Recent years have witnessed remarkable advancements in machine learning, but with these advances come concerns about data privacy. Machine learning inherently involves learning functions from data, and this process can potentially lead to information leakage through various attacks on the learned model. Additionally, the presence of malicious actors who may poison input data to manipulate the model has become a growing concern. Consequently, the ability to unlearn specific data samples on demand has become critically important. Federated Learning (FL) has emerged as a powerful approach to address these challenges. In FL, multiple participants or clients collaborate to train a single global machine learning model without sharing their training data. However, the issue of machine unlearning is particularly pertinent in FL, especially in scenarios where clients are not fully trustworthy. This paper delves into the investigation of the efficacy of solving machine unlearning problems within the FL framework. The central research question this work tackles is: How can we effectively unlearn the entire dataset from one or multiple clients once an FL training is completed, while maintaining privacy and without access to the data? To address this challenge, we introduce the concept of ”contribution,” which quantifies how much each client contributes to the training of the global FL model. In our implementation, we employ an Encoder-Decoder model on the server’s end to disentangle these contributions as the FL process progresses. Notably, our approach is unique in that there is no existing work that utilizes a similar concept nor similar models. Our findings, supported by extensive experiments on datasets MNIST and FashionMNIST, demonstrate that our proposed approach successfully solves the unlearning task in FL. Remarkably, it achieves results comparable to retraining from scratch without requiring the participation of the specific client whose data needs to be unlearned. Moreover, additional ablation studies indicate the sensitivity of the proposed model to specific structural hyperparameters. / Här har de senaste åren bevittnat enastående framsteg inom maskininlärning, men med dessa framsteg kommer bekymmer om dataskydd. Maskininlärning innebär i grunden att lära sig funktioner från data, och denna process kan potentiellt leda till läckage av information genom olika attacker mot den inlärda modellen. Dessutom har närvaron av illvilliga aktörer som kan förgifta indata för att manipulera modellen blivit en växande oro. Följaktligen har förmågan att avlära specifika datasatser på begäran blivit av avgörande betydelse. Federerad inlärning (FL) har framträtt som en kraftfull metod för att ta itu med dessa utmaningar. I FL samarbetar flera deltagare eller klienter för att träna en enda global maskininlärningsmodell utan att dela sina träningsdata. Emellertid är problemet med maskinavlärande särskilt relevant inom FL, särskilt i situationer där klienterna inte är fullt pålitliga. Denna artikel fördjupar sig i undersökningen av effektiviteten av att lösa problem med maskinavlärande inom FL-ramverket. Den centrala forskningsfråga som detta arbete behandlar är: Hur kan vi effektivt avlära hela datasamlingen från en eller flera klienter när FL-utbildningen är klar, samtidigt som vi bevarar integritet och inte har tillgång till datan? För att ta itu med denna utmaning introducerar vi begreppet ”bidrag,” som kvantifierar hur mycket varje klient bidrar till träningen av den globala FLmodellen. I vår implementering använder vi en Encoder-Decoder-modell på serverns sida för att reda ut dessa bidrag när FL-processen fortskrider. Det är värt att notera att vår metod är unik eftersom det inte finns något befintligt arbete som använder ett liknande koncept eller liknande modeller. Våra resultat, som stöds av omfattande experiment på dataseten MNIST och FashionMNIST, visar att vår föreslagna metod framgångsrikt löser avlärandeuppgiften i FL. Anmärkningsvärt uppnår den resultat som är jämförbara med att träna om från grunden utan att kräva deltagandet av den specifika klient vars data behöver avläras. Dessutom indikerar ytterligare avläggningsstudier känsligheten hos den föreslagna modellen för specifika strukturella hyperparametrar.

<p dir="ltr">Deep learning has achieved remarkable success across various domains, largely relyingon assumptions of ideal data conditions—such as balanced distributions, accurate labeling,and sufficient computational resources—that rarely hold in real-world applications. Thisthesis addresses the significant challenges posed by data non-idealities, including privacyconcerns, label noise, non-IID (Independent and Identically Distributed) data, and adversarial threats, which can compromise model performance and security. Additionally, weexplore the computational limitations inherent in traditional architectures by introducingin-memory computing techniques to mitigate the memory bottleneck in deep neural networkimplementations.We propose five novel contributions to tackle these challenges and enhance the efficiencyand robustness of deep learning models. First, we introduce a gradient-free machine unlearning algorithm to ensure data privacy by effectively forgetting specific classes withoutretraining. Second, we propose a corrective machine unlearning technique, SAP, that improves robustness against label noise using Scaled Activation Projections. Third, we presentthe Neighborhood Gradient Mean (NGM) method, a decentralized learning approach thatoptimizes performance on non-IID data with minimal computational overhead. Fourth, wedevelop TREND, an ensemble design strategy that leverages transferability metrics to enhance adversarial robustness. Finally, we explore an in-memory computing solution, IMAC,that enables energy-efficient and low-latency multiplication and accumulation operationsdirectly within 6T SRAM arrays.These contributions collectively advance the state-of-the-art in handling data non-idealitiesand computational efficiency in deep learning, providing robust, scalable, and privacypreserving solutions suitable for real-world deployment across diverse environments.</p>

Contemporary organisations face a raft of challenges in coping with competing demands and rapidly changing environments. With these demands and changes comes the need for those within the organisation to be adequately skilled to meet these challenges both now and into the future. There is a growing concern that the rate of change is such that learning will not be sufficient and that individuals will need to be skilled in unlearning or letting go of past practice and behaviour. This research investigated individual unlearning as it applies in the workplace, and enabled the development of a process model of unlearning that provides specific indication of factors affecting unlearning during times of change. In particular, this thesis highlights the critical importance of elements of a more personal and affective nature; often referred to as "soft" issues. Six key factors at the level of the individual were identified as impacting unlearning; positive prior outlook, individual inertia, feelings and expectations, positive experience and informal support, understanding the need for change, and assessment of the new way. Two factors emerged from the organisational level that also impact unlearning; organisational support and training and history of organisational change. Many change efforts will fail because of lack of attention to individuals, how they unlearn and the level of feelings and expectations that accompany change. This research demonstrates that organisations must provide resources and education to provide both those in supervisory roles and those impacted by change with the necessary skills to unlearn and to embrace change at an individual level.

No description available.

In 2015, the Truth and Reconciliation Commission of Canada (TRC) called on Ministries of Education, Faculties of Education, school administrators, and K-12 teachers to integrate Indigenous knowledges and pedagogies across the school curriculum. The TRC explicitly emphasized that education would be the intergenerational key to reconciliation in Canada and most provinces and territories quickly implemented curricula and developed resources to respond to the Calls to Action. Despite this mandate and these commitments, many teachers and teacher candidates continue to report that they do not have the skills, knowledge, or confidence to teach about the history of the Indian Residential Schooling system, Indigenous knowledges, or reconciliation. Research suggests that teacher resistance to "difficult knowledge" is a crucial contributing factor toward teachers avoiding, ignoring, and dismissing reconciliation work and upholding colonial logics. Moreover, teacher candidates and teachers often rely on the inaccurate and incomplete narratives they have learned about Canadians and First Nations, Inuit, and Métis Peoples. This impacts what and how they teach about these relationships, complicating the transformational changes the TRC urgently called for. How, then, might teachers unlearn these colonial stories and move from learning about Indigenous peoples to learning from them? Drawing on Donald’s concept of "ethical relationality", this study employed a qualitative approach to conduct conversational interviews with teacher candidates, teachers, staff, and students at two research sites. This study asks, "What are the curricular and pedagogical significances of ethical relationality to processes of unlearning colonialism?" Using a hermeneutic approach to interpret the stories shared, this study weaved within and between the landscapes of home and place. Findings reveal that teachers who experience supportive, multi-layered, and extended opportunities to unlearn settler colonialism and learn Indigenous wisdom traditions and knowledges from Indigenous peoples have the opportunity to understand a new story about Canadian-Indigenous relations. This study suggests that unless teachers begin to unlearn colonial logics, deeply understanding that they are implicated in ethical kinship relations with the places in which they live and with First Nations, Inuit, and Métis peoples, there is a significant possibility that curricula, professional development, and resources will not manifest in the transformational change that the TRC called for.

The establishment of the General Data Protection Regulation (GDPR) in Europe in 2018, including the "Right to be Forgotten" poses important questions about the necessity of efficient data deletion techniques for trained Machine Learning models to completely enforce this right, since retraining from scratch such models whenever a data point must be deleted seems impractical. We tackle such a problem for Gaussian Process Regression and define in this paper an efficient exact unlearning technique for Gaussian Process Regression which completely include the optimization of the hyperparameters of the kernel function. The method is based on an efficient retracing of past optimizations by the Resilient Backpropagation (Rprop) algorithm through the online formulation of a Gaussian Process regression. Furthermore, we develop an extension of the proposed method to the Product-of-Experts and Bayesian Committee Machines types of local approximations of Gaussian Process Regression, further enhancing the unlearning capabilities through a random partitioning of the dataset. The performance of the proposed method is largely dependent on the regression task. We show through multiple experiments on different problems that several iterations of such optimization can be recomputed without any need for kernel matrix inversions, at the cost of saving intermediate states of the training phase. We also offer different ideas to extend this method to an approximate unlearning scheme, even further improving its computational complexity. / Införandet av Dataskyddsförordningen (DSF) i Europa 2018, inklusive rätten att bli bortglömd, ställer viktiga frågor om nödvändigheten av effektiva dataraderingtekniker för tränade maskininlärningsmodeller för att följa denna rättighet, detta eftersom omskolning från grunden av tränade modeller när en datapunkt måste raderas verkar opraktiskt. Vi tacklar dataraderingsproblemet för regression av Gaussiska processer och vi definierar i detta dokument en effektiv exakt avlärningsteknik för Gaussisk process regression som inkluderar optimeringen av kärnfunktionens hyperparametrarna. Metoden är baserad på en effektiv omberäkning av tidigare optimeringar genom Resilient Backpropagation (Rprop)-algoritmen tack vare onlineformuleringen medelst en Gaussisk processregression. Dessutom utvecklar vi en utvidgning av den föreslagna metoden till produkter-av-experter och Bayesianska kommittémaskiner av lokala approximationer av Gaussiska processregression, för att förbättra avlärningskapaciteten genom att använda en slumpmässig partitionering av datasetet. Metodernas prestanda beror till stor del på regressionsuppgiften. Vi visar med flera experiment på olika problem att flera iterationer av optimeringarna kan omberäknas utan behov av kärnmatrisinversioner, men på bekostnad av att spara mellanstatus i träningsfasen. Vi föreslår också olika idéer för att utvidga denna metod till en approximativ avlärningsteknik, för att förbättra dess beräkningskomplexitet. / L’établissement du Règlement Général sur la Protection des Données (RGPD) en Europe en 2018, incluant le "Droit à l’Oubli" pose de sérieuses questions vis-à-vis de l’importance du développement de techniques permettant le "désapprentissage" de données specifiques d’un modéle entrainé. Réentrainer un modèle "from scratch" dés qu’une donnée doit être supprimée pose problème en pratique, ce qui justifie le besoin de méthodes plus efficaces pour répondre à ce problème. Nous abordons ce problème dans le contexte d’une Gaussian Process Regression, et définissons dans ce rapport une méthode efficace et exacte de désapprentissage pour une Gaussian Process Regression incluant l’optimisation des hyperparamètres du noyau. La méthode est basée sur un traçage efficace de l’optimisation faite par l’algorithme de Resilient Backpropagation (Rprop) grâce à la formulation Online d’une Gaussian Process Regression. De plus, nous développons une extension de cette première méthode pour la rendre applicable à des approximations locales telles que les Product-of-Experts ou Bayesian Committee Machines, ce qui permet d’améliorer d’avantage les performance de désapprentissage grâce à partitionement aléatoire du jeu de données. Du fait de la forte dépendence des performances de désapprentissage à la tâche de regression, nous montrons à travers de multiples expériences sur différents jeux de données qu’un nombre conséquent d’itérations peut être recalculé efficacement sans nécessiter d’inversion de matrices, au prix de la sauvegarde des états intermédiaires de la phase d’apprentissage.Nous donnons finalement des idées pour étendre cette méthode vers un désapprentissage approximatif, afin d’améliorer une fois de plus le temps de désapprentissage.

A l’occasion de deux études de cas, une démarche de tableau de bord stratégique hospitalier et une réorganisation de direction des systèmes d’information dans une entreprise de services, nous avons, dans cette recherche, analysé les freins qui apparaissent dans les démarches de changement technologique et managérial.Nous avons montré dans le premier cas comment le modèle du Balanced Scorecard peut être adapté pour piloter un programme de projets et promouvoir un travail coopératif, sous condition d’une évolution du rôle des dirigeants. Le deuxième cas nous a permis de montrer comment les démarches qualité s’enferment dans une logique de qualité-conformité contre-productive qui n’est pas due au manque de formation ou d’implication. Dans les deux cas, nous avons constaté que l’insuffisance de capacité méthodologique des ingénieurs et des managers renforce leur résistance au changement et leurs difficultés en situation complexe. Nous avons montré que la résistance organisationnelle se manifeste en deux temps : des difficultés de premier niveau, dérives nomopathes, apparaissent sous la forme d’un appauvrissement méthodologique du modèle choisi, puis les actions correctives sont suivies de difficultés de deuxième niveau constituant un phénomène de déviance computo-cogitationnelle, sorte de résilience organisationnelle qui voit la transformation de l’état final attendu en un état final hybride. Ces acquis méthodologiques nous ont permis, après une analyse historique de l’évolution des concepts de la qualité puis sur l’évolution du rôle des managers, de dresser des constats à un niveau méso reliant le niveau macro-économique (domination financière dans les entreprises, révolution numérique, renouveau bureaucratique) au niveau micro des entreprises (injonctions paradoxales, juridicisation de la société et conformité généralisée). Nous avons enfin proposé plusieurs modélisations, notamment le concept cyclique d’information-valeur, la figure tripolaire d’un Ingénieur-Stratège et l’ébauche d’un futur modèle de management des intangibles. / The two case studies concern an approach of the strategic dashboard of a hospital and the reorganization of an IT department in a services company. We have, in this research, analyzed the brakes which appear in the technological and managerial change approaches. In the first study case, we have showed how the Balanced Scorecard model can be adapted to pilot a projects program and to promote a cooperative way of working, under the condition of an evolution of the managers role. In the second case we have shown how the quality approaches lock themselves in a counter-productive logical of quality-conformity which is not due to a lack of training or implication. In the both cases we have noticed that the insufficiency of methodological capacity of the managers strengthens their resistance to change and their difficulties in complex situation We have noticed that the organizational resistance express itself in two times : first level difficulties, nomopathic drifts, appear as a methodological impoverishment of the choosen model, then the corrective actions are followed by second level difficulties who form a phenomena of computo-cogitational deviancy, a kind of organizational resistance who transforms the awaited final state to a hybrid final state. These methodological grants, after a historical analysis of the evolution of quality concepts then of the evolution of the managerial roles, succeeded to meso level reports joining the macro-economic level (financial domination in companies, numeric revolution, bureaucratic revival) to the micro-economic level in companies (paradoxical orders, juridical transformation of society and generalized conformity). A last, several modellings are proposed, in particular the cyclic concept of information-value, and the three-pole figure of the engineer-strategist, towards the sketch of a future model of management of intangibles.

This master essay starts with the author having an encounter with two 14-years old boys who claim that they are Nazis. In a school project where the pupils made films about norms in the society they made a film with a clearly racist and homophobic message. As a freelance artist the author was a part of a school development program in aesthetic learning, and the assignment in the actual school was supposed to teach the pupils about norms and how they effect people on an everyday basis. All the films fulfilled this purpose, but the actual film did something more: it revealed the zerotolerance rule in this particular school as a norm that silenced not only these boys, but also any pupil having a different opinion than the teachers. The situation described took place in a classroom in primary school in the area of Stockholm some years ago and even though the film was never shown to anybody that could be offended except the author herself, the author left the school with a feeling of total failure, and two questions craving for answers: Why couldn’t she, nor the teachers, find the time and space to meet the boys, taking their invitation to discuss the zere-tolerance-norm seriously? And what role did the fact that the school had a “zero-tolerance-for-racism rule” play in the cultivation of the boy’s feelings of exclusion and in the way the teachers treated their obvious need for recognition as sane and (soon-to-be) grown up men? Using Hannah Arendt’s ideas of action as the fundament of the investigation, the author poses questions about what space for action there is in the daily work of a teacher today, but also what happens when there is no room for action, when we become the blind administrators of homo faber. The method of the study includes 1) a dialogue seminar with teachers from elementary school, 2) examples and reflections from the author’s own teaching practice at the teacher education and 3) a philosophical investigation focusing on the concepts of aesthetic experience, aesthetic learning, not-knowing and unlearning. In dialogue with Sarah Ahmed, John Dewey, Hans-Georg Gadamer, Maurice Merleau-Ponty, Jacques Rancière, Cusanus and Jonna Bornemark, she further tries to understand what role the film, as an aesthetic experience, played in the example with the Nazi boys. Is there a radical possibility in art that can create space for action in Arendt’s sense? Further, in dialogue with the poet Wiszlava Szymborska, the clown Nalla Laanela, and two novelists; Rachel Cusk and Albert Camus, she tries to understand the possibilities that lie within aesthetic learning when it comes to re-thinking the role of the teacher aiming for a sustainable approach to teaching and learning in a society where teachers work themselves sick. / I inledningen till denna masteruppsats möter författaren två fjortonåriga pojkar som säger att de är nazister. I ett skolprojekt där eleverna skulle göra kortfilmer som gestaltade samhällsnormer gjorde de här pojkarna en film med ett tydligt rasistiskt och homofobiskt budskap. Författaren hade, i egenskap av frilansande konstnär inom skolutvecklingsprogrammet Skapa och Lära, uppdraget att leda eleverna i arbetet med filmerna och syftet var att eleverna skulle lära sig något om de olika normer som påverkar oss människor varje dag. Alla filmer uppfyllde syftet, men denna film gjorde något mer: den fick den här specifika skolans noll-tolerans-mot-rasism-regel att framstå som en norm som tystade inte bara de här två pojkarna, utan alla elever med –ur lärarnas perspektiv -avvikande åsikter. Den ovan beskrivna situationen utspelade sig i ett klassrum i Stockholmsområdet för ett antal år sedan och även om filmen aldrig visades för någon som hade kunnat ta illa upp –förutom författaren själv –lämnade författaren skolan med en känsla av totalt misslyckande och två frågor ringande i öronen: Varför kunde inte hon, och ingen av lärarna i skolan, ta sig tid att möta pojkarna genom att ta deras inbjudan till samtal om noll-tolerans-normen på allvar? Och vilken roll spelade det faktum att skolan hade en noll-tolerans-mot-rasism-regel i kultiverandet av pojkarnas känslor av exkludering och i lärarnas hanterande av deras uppenbara behov av erkännande som sunda, snart vuxna unga män? Med utgångspunkt i Hannah Arendts idéer om handlandet ställer författaren frågor om vilket handlingsutrymme lärare i dagens skola har, men hon undersöker också vad som händer när det inte finns något handlingsutrymme, när läraren blir en blind administratör, en homo faber. Metodologiskt använder sig författaren av 1) ett dialogseminarium med lärare från grundskolan, 2) exempel och reflektioner från hennes egen lärarpraktik, och 3) en filosofisk undersökning med fokus på begreppen estetisk erfarenhet, estetiska lärprocesser, icke-vetande och av-lärande. I dialog med Sarah Ahmed, John Dewey, Hans-Georg Gadamer, Maurice Merleau-Ponty, Jacques Rancière, Cusanus och Jonna Bornemark, försöker hon vidare förstå vilken roll filmen som en estetisk erfarenhet spelade I exemplet med de två nazistpojkarna. Finns det en radikal potential i konstnärliga uttryck som kan skapa handlingsutrymme i Arendts mening? Vidare,i dialog med poeten Wiszlava Szymborska, clownen Nalle Laanela, och två romanförfattare; Rachel Cusk and Albert Camus, försöker författaren utröna vilka möjligheter som kan rymmas i estetiska lärprocesser vad gäller att tänka om (om-tänka) lärarrollen med målet att finna ett hållbart förhållningssätt till undervisning och lärande i ett samhälle där lärare arbetar sig sjuka.