuroscience to describe the cellular and molecular mechanisms by which the complex nervous system forms in humans, develops during prenatal development, and continues to develop postnatally.

Some landmarks of neural development in the embryo include the formation and differentiation of neurons from stem cell precursors (neurogenesis); the migration of immature neurons from their birthplaces in the embryo to their final positions; the outgrowth of axons from neurons and guidance of the motile growth cone through the embryo towards postsynaptic partners, the generation of synapses between these axons and their postsynaptic partners, the synaptic pruning that occurs in adolescence, and finally the lifelong changes in synapses which are thought to underlie learning and memory.

Typically, these neurodevelopmental processes can be broadly divided into two classes: activity-independent mechanisms and activity-dependent mechanisms. Activity-independent mechanisms are generally believed to occur as hardwired processes determined by genetic programs played out within individual neurons. These include differentiation, migration and axon guidance to their initial target areas. These processes are thought of as being independent of neural activity and sensory experience. Once axons reach their target areas, activity-dependent mechanisms come into play. Neural activity and sensory experience will mediate formation of new synapses, as well as synaptic plasticity, which will be responsible for refinement of the nascent neural circuits.

Development of the human brain

Further information: Human embryonic development and Human brain § Development

Highly schematic flowchart of human brain development

Overview

The central nervous system (CNS) is derived from the ectoderm—the outermost tissue layer of the embryo. In the third week of human embryonic development the neuroectoderm appears and forms the neural plate along the dorsal side of the embryo. The neural plate is the source of the majority of neurons and glial cells of the CNS. A groove forms along the long axis of the neural plate and, by week four of development, the neural plate wraps in on itself to give rise to the neural tube, which is filled with cerebrospinal fluid (CSF). As the embryo develops, the anterior part of the neural tube forms three primary brain vesicles, which become the primary anatomical regions of the brain: the forebrain (prosencephalon), midbrain (mesencephalon), and hindbrain (rhombencephalon). These simple, early vesicles enlarge and further divide into the five secondary brain vesicles – the telencephalon (future cerebral cortex and basal ganglia), diencephalon (future thalamus and hypothalamus), mesencephalon (future colliculi), metencephalon (future pons and cerebellum), and myelencephalon (future medulla). The CSF-filled central chamber is continuous from the telencephalon to the spinal cord, and constitutes the developing ventricular system of the CNS. Because the neural tube gives rise to the brain and spinal cord any mutations at this stage in development can lead to fatal deformities like anencephaly or lifelong disabilities like spina bifida. During this time, the walls of the neural tube contain neural stem cells, which drive brain growth as they divide many times. Gradually some of the cells stop dividing and differentiate into neurons and glial cells, which are the main cellular components of the CNS. The newly generated neurons migrate to different parts of the developing brain to self-organize into different brain structures. Once the neurons have reached their regional positions, they extend axons and dendrites, which allow them to communicate with other neurons via synapses. Synaptic communication between neurons leads to the establishment of functional neural circuits that mediate sensory and motor processing, and underlie behavior.

Neural induction

During early embryonic development the ectoderm becomes specified to give rise to the epidermis (skin) and the neural plate. The conversion of undifferentiated ectoderm to neuro-ectoderm requires signals from the mesoderm. At the onset of gastrulation presumptive mesodermal cells move through the dorsal blastopore lip and form a layer in between the endoderm and the ectoderm. These mesodermal cells that migrate along the dorsal midline give rise to a structure called the notochord. Ectodermal cells overlying the notochord develop into the neural plate in response to a diffusible signal produced by the notochord. The remainder of the ectoderm gives rise to the epidermis (skin). The ability of the mesoderm to convert the overlying ectoderm into neural tissue is called neural induction.

The neural plate folds outwards during the third week of gestation to form the neural groove. Beginning in the future neck region, the neural folds of this groove close to create the neural tube. The formation of the neural tube from the ectoderm is called neurulation. The ventral part of the neural tube is called the basal plate; the dorsal part is called the alar plate. The hollow interior is called the neural canal. By the end of the fourth week of gestation, the open ends of the neural tube, called the neuropores, close off.

A transplanted blastopore lip can convert ectoderm into neural tissue and is said to have an inductive effect. Neural inducers are molecules that can induce the expression of neural genes in ectoderm explants without inducing mesodermal genes as well. Neural induction is often studied in xenopus embryos since they have a simple body pattern and there are good markers to distinguish between neural and non-neural tissue. Examples of neural inducers are the molecules noggin and chordin.

When embryonic ectodermal cells are cultured at low density in the absence of mesodermal cells they undergo neural differentiation (express neural genes), suggesting that neural differentiation is the default fate of ectodermal cells. In explant cultures (which allow direct cell-cell interactions) the same cells differentiate into epidermis. This is due to the action of BMP4 (a TGF-β family protein) that induces ectodermal cultures to differentiate into epidermis. During neural induction, noggin and chordin are produced by the dorsal mesoderm (notochord) and diffuse into the overlying ectoderm to inhibit the activity of BMP4. This inhibition of BMP4 causes the cells to differentiate into neural cells. Inhibition of TGF-β and BMP (bone morphogenetic protein) signaling can efficiently induce neural tissue from human pluripotent stem cells, a model of early human development.

The early brain

Late in the fourth week, the superior part of the neural tube flexes at the level of the future midbrain—the mesencephalon. Above the mesencephalon is the prosencephalon (future forebrain) and beneath it is the rhombencephalon (future hindbrain). The optical vesicle (which will eventually become the optic nerve, retina and iris) forms at the basal plate of the prosencephalon.

The spinal cord forms from the lower part of the neural tube. The wall of the neural tube consists of neuroepithelial cells, which differentiate into neuroblasts, forming the mantle layer (the gray matter). Nerve fibers emerge from these neuroblasts to form the marginal layer (the white matter). The ventral part of the mantle layer (the basal plates) forms the motor areas of the spinal cord, whilst the dorsal part (the alar plates) forms the sensory areas. Between the basal and alar plates is an intermediate layer that contains neurons of the autonomic nervous system.

In the fifth week, the alar plate of the prosencephalon expands to form the cerebral hemispheres (the telencephalon). The basal plate becomes the diencephalon.

The diencephalon, mesencephalon and rhombencephalon constitute the brain stem of the embryo. It continues to flex at the mesencephalon. The rhombencephalon folds posteriorly, which causes its alar plate to flare and form the fourth ventricle of the brain. The pons and the cerebellum form in the upper part of the rhombencephalon, whilst the medulla oblongata forms in the lower part.

• 
  The embryo's brain at four weeks
   
• 
  Cross-section of a developing spinal cord at four weeks
   
• 
  The embryo's nervous system at six weeks

Neuroimaging

Neuroimaging is responsible for great advancements in understanding how the brain develops. EEG and ERP are effective imaging processes used mainly on babies and young children since they are more gentle. Infants are generally tested with fNIRS. The MRI and fMRI are widely used for research on the brain due to the quality of images and analysis possible from them.

Magnetic resonance imaging

MRI's are helpful in analyzing many aspects of the brain. The magnetization-transfer ratio (MTR) measures integrity using magnetization. Fractional anisotropy (FA) measures organization using the diffusion of water molecules. Additionally, mean diffusivity (MD) measures the strength of white matter tracts.

Structural magnetic resonance imaging

Using structural MRI, quantitative assessment of a number of developmental processes can be carried out including defining growth patterns, and characterizing the sequence of myelination. These data complement evidence from Diffusion Tensor Imaging (DTI) studies that have been widely used to investigate the development of white matter.

Functional magnetic resonance imaging

fMRI's test mentalising which is the theory of the mind by activating a network. The posterior superior temporal sulcus (pSTS) and temporo-parietal junction (TPJ) are helpful in predicting movement. In adults, the right pSTS showed greater response than the same region in adolescents when tested on intentional causality. These regions were also activated during the "mind in the eyes" exercise where emotion must be judged based on different images of eyes. Another key region is the anterior temporal cortex (ATC) in the posterior region. In adults, the left ATC showed greater response than the same region in adolescents when tested on emotional tests of mentalising. Finally, the medial prefrontal cortex (MPFC) and the anterior dorsal MPFC (dMPFC) are activated when the mind is stimulated by psychology.

Three-dimensional sonography

Higher resolution imaging has allowed three-dimensional ultrasound to help identify human brain development during the embryonic stages. Studies report that three primary structures are formed in the sixth gestational week. These are the forebrain, the midbrain, and the hindbrain, also known as the prosencephalon, mesencephalon, and the rhombencephalon respectively. Five secondary structures from these in the seventh gestational week. These are the telencephalon, diencephalon, mesencephalon, metencephalon, and myelencephalon which later become the lateral ventricles, third ventricles, aqueduct, and upper and lower parts of the fourth ventricle from the telencephalon to the myelencephalon, during adulthood. 3D ultrasound imaging allows in-vivo depictions of ideal brain development which can help tp recognize irregularities during gestation.

White matter development

Using MRI, studies showed that while white matter increases from childhood (~9 years) to adolescence (~14 years), grey matter decreases. This was observed primarily in the frontal and parietal cortices. Theories as to why this occurs vary. One thought is that the intracortical myelination paired with increased axonal calibre increases the volume of white matter tissue. Another is that synaptic reorganization occurs from proliferation and then pruning.

Grey matter development

The rise and fall of the volume of grey matter in the frontal and parietal lobes peaked at ~12 years of age. The peak for the temporal lobes was ~17 years with the superior temporal cortex being last to mature. The sensory and motor regions matured first after which the rest of the cortex developed. This was characterized by loss of grey matter and it occurred from the posterior to the anterior region. This loss of grey matter and increase of white matter may occur throughout a lifetime though the more robust changes occur from childhood to adolescence.

Neuronal migration

Neuronal migration is the method by which neurons travel from their origin or birthplace to their final position in the brain. Their most common means of migration are radial and tangential migration.

Radial migration

Neural stem cells proliferate in the ventricular zone of the developing neocortex. The first postmitotic cells to migrate from the preplate which are destined to become Cajal–Retzius cells and subplate neurons. These cells do so by somal translocation. Neurons migrating with this mode of locomotion are bipolar and attach the leading edge of the process to the pia. The soma is then transported to the pial surface by nucleokenisis, a process by which a microtubule "cage" around the nucleus elongates and contracts in association with the centrosome to guide the nucleus to its final destination. Radial fibres (also known as radial glia) can translocate to the cortical plate and differentiate either into astrocytes or neurons. Somal translocation can occur at any time during development.

Subsequent waves of neurons split the preplate by migrating along radial glial fibres to form the cortical plate. Each wave of migrating cells travel past their predecessors forming layers in an inside-out manner, meaning that the youngest neurons are the closest to the surface. It is estimated that glial guided migration represents 80-90% of migrating neurons.

Axophilic migration

Many neurons migrating along the anterior-posterior axis of the body use existing axon tracts to migrate along in a process called axophilic migration. An example of this mode of migration is in GnRH-expressing neurons, which make a long journey from their birthplace in the nose, through the forebrain, and into the hypothalamus. Many of the mechanisms of this migration have been worked out, starting with the extracellular guidance cues that trigger intracellular signaling. These intracellular signals, such as calcium signaling, lead to actin and microtubule cytoskeletal dynamics, which produce cellular forces that interact with the extracellular environment through cell adhesion proteins to cause the movement of these cells. Neurophilic migration refers to the migration of neurons along an axon belonging to a different nerve. Gliophilic migration is the migration of glia along glial fibres.

Tangential migration

Most interneurons migrate tangentially through multiple modes of migration to reach their appropriate location in the cortex. An example of tangential migration is the movement of Cajal–Retzius cells within the marginal zone of the cortical neuroepithelium.

Others

There is also a method of neuronal migration called multipolar migration. This is seen in multipolar cells, which are abundantly present in the cortical intermediate zone. They do not resemble the cells migrating by locomotion or somal translocation. Instead these multipolar cells express neuronal markers and extend multiple thin processes in various directions independently of the radial glial fibers.

Neurotrophic factors

Neurotrophic factors are molecules which promote and regulate neuronal survival in the developing nervous system. They are distinguished from ubiquitous metabolites necessary for cellular maintenance and growth by their specificity; each neurotrophic factor promotes the survival of only certain kinds of neurons during a particular stage of their development. In addition, it has been argued that neurotrophic factors are involved in many other aspects of neuronal development ranging from axonal guidance to regulation of neurotransmitter synthesis.

Adult neural development

Main articles: Neuroregeneration and Adult neurogenesis

Neurodevelopment in the adult nervous system includes mechanisms such as remyelination, generation of new neurons, glia, axons, myelin or synapses. Neuroregeneration differs between the peripheral nervous system (PNS) and the central nervous system (CNS) by the functional mechanisms and especially, the extent and speed.

The nervous system continues to develop during adulthood until brain death. For example:

• physical exercise has neurobiological effects
• the consumption of foods (or nutrients), obesity, alterations of the microbiome, drinks, dietary supplements, recreational drugs and medications may possibly also have effects on the development of the nervous system
• various diseases, such as COVID-19, have effects on the development of the nervous system
  • For example, several genes have been identified as to being associated with changes in brain structure over lifetime and are potential Alzheimer's disease therapy-targets.
• psychological events such as mental trauma and resilience-building
• exposure to environmental pollution and toxins such as air pollution may have effects on the further development of the nervous system
• other activities may also have effects on the development of the nervous system, such as lifelong learning, retraining, and types of media- and economic activities
• broadly, brain aging

Research, treatments and policies often distinguish between "mature" brains and "developing" brains while scientists have pointed out that "the complex nature of neurodevelopment itself poses challenges to establishing a point of reference that would indicate when a brain is mature" and that various structural brain measures change constantly throughout the adult phase of life, albeit childhood neuroplasticity-levels may not be reached again and it is thought that there are various critical and sensitive periods of brain development.

Differences to children's learning

Learning is often more efficient in children and takes longer or is more difficult with age. A study using neuroimaging identified rapid neurotransmitter GABA boosting as a major potential explanation-component for why that is.

Children's brains contain more "silent synapses" that are inactive until recruited as part of neuroplasticity and flexible learning or memories. Neuroplasticity is heightened during critical or sensitive periods of brain development, mainly referring to brain development during child development.

What humans learn at the early stages, and what they learn to apply, sets humans on course for life or has a disproportional impact. Adults usually have a higher capacity to select what they learn, to what extent and how. For example, children may learn the given subjects and topics of school curricula via classroom blackboard-transcription handwriting, instead of being able to choose specific topics/skills or jobs to learn and the styles of learning. For instance, children may not have developed consolidated interests, ethics, interest in purpose and meaningful activities, knowledge about real-world requirements and demands, and priorities.

Research

Spatio-temporal modeling of brain development

In early development (before birth and during the first few months), the brain undergoes more changes in size, shape and structure than at any other time in life. Improved understanding of cerebral development during this critical period is important for mapping normal growth, and for investigating mechanisms of injury associated with risk factors for maldevelopment such as premature birth. Hence, there is a need for dense coverage of this age range with a time-varying, age-dependent atlas. Such a spatio-temporal atlases can accurately represent the dynamic changes occurring during early brain development, and can be used as a normative reference space.

Furthermore, large scale gene expression studies of different brain regions from early gestation to aging have been performed. This kind of data provides a unique insight into changes that happen in brain during this long period. This approach showed that 86 per cent of the genes were expressed, and that 90 per cent of these were differentially regulated at the whole-transcript or exon level across brain regions and/or time. The majority of these spatio-temporal differences were detected before birth, with subsequent increases in the similarity among regional transcriptomes. Furthermore, interareal differences exhibit a temporal hourglass pattern, dividing the human neocortical development into three major phases. During the first phase, in the first six months after conception, general architecture of brain regions is largely formed by a burst of genetic activity, which is distinct for specific regions of the neocortex. This rush is followed by a sort of intermission beginning in the third trimester of pregnancy. During this period, most genes that are active in specific brain regions are quieted — except for genes that spur connections between all neocortex regions. Then in late childhood and early adolescence, the genetic orchestra begins again and helps subtly shape neocortex regions that progressively perform more specialized tasks, a process that continues into adulthood.

Embryonic brain development research

Approaches to investigate the organogenesis and early development of the human brain or nervous system include:

• Brain organoids and 'assembloids' (see also: Genetic factors of recent brain evolution)
• Synthetic embryos/embryo models
• Model animals
• Post-mortem studies
• Non-invasive in vivo imaging As of 2014 imaging in utero is not commonly done without strong medical arguments; in 2019 a study reported that "neuroimaging approaches have contributed significantly to our understanding of early brain development"

Human tissue inaccessibility has impeded molecular understanding of the formation of cognitive capacities. The placenta is researched as well.

Better understanding of the development may potentially enable insights into nervous system diseases, improving intelligence, and better protection against harmful impacts from identified factors of fetal development (potentially including from diseases of the mother, various events and xenobiotics).

Specific regions

Research has been able to make new discoveries for various parts of the brain thanks to the noninvasive imaging available.

• Medial Prefrontal Cortex (MPFC)

In this region, more activity is noted in adolescents than in adults when faced with tests on mentalising tasks as well as communicative and personal intent. Decreased activity from adolescence to adulthood. In a mentalising task employing animation, the dMPFC was more stimulated in adults while the ventral MPFC was more stimulated in children. They can be attributed to the use of objective strategy associated with the dMPFC. Theories for decrease in activity from adolescence to adulthood vary. One theory is that cognitive strategy becomes more automatic with age and another is that functional change occurs parallel to neuroanatomical change which is characterized by synaptogenesis and pruning.

The MPFC is an example of one specific region that has become better understood using current imaging techniques. Current research provides many more findings like this.

Early life stress

Early life stress is defined as exposure to circumstances during childhood that overwhelm a child's coping resources and lead to sustained periods of stress. Results from multiple studies indicate that the effects of early life stress on the developing brain are significant and include, but are not limited to the following: increased amygdala volume, decreased activity in frontal cortical and limbic brain structures, and altered white matter structures.

Early life stress is believed to produce changes in brain development by interfering with neurogenesis, synaptic production, and pruning of synapses and receptors. Interference with these processes could result in increased or decreased brain region volumes, potentially explaining the findings that early life stress is associated with increased amygdala volume and decreased anterior cingulate cortex volume.

From the literature, several important conclusions have been drawn. Brain areas that undergo significant post-natal development, such as those involved in memory and emotion are more vulnerable to effects of early life stress. For example, the hippocampus continues to develop after birth and is a structure that is affected by childhood maltreatment. Early life stress seems to interfere with the overproduction of synapses that is typical in childhood, but does not interfere with synaptic pruning in adolescence. This results in smaller hippocampal volumes, potentially explaining the association between early life stress and reduced hippocampal volume. This volume reduction may be associated with the emotion regulation deficits seen in those exposed to early life stress.

The amygdala is particularly vulnerable to early life stress. The amygdala also undergoes significant development during childhood, is structurally and functionally altered in individuals that have experienced early life stress, and is associated with the socioemotional difficulties linked with early life stress.

Receptor type is another consideration when determining whether or not a brain region is sensitive to the effects of early life stress. Brain regions with a high density of glucocorticoid receptors are especially vulnerable to the effects of early life stress, likely because glucocorticoids bind to these receptors during stress exposure, facilitating the development of survival responses at the cost of other important neural pathways. Some examples of brain regions with high glucocorticoid receptor density are the hippocampus and cerebellar vermis. Stress activates the HPA axis, and results in the production of glucocorticoids. Increased glucocorticoid production results in increased activation of these brain regions, facilitating the development of certain neural pathways at the cost of others.

Abnormalities in brain structure and function are often associated with deficits that may persist for years after the stress is removed, and may be a risk factor for future psychopathology. The brain regions most sensitive to early life stress are those undergoing developmental changes during the stress exposure. As a result, stress alters the developmental trajectory of that brain region, producing long-lasting alterations in structure and function.

Common types of early life stress that are documented include maltreatment, neglect, and previous institutionalization. Living in poverty has also been shown to similarly influence brain function.