Добірка наукової літератури з теми "Machine Learning, Learning from Constraints, First-Order Logic, Convex Logical Constraints"

Information extraction (IE) aims to produce structured information from an input text, e.g., Named Entity Recognition and Relation Extraction. Various attempts have been proposed for IE via feature engineering or deep learning. However, most of them fail to associate the complex relationships inherent in the task itself, which has proven to be especially crucial. For example, the relation between 2 entities is highly dependent on their entity types. These dependencies can be regarded as complex constraints that can be efficiently expressed as logical rules. To combine such logic reasoning capabilities with learning capabilities of deep neural networks, we propose to integrate logical knowledge in the form of first-order logic into a deep learning system, which can be trained jointly in an end-to-end manner. The integrated framework is able to enhance neural outputs with knowledge regularization via logic rules, and at the same time update the weights of logic rules to comply with the characteristics of the training data. We demonstrate the effectiveness and generalization of the proposed model on multiple IE tasks.

In many optimization problems arising from scientific, engineering and artificial intelligence applications, objective and constraint functions are available only as the output of a black-box or simulation oracle that does not provide derivative information. Such settings necessitate the use of methods for derivative-free, or zeroth-order, optimization. We provide a review and perspectives on developments in these methods, with an emphasis on highlighting recent developments and on unifying treatment of such problems in the non-linear optimization and machine learning literature. We categorize methods based on assumed properties of the black-box functions, as well as features of the methods. We first overview the primary setting of deterministic methods applied to unconstrained, non-convex optimization problems where the objective function is defined by a deterministic black-box oracle. We then discuss developments in randomized methods, methods that assume some additional structure about the objective (including convexity, separability and general non-smooth compositions), methods for problems where the output of the black-box oracle is stochastic, and methods for handling different types of constraints.

The beginning of the twenty-first century could be characterised by the externalisation of attention, following the externalisation of our other faculties: the term 'exo-attention' can be used to refer to the increasing number of electrical devices performing attentional tasks for us outside of our bodies. At the same time, the logic of industrial production continues to demand human beings to develop automated gestures commanded by the planetary assembly line, intellectual gestures being now added to bodily gestures. This automation of our 'endo-attention' cannot be considered as a temporary step in the process leading to full automation. On the one hand, it coexists and goes along with the logic of 'heteromation', whereby supposedly automated procedures are actually performed by micro-taskers, click farms and Mechanical Turks. On the other hand, the precarisation of labour conditions analysed by Franco Berardi tends to segment our activity into pre-formatted time-cells which alienate us from the very tasks we accomplish. While our endo-attention threatens to be automated through and through, progress in deep learning programming allows exo-attention to become creative: what used to be the specificity of human attention (i.e., its capacity to extract a meaningful figure from a given background) can now be obtained by unsupervised machine learning. Does all this mean that the creativity of human attention has been merely displaced, from creatively paying attention to (a limited number of) things, to creatively devising algorithms that pay attention to (a higher number of) things? This perspective could be technologically attractive, if it weren't trapped within the constraints of neoliberal capitalism. Social – not technological – logics should be the main cause of our concern (and anxiety) about automation. Neoliberal capitalism tends to globally align the infinite diversity of our individual attentions under one single hegemonic imperative to maximise financial profit. This is both egocidal, as it automatises our endo-attention subjected to segmented tasks that no longer make sense to us (pre-empting emancipatory forms of subjectification), and eco-cidal, as the race for short-term profit vandalises our social and natural environments. We therefore need to sharpen our analyses (and anxieties), in order to deflect our fear of automation towards a rejection of neoliberal capitalism.

In many real tasks there are human knowledge expressed in logic formulae as well as data samples described by raw features (e.g., pixels, strings). It is popular to apply SRL or PILPtechniques to exploit human knowledge through learning of symbolic data, or statistical learning techniques to learn from the raw data samples; however, it is often desired to directly exploit these logic formulae on raw data processing, like human beings utilizing knowledge to guide perception. In this paper, we propose an approach, LASIN, which combines Logical Abduction and Statistical Induction. The LASIN approach generates candidate hypotheses based on the abduction of first-order formulae, and then, the hypotheses are exploited as constraints for statistical induction. We apply theLASIN approach to the learning of representation of written primitives, where a primitive is a basic component in human writing. Our results show that the discovered primitives are reasonable for human perception, and these primitives, if used in learning tasks such as classification and domain adaptation, lead to better performances than simply applying feature learning based on raw data only.

AbstractAccording to the objective Bayesian approach to inductive logic, premisses inductively entail a conclusion just when every probability function with maximal entropy, from all those that satisfy the premisses, satisfies the conclusion. When premisses and conclusion are constraints on probabilities of sentences of a first-order predicate language, however, it is by no means obvious how to determine these maximal entropy functions. This paper makes progress on the problem in the following ways. Firstly, we introduce the concept of a limit in entropy and show that, if the set of probability functions satisfying the premisses contains a limit in entropy, then this limit point is unique and is the maximal entropy probability function. Next, we turn to the special case in which the premisses are categorical sentences of the logical language. We show that if the uniform probability function gives the premisses positive probability, then the maximal entropy function can be found by simply conditionalising this uniform prior on the premisses. We generalise our results to demonstrate agreement between the maximal entropy approach and Jeffrey conditionalisation in the case in which there is a single premiss that specifies the probability of a sentence of the language. We show that, after learning such a premiss, certain inferences are preserved, namely inferences to inductive tautologies. Finally, we consider potential pathologies of the approach: we explore the extent to which the maximal entropy approach is invariant under permutations of the constants of the language, and we discuss some cases in which there is no maximal entropy probability function.

Every crisis at the same time reveals, forewarns and implies changes with cyclical trends that can be analyzed from different disciplinary perspectives, building scenarios to anticipate the future, despite uncertainties and risks. And the current crisis certainly appears as one of the most problematic of the modern era: recently, Luigi Ferrara, Director of the School of Design at the George Brown College in Toronto and of the connected Institute without Boundaries, highlighted how the pandemic has simply accelerated undergoing dynamics, exacerbating other crises – climatic, environmental, social, economic – which had already been going on for a long time both locally and globally. In the most economically developed contexts, from North America to Europe, the Covid emergency has led, for example, to the closure of almost 30% of the retail trade, as well as to the disposal and sale of many churches. Places of care and assistance, such as hospitals and elderly houses, have become places of death and isolation for over a year, or have been closed. At the same time, the pandemic has imposed the revolution of the remote working and education, which was heralded – without much success – more than twenty years ago. In these even contradictory dynamics, Ferrara sees many possibilities: new roles for stronger and more capable public institutions as well as the opportunity to rethink and redesign the built environment and the landscape. Last but not least, against a future that could be configured as dystopian, a unique chance to enable forms of citizenship and communities capable of inhabiting more sustainable, intelligent and ethical cities and territories; and architects capable of designing them. This multifactorial and pervasive crisis seems therefore to impose a deep review of the current unequal development models, in the perspective of that “creative destruction” that Schumpeter placed at the basis of the dynamic entrepreneurial push: «To produce means to combine materials and forces within our reach. To produce other things, or the same things by a different method, means to combine these materials and forces differently» (Schumpeter, 1912). A concept well suiting to the design practice as a response to social needs and improving the living conditions. This is the perspective of Architectural Technology, in its various forms, which has always placed the experimental method at the center of its action. As Eduardo Vittoria already pointed out: «The specific contribution of the technological project to the development of an industrial culture is aimed at balancing the emotional-aesthetic data of the design with the technical-productive data of the industry. Design becomes a place of convergence of ideas and skills related to factuality, based on a multidisciplinary intelligence» (Vittoria, 1999). A lucid and appropriate critique of the many formalistic emphases that have invested contemporary architecture. In the most acute phases of the pandemic, the radical nature of this polycrisis has been repeatedly invoked as a lever for an equally radical modification of the development models, for the definitive defeat of conjunctural and emergency modes of action. With particular reference to the Italian context, however, it seems improper to talk about a “change of models” – whether economic, social, productive or programming, rather than technological innovation – since in the national reality the models and reference systems prove to not to be actually structured. The current socio-economic and productive framework, and the political and planning actions themselves, are rather a variegated and disordered set of consolidated practices, habits often distorted when not deleterious, that correspond to stratified regulatory apparatuses, which are inconsistent and often ineffective. It is even more difficult to talk about programmatic rationality models in the specific sector of construction and built environment transformation, where the enunciation of objectives and the prospection of planning actions rarely achieve adequate projects and certain implementation processes, verified for the consistency of the results obtained and monitored for the ability in maintaining the required performance over time. Rather than “changing the model”, in the Italian case, we should therefore talk about giving shape and implementation to an organic and rational system of multilevel and inter-sectorial governance models, which assumes the principles of subsidiarity, administrative decentralization, inter-institutional and public-private cooperation. But, even in the current situation, with the pandemic not yet over, we are already experiencing a sort of “return to order”: after having envisaged radical changes – new urban models environmentally and climatically more sustainable, residential systems and public spaces more responsive to the pressing needs of social demand, priority actions to redevelop the suburbs and to strength infrastructures and ecosystem services, new advanced forms of decision-making decentralization for the co-planning of urban and territorial transformations, and so on – everything seems to has been reset to zero. This is evident from the list of actions and projects proposed by the National Recovery and Resilience Plan (NRRP), where no clear national strategy for green transition emerges, even though it is repeatedly mentioned. As highlighted by the Coordination of Technical-Scientific Associations for the Environment and Landscape1, and as required by EU guidelines2, this transition requires a paradigm shift that assumes eco-sustainability as a transversal guideline for all actions. With the primary objective of protecting ecosystem balances, improving and enhancing the natural and landscape capital, as well as protecting citizen health and well-being from environmental risks and from those generated by improper anthropization phenomena. The contents of the Plan explicitly emphases the need to «repair the economic and social damage of the pandemic crisis» and to «contribute to addressing the structural weaknesses of the Italian economy», two certainly relevant objectives, the pursuit of which, however, could paradoxically contrast precisely with the transition to a more sustainable development. In the Plan, the green revolution and the ecological transition are resolved in a dedicated axis (waste management, hydrogen, energy efficiency of buildings, without however specific reform guidelines of the broader “energy” sector), while «only one of the projects of the Plan regards directly the theme Biodiversity / Ecosystem / Landscape, and in a completely marginal way» (CATAP, 2021). Actions are also limited for assessing the environmental sustainability of the interventions, except the provision of an ad hoc Commission for the streamlining of some procedural steps and a generic indication of compliance with the DNSH-Do not significant Harm criterion (do not cause any significant damage), without specific guidelines on the evaluation methods. Moreover, little or nothing in the Plan refers on actions and investments in urban renewal, abandoned heritage recovery3, of in protecting and enhancing areas characterized by environmental sensitivity/fragility; situations widely present on the national territory, which are instead the first resource for a structural environmental transition. Finally yet importantly, the well-known inability to manage expenditure and the public administration inefficiencies must be considered: a limit not only to the effective implementation of projects, but also to the control of the relationship between time, costs and quality (also environmental) of the interventions. In many places, the Plan has been talked about as an opportunity for a real “reconstruction”, similar to that of post-war Italy; forgetting that the socio-economic renaissance was driven by the INA-Casa Plan4, but also by a considerable robustness of the cultural approach in the research and experimentation of new housing models (Schiaffonati, 2014)5. A possible “model”, which – appropriately updated in socio-technical and environmental terms – could be a reference for an incisive governmental action aiming at answering to a question – the one of the housing – far from being resolved and still a priority, if not an emergency. The crisis also implies the deployment of new skills, with a review of outdated disciplinary approaches, abandoning all corporate resistances and subcultures that have long prevented the change. A particularly deep fracture in our country, which has implications in research, education and professions, dramatically evident in the disciplines of architectural and urban design. Coherently with the EU Strategic Agenda 2019-2024 and the European Pillar of Social Rights, the action plan presented by the Commission in March 2021, with the commitment of the Declaration of Porto on May 7, sets three main objectives for 2030: an employment rate higher than 78%, the participation of more than 60% of adults in training courses every year and at least 15 million fewer people at risk of social exclusion or poverty6. Education, training and retraining, lifelong learning and employment-oriented skills, placed at the center of EU policy action, now require large investments, to stimulate employment transitions towards the emerging sectors of green, circular and digital economies (environmental design and assessment, risk assessment & management, safety, durability and maintainability, design and management of the life cycle of plans, projects, building systems and components: contents that are completely marginal or absent in the current training offer of Architecture). Departments and PhDs in the Technological Area have actively worked with considerable effectiveness in this field. In these regards, we have to recall the role played by Romano Del Nord «protagonist for commitment and clarity in identifying fundamental strategic lines for the cultural and professional training of architects, in the face of unprecedented changes of the environmental and production context» (Schiaffonati, 2021). Today, on the other hand, the axis of permanent and technical training is almost forgotten by ministerial and university policies for the reorganization of teaching systems, with a lack of strategic visions for bridging the deficit of skills that characterizes the area of architecture on the facing environmental and socio-economic challenges. Also and precisely in the dual perspective of greater interaction with the research systems and with the world of companies and institutions, and of that trans- and multi-disciplinary dimension of knowledge, methods and techniques necessary for the ecological transition of settlement systems and construction sector. Due to the high awareness of the Technological Area about the multifactorial and multi-scale dimension of the crises that recurrently affect our territories, SITdA has been configured since its foundation as a place for scientific and cultural debate on the research and training themes. With a critical approach to the consoling academic attitude looking for a “specific disciplinary” external and extraneous to the social production of goods and services. Finalizing the action of our community to «activate relationships between universities, professions, institutions through the promotion of the technological culture of architecture [...], to offer scientific-cultural resources for the training and qualification of young researchers [...], in collaboration with the national education system in order to advance training in the areas of technology and innovation in architecture» (SITdA Statute, 2007). Goals and topics which seem to be current, which Techne intends to resume and develop in the next issues, and already widely present in this n. 22 dedicated to the Circular Economy. A theme that, as emerges from the contributions, permeates the entire field of action of the project: housing, services, public space, suburbs, infrastructures, production, buildings. All contexts in which technological innovation invests both processes and products: artificial intelligence, robotics and automation, internet of things, 3D printing, sensors, nano and biotechnology, biomaterials, biogenetics and neuroscience feed advanced experiments that cross-fertilize different contributions towards common objectives of circularity and sustainability. In this context, the issue of waste, the superfluous, abandonment and waste, emerge, raising the question of re-purpose: an action that crosses a large panel of cases, due to the presence of a vast heritage of resources – materials, artefacts, spaces and entire territories – to be recovered and re-functionalized, transforming, adapting, reusing, reconverting, reactivating the existing for new purposes and uses, or adapting it to new and changing needs. Therefore, by adopting strategies and techniques of reconversion and reuse, of re-manufacturing and recycling of construction and demolition waste, of design for disassembly that operate along even unprecedented supply chains and which are accompanied by actions to extend the useful life cycle of materials , components and building systems, as well as product service logic also extended to durable goods such as the housing. These are complex perspectives but considerably interesting, feasible through the activation of adequate and updated skills systems, for a necessary and possible future, precisely starting from the ability – as designers, researchers and teachers in the area of Architectural Technology – to read the space and conceive a project within a system of rationalities, albeit limited, but substantially founded, which qualify the interventions through approaches validated in research and experimental verification. Contrarily to any ineffective academicism, which corresponds in fact to a condition of subordination caused by the hegemonic dynamics at the base of the crisis itself, but also by a loss of authority that derives from the inadequate preparation of the architects. An expropriation that legitimizes the worst ignorance in the government of the territories, cities and artifacts. Education in Architecture, strictly connected to the research from which contents and methods derive, has its central pivot in the project didactic: activity by its nature of a practical and experimental type, applied to specific places and contexts, concrete and material, and characterized by considerable complexity, due to the multiplicity of factors involved. This is what differentiates the construction sector, delegated to territorial and urban transformations, from any other sector. A sector that borrows its knowledge from other production processes, importing technologies and materials. With a complex integration of which the project is charged, for the realization of the buildings, along a succession of phases for corresponding to multiple regulatory and procedural constraints. The knowledge and rationalization of these processes are the basis of the evolution of the design and construction production approaches, as well as merely intuitive logics. These aspects were the subject of in-depth study at the SITdA National Conference on “Producing Project” (Reggio Calabria, 2018), and relaunched in a new perspective by the International Conference “The project in the digital age. Technology, Nature, Culture” scheduled in Naples on the 1st-2nd of July 2021. A reflection that Techne intends to further develop through the sharing of knowledge and scientific debate, selecting topics of great importance, to give voice to a new phase and recalling the practice of design research, in connection with the production context, institutions and social demand. “Inside the Polycrisis. The possible necessary” is the theme of the call we launched for n. 23, to plan the future despite the uncertainties and risks, foreshadowing strategies that support a unavoidable change, also by operating within the dynamics that, for better or for worse, will be triggered by the significant resources committed to the implementation of the Recovery Plan. To envisage systematic actions based on the centrality of a rational programming, of environmentally appropriate design at the architectural, urban and territorial scales, and of a continuous monitoring of the implementation processes. With the commitment also to promote, after each release, a public moment of reflection and critical assessment on the research progresses. NOTES 1 “Osservazioni del Coordinamento delle Associazioni Tecnico-scientifiche per l’Ambiente e il Paesaggio al PNRR”, 2021. 2 EU Guidelines, SWD-2021-12 final, 21.1.2021. 3 For instance, we can consider the 7,000 km of dismissed railways, with related buildings and areas. 4 The two seven-year activities of the Plan (1949-1963) promoted by Amintore Fanfani, Minister of Labor and Social Security at the time, represented both an employment and a social maneuver, which left us the important legacy of neighborhoods that still today they have their own precise identity, testimony of the architectural culture of the Italian twentieth century. But also a «grandiose machine for the housing» (Samonà, 1949), based on a clear institutional and organizational reorganization, with the establishment of a single body (articulated in the plan implementation committee, led by Filiberto Guala, with regulatory functions of disbursement of funds, assignment of tasks and supervision, and in the INA-Casa Management directed by the architect Arnaldo Foschini, then dean of the Faculty of Architecture), which led to the construction of two million rooms for over 350,000 families. See Di Biagi F. (2013), Il Contributo italiano alla storia del Pensiero – Tecnica, Enciclopedia Treccani. 5 From Quaderni of the Centro Studi INA-Casa, to Gescal and in the Eighties to the activity of CER. Complex theme investigated by Fabrizio Schiaffonati in Il progetto della residenza sociale, edited by Raffaella Riva. 6 Ferruccio De Bortoli underlines in Corriere della Sera of 15 May 2021: «The revolution of lifelong learning (which) is no less important for Brussels than the digital or green one. By 2030, at least 60 per cent of the active population will have to participate in training courses every year. It will be said: but 2030 is far away. There’s time. No, because most people have escaped that to achieve this goal, by 2025 – that is, in less than four years – 120 million Europeans will ideally return to school. A kind of great educational vaccination campaign. Day after tomorrow».

On 17 January 2013, Wired chose the smart contact lens as one of “7 Massive Ideas That Could Change the World” describing a Google-led research project. Wired explains that the inventor, Dr. Babak Parviz, wants to build a microsystem on a contact lens: “Using radios no wider than a few human hairs, he thinks these lenses can augment reality and incidentally eliminate the need for displays on phones, PCs, and widescreen TVs”. Explained further in other sources, the technology entails an antenna, circuits embedded into a contact lens, GPS, and an LED to project images on the eye, creating a virtual display (Solve for X). Wi-Fi would stream content through a transparent screen over the eye. One patent describes a camera embedded in the lens (Etherington). Another mentions medical sensing, such as glucose monitoring of tears (Goldman). In other words, Google proposes an imagined future when we use contact lenses to search the Internet (and be searched by it), shop online, communicate with friends, work, navigate maps, swipe through Tinder, monitor our health, watch television, and, by that time, probably engage in a host of activities not yet invented. Often referred to as a bionic contact, the smart contact lens would signal a weighty shift in the way we work, socialize, and frame our online identities. However, speculative discussion over this radical shift in personal computing, rarely if ever, includes consideration of how the body, acting as a host to digital information, will manage to assimilate not only significant affordances, but also significant constraints and vulnerabilities. At this point, for most people, the smart contact lens is just an idea. Is a new medium of communication started when it is launched in an advertising campaign? When we Like it on Facebook? If we chat about it during a party amongst friends? Or, do a critical mass of people actually have to be using it to say it has started? One might say that Apple’s Macintosh computer started as a media platform when the world heard about the famous 1984 television advertisement aired during the American NFL Super Bowl of that year. Directed by Ridley Scott, the ad entails an athlete running down a passageway and hurling a hammer at a massive screen depicting cold war style rulers expounding state propaganda. The screen explodes freeing those imprisoned from their concentration camp existence. The direct reference to Orwell’s 1984 serves as a metaphor for IBM in 1984. PC users were made analogous to political prisoners and IBM served to represent the totalitarian government. The Mac became a something that, at the time, challenged IBM, and suggested an alternative use for the desktop computer that had previously been relegated for work rather than life. Not everyone bought a Mac, but the polemical ad fostered the idea that Mac was certainly the start of new expectations, civic identities, value-systems, and personal uses for computers. The smart contact lens is another startling start. News of it shocks us, initiates social media clicks and forwards, and instigates dialogue. But, it also indicates the start of a new media paradigm that is already undergoing popular adoption as it is announced in mainstream news and circulated algorithmically across media channels. Since 2008, news outlets like CNN, The New York Times, The Globe and Mail, Asian International News, United News of India, The Times of London and The Washington Post have carried it, feeding the buzz in circulation that Google intends. Attached to the wave of current popular interest generated around any technology claiming to be “wearable,” a smart contact lens also seems surreptitious. We would no longer hold smartphones, but hide all of that digital functionality beneath our eyelids. Its emergence reveals the way commercial models have dramatically changed. The smart contact lens is a futuristic invention imagined for us and about us, but also a sensationalized idea socializing us to a future that includes it. It is also a real device that Parviz (with Google) has been inventing, promoting, and patenting for commercial applications. All of these workings speak to a broader digital culture phenomenon. We argue that the smart contact lens discloses a process of nascent posthuman adaptation, launched in an era that celebrates wearable media as simultaneously astonishing and banal. More specifically, we adopt technology based on our adaptation to it within our personal, political, medial, social, and biological contexts, which also function in a state of flux. N. Katherine Hayles writes that “Contemporary technogenesis, like evolution in general, is not about progress ... rather, contemporary technogenesis is about adaptation, the fit between organisms and their environments, recognizing that both sides of the engagement (human and technologies) are undergoing coordinated transformations” (81). This article attends to the idea that in these early stages, symbolic acts of adaptation signal an emergent medium through rhetorical processes that society both draws from and contributes to. In terms of project scope, this article contributes a focused analysis to a much larger ongoing digital rhetoric project. For the larger project, we conducted a discourse analysis on a collection of international publications concerning Babak Parviz and the invention. We searched for and collected newspaper stories, news broadcasts, YouTube videos from various sources, academic journal publications, inventors’ conference presentations, and advertising, all published between January 2008 and May 2014, generating a corpus of more than 600 relevant artifacts. Shortly after this time, Dr. Parviz, a Professor at the University of Washington, left the secretive GoogleX lab and joined Amazon.com (Mac). For this article we focus specifically on the idea of beginnings or genesis and how digital spaces increasingly serve as the grounds for emergent digital cultural phenomena that are rarely recognized as starting points. We searched through the corpus to identify a few exemplary international mainstream news stories to foreground predominant tropes in support of the claim we make that smart contacts lenses are a startling idea. Content producers deliberately use astonishment as a persuasive device. We characterize the idea of a smart contact lens cast in rhetorical terms in order to reveal how its allure works as a process of adaptation. Rhetorician and philosopher, Kenneth Burke writes that “rhetorical language is inducement to action (or to attitude)” (42). A rhetorical approach is instrumental because it offers a model to explain how we deploy, often times, manipulative meaning as senders and receivers while negotiating highly complex constellations of resources and contexts. Burke’s rhetorical theory can show how messages influence and become influenced by powerful hierarchies in discourse that seem transparent or neutral, ones that seem to fade into the background of our consciousness. For this article, we also concentrate on rhetorical devices such as ethos and the inventor’s own appeals through different modes of communication. Ethos was originally proposed by Aristotle to identify speaker credibility as a persuasive tactic. Addressed by scholars of rhetoric for centuries, ethos has been reconfigured by many critical theorists (Burke; Baumlin Ethos; Hyde). Baumlin and Baumlin suggest that “ethos describes an audience’s projection of authority and trustworthiness onto the speaker ... ethos suggests that the ethical appeal to be a radically psychological event situated in the mental processes of the audience – as belonging as much to the audience as to the actual character of a speaker” (Psychology 99). Discussed in the next section, our impression of Parviz and his position as inventor plays a dramatic role in the surfacing of the smart contact lens. Digital Rhetoric is an “emerging scholarly discipline concerned with the interpretation of computer-generated media as objects of study” (Losh 48). In an era when machine-learning algorithms become the messengers for our messages, which have become commodity items operating across globalized, capitalist networks, digital rhetoric provides a stable model for our approach. It leads us to demonstrate how this emergent medium and invention, the smart contact lens, is born amid new digital genres of speculative communication circulated in the everyday forums we engage on a daily basis. Smart Contact Lenses, Sensationalism, and Identity One relevant site for exploration into how an invention gains ethos is through writing or video penned or produced by the inventor. An article authored by Parviz in 2009 discusses his invention and the technical advancements that need to be made before the smart contact lens could work. He opens the article using a fictional and sensationalized analogy to encourage the adoption of his invention: The human eye is a perceptual powerhouse. It can see millions of colors, adjust easily to shifting light conditions, and transmit information to the brain at a rate exceeding that of a high-speed Internet connection.But why stop there?In the Terminator movies, Arnold Schwarzenegger’s character sees the world with data superimposed on his visual field—virtual captions that enhance the cyborg’s scan of a scene. In stories by the science fiction author Vernor Vinge, characters rely on electronic contact lenses, rather than smartphones or brain implants, for seamless access to information that appears right before their eyes. Identity building is made to correlate with smart contact lenses in a manner that frames them as exciting. Coming to terms with them often involves casting us as superhumans, wielding abilities that we do not currently possess. One reason for embellishment is because we do not need digital displays on the eyes, so the motive to use them must always be geared to transcending our assumed present condition as humans and society members. Consequently, imagination is used to justify a shift in human identity along a future trajectory.This passage above also instantiates a transformation from humanist to posthumanist posturing (i.e. “the cyborg”) in order to incent the adoption of smart contact lenses. It begins with the bold declarative statement, “The human eye is a perceptual powerhouse,” which is a comforting claim about our seemingly human superiority. Indexing abstract humanist values, Parviz emphasizes skills we already possess, including seeing a plethora of colours, adjusting to light on the fly, and thinking fast, indeed faster than “a high-speed Internet connection”. However, the text goes on to summon the Terminator character and his optic feats from the franchise of films. Filmic cyborg characters fulfill the excitement that posthuman rhetoric often seems to demand, but there is more here than sensationalism. Parviz raises the issue of augmenting human vision using science fiction as his contextualizing vehicle because he lacks another way to imbricate the idea. Most interesting in this passage is the inventor’s query “But why stop there?” to yoke the two claims, one biological (i.e., “The human eye is a perceptual powerhouse”) and one fictional (i.e. Terminator, Vernor Vinge characters). The query suggests, Why stop with human superiority, we may as well progress to the next level and embrace a smart contact lens just as fictional cyborgs do. The non-threatening use of fiction makes the concept seem simultaneously exciting and banal, especially because the inventor follows with a clear description of the necessary scientific engineering in the rest of the article. This rhetorical act signifies the voice of a technoelite, a heavily-funded cohort responding to global capitalist imperatives armed with a team of technologists who can access technological advancements and imbue comments with an authority that may extend beyond their fields of expertise, such as communication studies, sociology, psychology, or medicine. The result is a powerful ethos. The idea behind the smart contact lens maintains a degree of respectability long before a public is invited to use it.Parviz exhumes much cultural baggage when he brings to life the Terminator character to pitch smart contact lenses. The Terminator series of films has established the “Arnold Schwarzenegger” character a cultural mainstay. Each new film reinvented him, but ultimately promoted him within a convincing dystopian future across the whole series: The Terminator (Cameron), Terminator 2: Judgment Day (Cameron), Terminator 3: Rise of the Machines (Mostow), Terminator Salvation (McG) and Terminator Genisys (Taylor) (which appeared in 2015 after Parviz’s article). Recently, several writers have addressed how cyborg characters figure significantly in our cultural psyche (Haraway, Bukatman; Leaver). Tama Leaver’s Artificial Culture explores the way popular, contemporary, cinematic, science fiction depictions of embodied Artificial Intelligence, such as the Terminator cyborgs, “can act as a matrix which, rather than separating or demarcating minds and bodies or humanity and the digital, reinforce the symbiotic connection between people, bodies, and technologies” (31). Pointing out the violent and ultimately technophobic motive of The Terminator films, Leaver reads across them to conclude nevertheless that science fiction “proves an extremely fertile context in which to address the significance of representations of Artificial Intelligence” (63).Posthumanism and TechnogenesisOne reason this invention enters the public’s consciousness is its announcement alongside a host of other technologies, which seem like parts of a whole. We argue that this constant grouping of technologies in the news is one process indicative of technogenesis. For example, City A.M., London’s largest free commuter daily newspaper, reports on the future of business technology as a hodgepodge of what ifs: As Facebook turns ten, and with Bill Gates stepping down as Microsoft chairman, it feels like something is drawing to an end. But if so, it is only the end of the technological revolution’s beginning ... Try to look ahead ten years from now and the future is dark. Not because it is bleak, but because the sheer profusion of potential is blinding. Smartphones are set to outnumber PCs within months. After just a few more years, there are likely to be 3bn in use across the planet. In ten years, who knows – wearables? smart contact lenses? implants? And that’s just the start. The Internet of Things is projected to be a $300bn (£183bn) industry by 2020. (Sidwell) This reporting is a common means to frame the commodification of technology in globalized business news that seeks circulation as much as it does readership. But as a text, it also posits how individuals frame the future and their participation with it (Pedersen). Smart contacts appear to move along this exciting, unstoppable trajectory where the “potential is blinding”. The motive is to excite and scare. However, simultaneously, the effect is predictable. We are quite accustomed to this march of innovations that appears everyday in the morning paper. We are asked to adapt rather than question, consequently, we never separate the parts from the whole (e.g., “wearables? smart contact lenses? Implants”) in order to look at them critically.In coming to terms with Cary Wolf’s definition of posthumanism, Greg Pollock writes that posthumanism is the questioning that goes on “when we can no longer rely on ‘the human’ as an autonomous, rational being who provides an Archimedean point for knowing about the world (in contrast to “humanism,” which uses such a figure to ground further claims)” (208). With similar intent, N. Katherine Hayles formulating the term technogenesis suggests that we are not really progressing to another level of autonomous human existence when we adopt media, we are in effect, adapting to media and media are also in a process of adapting to us. She writes: As digital media, including networked and programmable desktop stations, mobile devices, and other computational media embedded in the environment, become more pervasive, they push us in the direction of faster communication, more intense and varied information streams, more integration of humans and intelligent machines, and more interactions of language with code. These environmental changes have significant neurological consequences, many of which are now becoming evident in young people and to a lesser degree in almost everyone who interacts with digital media on a regular basis. (11) Following Hayles, three actions or traits characterize adaptation in a manner germane to the technogenesis of media like smart contact lenses. The first is “media embedded in the environment”. The trait of embedding technology in the form of sensors and chips into external spaces evokes the onset of The Internet of Things (IoT) foundations. Extensive data-gathering sensors, wireless technologies, mobile and wearable components integrated with the Internet, all contribute to the IoT. Emerging from cloud computing infrastructures and data models, The IoT, in its most extreme, involves a scenario whereby people, places, animals, and objects are given unique “embedded” identifiers so that they can embark on constant data transfer over a network. In a sense, the lenses are adapted artifacts responding to a world that expects ubiquitous networked access for both humans and machines. Smart contact lenses will essentially be attached to the user who must adapt to these dynamic and heavily mediated contexts.Following closely on the first, the second point Hayles makes is “integration of humans and intelligent machines”. The camera embedded in the smart contact lens, really an adapted smartphone camera, turns the eye itself into an image capture device. By incorporating them under the eyelids, smart contact lenses signify integration in complex ways. Human-machine amalgamation follows biological, cognitive, and social contexts. Third, Hayles points to “more interactions of language with code.” We assert that with smart contact lenses, code will eventually govern interaction between countless agents in accordance with other smart devices, such as: (1) exchanges of code between people and external nonhuman networks of actors through machine algorithms and massive amalgamations of big data distributed on the Internet;(2) exchanges of code amongst people, human social actors in direct communication with each other over social media; and (3) exchanges of coding and decoding between people and their own biological processes (e.g. monitoring breathing, consuming nutrients, translating brainwaves) and phenomenological (but no less material) practices (e.g., remembering, grieving, or celebrating). The allure of the smart contact lens is the quietly pressing proposition that communication models such as these will be radically transformed because they will have to be adapted to use with the human eye, as the method of input and output of information. Focusing on genetic engineering, Eugene Thacker fittingly defines biomedia as “entail[ing] the informatic recontextualization of biological components and processes, for ends that may be medical or nonmedical (economic, technical) and with effects that are as much cultural, social, and political as they are scientific” (123). He specifies, “biomedia are not computers that simply work on or manipulate biological compounds. Rather, the aim is to provide the right conditions, such that biological life is able to demonstrate or express itself in a particular way” (123). Smart contact lenses sit on the cusp of emergence as a biomedia device that will enable us to decode bodily processes in significant new ways. The bold, technical discourse that announces it however, has not yet begun to attend to the seemingly dramatic “cultural, social, and political” effects percolating under the surface. Through technogenesis, media acclimatizes rapidly to change without establishing a logic of the consequences, nor a design plan for emergence. Following from this, we should mention issues such as the intrusion of surveillance algorithms deployed by corporations, governments, and other hegemonic entities that this invention risks. If smart contact lenses are biomedia devices inspiring us to decode bodily processes and communicate that data for analysis, for ourselves, and others in our trust (e.g., doctors, family, friends), we also need to be wary of them. David Lyon warns: Surveillance has spilled out of its old nation-state containers to become a feature of everyday life, at work, at home, at play, on the move. So far from the single all-seeing eye of Big Brother, myriad agencies now trace and track mundane activities for a plethora of purposes. Abstract data, now including video, biometric, and genetic as well as computerized administrative files, are manipulated to produce profiles and risk categories in a liquid, networked system. The point is to plan, predict, and prevent by classifying and assessing those profiles and risks. (13) In simple terms, the smart contact lens might disclose the most intimate information we possess and leave us vulnerable to profiling, tracking, and theft. Irma van der Ploeg presupposed this predicament when she wrote: “The capacity of certain technologies to change the boundary, not just between what is public and private information but, on top of that, between what is inside and outside the human body, appears to leave our normative concepts wanting” (71). The smart contact lens, with its implied motive to encode and disclose internal bodily information, needs considerations on many levels. Conclusion The smart contact lens has made a digital beginning. We accept it through the mass consumption of the idea, which acts as a rhetorical motivator for media adoption, taking place long before the device materializes in the marketplace. This occurrence may also be a sign of our “posthuman predicament” (Braidotti). We have argued that the smart contact lens concept reveals our posthuman adaptation to media rather than our reasoned acceptance or agreement with it as a logical proposition. By the time we actually squabble over the price, express fears for our privacy, and buy them, smart contact lenses will long be part of our everyday culture. References Baumlin, James S., and Tita F. Baumlin. “On the Psychology of the Pisteis: Mapping the Terrains of Mind and Rhetoric.” Ethos: New Essays in Rhetorical and Critical Theory. Eds. James S. Baumlin and Tita F. Baumlin. Dallas: Southern Methodist University Press, 1994. 91-112. Baumlin, James S., and Tita F. Baumlin, eds. Ethos: New Essays in Rhetorical and Critical Theory. Dallas: Southern Methodist University Press, 1994. Bilton, Nick. “A Rose-Colored View May Come Standard.” The New York Times, 4 Apr. 2012. Braidotti, Rosi. The Posthuman. Cambridge: Polity, 2013. Bukatman, Scott. Terminal Identity: The Virtual Subject in Postmodern Science Fiction. Durham: Duke University Press, 1993. Burke, Kenneth. A Rhetoric of Motives. Berkeley: University of California Press, 1950. Cameron, James, dir. The Terminator. Orion Pictures, 1984. DVD. Cameron, James, dir. Terminator 2: Judgment Day. Artisan Home Entertainment, 2003. DVD. Etherington, Darrell. “Google Patents Tiny Cameras Embedded in Contact Lenses.” TechCrunch, 14 Apr. 2014. Goldman, David. “Google to Make Smart Contact Lenses.” CNN Money 17 Jan. 2014. Haraway, Donna. Simians, Cyborgs and Women: The Reinvention of Nature. London: Free Association Books, 1991. Hayles, N. Katherine. How We Think: Digital Media and Contemporary Technogenesis. Chicago: University of Chicago, 2012. Hyde, Michael. The Ethos of Rhetoric. Columbia: University of South Carolina Press, 2004. Leaver, Tama. Artificial Culture: Identity, Technology, and Bodies. New York: Routledge, 2012. Losh, Elizabeth. Virtualpolitik: An Electronic History of Government Media-Making in a Time of War, Scandal, Disaster, Miscommunication, and Mistakes. Boston: MIT Press. 2009. Lyon, David, ed. Surveillance as Social Sorting: Privacy, Risk and Digital Discrimination. New York: Routledge, 2003. Mac, Ryan. “Amazon Lures Google Glass Creator Following Phone Launch.” Forbes.com, 14 July 2014. McG, dir. Terminator Salvation. Warner Brothers, 2009. DVD. Mostow, Jonathan, dir. Terminator 3: Rise of the Machines. Warner Brothers, 2003. DVD. Parviz, Babak A. “Augmented Reality in a Contact Lens.” IEEE Spectrum, 1 Sep. 2009. Pedersen, Isabel. Ready to Wear: A Rhetoric of Wearable Computers and Reality-Shifting Media. Anderson, South Carolina: Parlor Press, 2013. Pollock, Greg. “What Is Posthumanism by Cary Wolfe (2009).” Rev. of What is Posthumanism?, by Cary Wolfe. Journal for Critical Animal Studies 9.1/2 (2011): 235-241. Sidwell, Marc. “The Long View: Bill Gates Is Gone and the Dot-com Era Is Over: It's Only the End of the Beginning.” City A.M., 7 Feb. 2014. “Solve for X: Babak Parviz on Building Microsystems on the Eye.” YouTube, 7 Feb. 2012. Taylor, Alan, dir. Terminator: Genisys. Paramount Pictures, 2015. DVD. Thacker, Eugene “Biomedia.” Critical Terms for Media Studies. Eds. W.J.T Mitchell and Mark Hansen, Chicago: Chicago Press, 2010. 117-130. Van der Ploeg, Irma. “Biometrics and the Body as Information.” Surveillance as Social Sorting: Privacy, Risk and Digital Discrimination. Ed. David Lyon. New York: Routledge, 2003. 57-73. Wired Staff. “7 Massive Ideas That Could Change the World.” Wired.com, 17 Jan. 2013.

A key point in the success of machine learning, and in particular deep learning, has been the availability of high-performance computing architectures allowing to process a large amount of data. However, this potentially prevents a wider application of machine learning in real world applications, where the collection of training data is often a slow and expensive process, requiring an extensive human intervention. This suggests to look at possible ways to overcome this limitation, for instance injecting prior knowledge into a learning problem to express some desired behaviors for the functions to be learned. In this thesis, we consider the case of prior knowledge expressed by means of first-order logic formulas to be integrated into a learning problem. In particular, at first the formulas are converted into real-valued functions by means of t-norm fuzzy logic operators. Thereafter, a loss component (a constraint) is assigned to any function representing a formula and all these components are aggregated (e.g. summed) together with other possible loss components, e.g. a regularization term or some loss components associated to supervisions, if they are available for the functions to be learned. Both the functional representation of a formula and the mapping into a loss component have been investigated, and some theoretical results are discussed to get an insight on how to bring some benefits for different learning schemata. In particular we define a fragment of Lukasiewicz logic that guarantees to yield convex functional constraints given any knowledge base made of first-order logic formulas. The convexity of these constraints is exploited to formulate collective classification as a quadratic optimization problem and some experimental results are discussed. In addition, we extend classic Support Vector Machines with logical constraints, still preserving quadratic programming resolution. Since formulas may be logically depending on each other, some of the constraints may turn out to be unnecessary with respect to the learning process. This suggests to generalize the notion of support vector to support constraint, and we provide both logical and algebraic criteria to determine the constraints that are unnecessary. Finally we present LYRICS, a general interface implemented in TensorFlow to integrate both deep learning architectures and a first-order logic representation of knowledge for a learning problem. In particular, we show several learning tasks that may be addressed in LYRICS, with a special discussion for the case of visual generation.

Semantic Image Interpretation (SII) is the task of extracting structured semantic descriptions from images. It is widely agreed that the combined use of visual data and background knowledge is of great importance for SII. Recently, Statistical Relational Learning (SRL) approaches have been developed for reasoning under uncertainty and learning in the presence of data and rich knowledge. Logic Tensor Networks (LTNs) are a SRL framework which integrates neural networks with first-order fuzzy logic to allow (i) efficient learning from noisy data in the presence of logical constraints, and (ii) reasoning with logical formulas describing general properties of the data. In this paper, we develop and apply LTNs to two of the main tasks of SII, namely, the classification of an image's bounding boxes and the detection of the relevant part-of relations between objects. To the best of our knowledge, this is the first successful application of SRL to such SII tasks. The proposed approach is evaluated on a standard image processing benchmark. Experiments show that background knowledge in the form of logical constraints can improve the performance of purely data-driven approaches, including the state-of-the-art Fast Region-based Convolutional Neural Networks (Fast R-CNN). Moreover, we show that the use of logical background knowledge adds robustness to the learning system when errors are present in the labels of the training data.