Introduction

In the previous chapter, we ended with the question whether flexibility stigma exists across all countries, and whether in all countries women will be the ones who suffer more from its prevalence. Given that norms around the ‘ideal worker’ are different across countries, and as countries differ in the extent to which traditional gender roles exist, we can expect some variations across countries. Flexible working is not used in a vacuum, and the socio-economic, cultural and institutional context in which it is used matter. As we have discussed in the previous chapters, according to capabilities approach theories, a person's capacity to use the ‘freedom’ given to oneself is limited by the context in which that individual is embedded (Hobson, 2011). The same could be found if we examine Foucault's (2010) theory of the subjectification of self and the rise of the homo-economicus which enables the flexibility paradox to occur. The crux of the argument lies in the context of widespread neo-liberalism and the shifts found in societal norms – and the individual's own identity – towards one that privilege capitalist market exchange values above all else. However, there are a variety of capitalisms (Hall and Soskice, 2001) and neo-liberalistic ideals are not as prevalent across all countries. In fact, examining some of the evidence of the flexibility and autonomy paradox, we see that most previous studies are from countries that are typically considered liberal countries (Esping-Andersen, 1990). The question arises then whether we would not see similar patterns in other countries where norms around work and work-life balance are very different. Similarly, we can expect to find variations in the degree to which the gendered flexibility paradox occurs across countries. One main reason why we expect to find and do find gendered patterns of the flexibility paradox was largely due to the patriarchal societal structures with strict gender norms around men and women's roles. Thus, in countries where such traditional gender norms do not exist, the gendered flexibility paradox may also not occur.

Introduction

This book set out to explore flexible working in a more critical way, asking the question whether flexible working actually provides positive outcomes for workers in terms of work-life balance, workers’ well-being and gender equality as many expect it to. The results of the previous chapters show that paradoxically rather than improving workers’ work-life balance, flexible working increased feelings of conflict between work and family. The reason behind this phenomenon was explained through the flexibility paradox, that flexible working can lead to further exploitation of workers’ labour. This exploitation pattern is gendered. Men expanded their employment hours, namely overtime hours, to fulfil their ideal worker and breadwinner masculine image. Women expanded their unpaid working hours, namely increased time spent on housework and childcare adhering to the social norms around their roles as caregivers. What is more, due to these gendered patterns of flexible working or more so the assumptions behind such patterns, women end up being penalised further when working flexibly despite the fact that they are also likely to work longer and harder on their paid work when working flexibly.

However, I have also shown that the take-up and outcomes of flexible working largely depends on the contexts in which it is used. The way we think about work, work-life balance, and gender roles, workers’ bargaining power and insecurity all help shape the outcomes of flexible working. The book also showed that as flexible working becomes more widely used, we see a shift in the attitudes towards flexible working – namely through the decline in flexibility stigma. The pandemic has provided us with some evidence of this, where the large-scale introduction of homeworking has led to profound changes not only in the perceptions towards and practices of flexible working, but also partly the gendered outcomes of flexible working. However, we still see signs that the flexibility paradox still exists, given that many other cultural normative and institutional factors remain largely the same during this period, and because the pandemic has given rise to higher levels of insecurity among workers which may exacerbate the problem.

Introduction

One of the key findings drawn from the previous chapter was that as flexible working becomes more widespread, people are less likely to hold stigmatised views against flexible workers, and it is less likely to lead to negative outcomes in terms of work-life balance. The results were based on cross-national studies which meant that although we do see strong associations we cannot guarantee the direction of the relationship (for example, which came first, stigma or prevalence of flexible working?). We also cannot be certain if the more widespread use of flexible working or changes in contexts are the real causes or if it has to do with something else we failed to observe. In other words, the question arises whether we would see positive changes to flexible working practices in countries like the UK and the US if we were to change some of the contexts. These are difficult questions to answer given that cultures, policies and the take-up of flexible working do not usually change rapidly enough for us to properly answer them.

Then the COVID-19 pandemic happened and provided us with a very unique experimental opportunity to answer some of these difficult questions: What happens if a large group of workers starts working from home? How would this sudden rise of flexible working change stigmatised views towards flexible workers? How would this change the flexibility paradox patterns we have observed previously? How would this change the gender dynamics of the outcomes of flexible working? Just to clarify, I am not making light of the devastating impact the pandemic had in terms of not only deaths but the health, mental health and economic impact it has had on millions of families. However, given the scope of this book, the COVID-19 pandemic provided us with a once in a lifetime opportunity to better understand how (drastic) changes in contexts may change much of our existing understanding about the nature and outcomes of flexible working. This chapter aims to explore these questions by summarising key studies carried out during the pandemic. The conclusion shows that the widespread use of flexible working helped change the perception towards flexible working to be more positive.

This chapter introduces the background, development history, and typical applications of edge learning. It also specifies the main challenges faced by edge learning from the aspects of data, communication, and computation.

In this chapter, we first provide convergence results of Stochastic Gradient Descent (SGD) methods that are usually adopted to solve the machine learning problem. Then, we introduce advanced training algorithms including momentum SGD, Hyper-parameter-based algorithms, and optimization algorithms for deep learning models. At last, we give theoretical frameworks about how to deal with the staleness gradient incurred by ASP or SSP.

This chapter first focuses on model compression and hardware acceleration for edge learning. It covers many aspects, including the learning algorithms, learning-oriented communication, distributed machine learning with hardware adaptation, TEE-based privacy protection, algorithm, and hardware joint optimization, etc. The essential objective is to implement an integrated algorithm-hardware platform, to optimize the implementation of emerging machine learning algorithms, to fully explore the potential of modern computation hardware, and to promote novel intelligent applications for sophisticated services. Then, we introduce straggler tolerance schemes that can avoid the overall training performance seriously degraded by faulty nodes, and can adequately utilize the computation power of slow nodes. At last, we introduce computation acceleration technologies for inference at the edge.

How to build a benign ecosystem for sustainable development of edge learning is a crucial issue. This chapter first introduces incentive mechanisms for edge learning to motivate edge nodes to contribute model training. Specifically, in parameter server architecture, we introduce a deep reinforcement learning-based (DRL) incentive mechanism to determine the optimal pricing strategy for the parameter server and the optimal training strategies for edge nodes. Finally, we discuss future directions.

Big data and AI are enabling technologies for smart decision-making, automation, and resource optimization. These technologies collectively promote intelligent services from concepts to practical applications. It is widely recognized that Intelligent Services meet the strategic development of emerging industries, meanwhile enrich people’s lifestyle and make a convenient and efficient life. This chapter introduces the popular programming frameworks for Edge Learning. Then, we give some examples of emerging intelligent applications in the edge, e.g., smart health, self-driving, smart surveillance, and smart transportation.

Edge learning has enabled the training of large-scale machine learning models on a big dataset by implementing data parallelism in multiple nodes. However, the iterative interaction generated by multiple learning nodes together with the considerable quantity of communication data on each interaction yields huge communication overhead, which greatly hinders the scalability of Edge Learning. In this chapter, we introduce the mainstream approaches to achieve communication efficiency of edge training, including compressing communication data, reducing the synchronous frequency, overlapping computation and communication, and optimizing the transmission network. Specifically, we propose two hybrid mechanisms for communication-efficient Edge Learning. The first one is QOSP that integrates gradient quantization for communication compression and overlap synchronization parallel for simultaneous computation and communication. The second mechanism improves communication efficiency during the aggregation of client-side updates by quantizing the gradients and exploiting the inherent superposition of radio frequency signals. Finally, we discuss the future directions of communication-efficient edge learning.

In a cloud-edge environment, data are generated by different types of devices, and these devices have various computation capabilities and storage sizes. It is unrealistic to execute all the tasks in the cloud, instead, putting some work into edge servers that are close to end-users would be more reasonable. Edge Learning is a powerful paradigm for big data analytics in the cloud-edge environment. Edge Learning exploits pervasive data generated not only by user devices but also by other sensing devices and those stored in the cloud/edge servers (e.g., data from social networks). Moreover, EL leverages various computing entities (all the devices with computing capabilities ranging from cloud, edge servers, to various edge devices) in an efficient, reliable, and robust manner.

In this chapter, we first introduce the deep learning models that are widely used in Edge Learning. Then, we introduce the basic machine learning algorithms, architectures, and synchronization mode for Edge Learning.

Conventional distributed machine learning manages the training data in a centralized mode without considering the privacy and security problems during training or inference. With the rapid development and wide deployment of artificial intelligence technology these days, privacy protection has gained more and more attention. Moreover, EL participants usually are small devices (e.g., smartphones, sensors) that have weak defense ability and can be easily compromised under possible attacks. In this chapter, we first introduce a security guarantee mechanism in Edge Learning including the defense methods for data-oriented attacks and model-oriented attacks. Then, we summarize the mainstream methods of privacy protection including differential privacy, secure multi-party computation, and homomorphic encryption. Finally, we discuss future directions in this field.

With the growth of model complexity and computational overhead, modern ML applications are usually handled by the distributed systems, where the training procedure is conducted in parallel. Basically, the datasets and models are partitioned to different workers in data parallelism and model parallelism, respectively.

In this chapter, we present the details of these two schemes. Moreover, considering some latest researches that handle distributed training via multiple primitives, we also discuss the extension of training parallelism, i.e., learning frameworks and efficient synchronization mechanisms over the hierarchical architecture.

In Edge Learning, training data are non-independently and identically distributed (non-IID). Applying the same learning strategy for all workers fails to work efficiently. In this chapter, we introduce federated learning, where the training data are always non-IID due to data isolation. Then, we summarize enabling technologies for efficient training with non-IID data. We also propose a reinforcement learning-based method that takes non-IID property and resource-constrain into consideration and adjusts the hyper-parameters to accelerate the loss descent efficiency.