Towards Exploring the Limitations of Active Learning: An Empirical Study

in The 36th IEEE/ACM International Conference on Automated Software Engineering. (2021)

Deep neural networks (DNNs) are being increasingly deployed as integral parts of software systems. However, due to the complex interconnections among hidden layers and massive hyperparameters, DNNs ... [more ▼]

Deep neural networks (DNNs) are being increasingly deployed as integral parts of software systems. However, due to the complex interconnections among hidden layers and massive hyperparameters, DNNs require being trained using a large number of labeled inputs, which calls for extensive human effort for collecting and labeling data. Spontaneously, to alleviate this growing demand, a surge of state-of-the-art studies comes up with different metrics to select a small yet informative dataset for the model training. These research works have demonstrated that DNN models can achieve competitive performance using a carefully selected small set of data. However, the literature lacks proper investigation of the limitations of data selection metrics, which is crucial to apply them in practice. In this paper, we fill this gap and conduct an extensive empirical study to explore the limits of selection metrics. Our study involves 15 selection metrics evaluated over 5 datasets (2 image classification tasks and 3 text classification tasks), 10 DNN architectures, and 20 labeling budgets (ratio of training data being labeled). Our findings reveal that, while selection metrics are usually effective in producing accurate models, they may induce a loss of model robustness (against adversarial examples) and resilience to compression. Overall, we demonstrate the existence of a trade-off between labeling effort and different model qualities. This paves the way for future research in devising selection metrics considering multiple quality criteria. [less ▲]

Electricity Spot Market Design 2030-2050

Report (2021)

Driven by the climate conference in Paris in December 2015 countries worldwide are confronted with the question of how to shape their power system and how to establish alternative technologies to reduce ... [more ▼]

Driven by the climate conference in Paris in December 2015 countries worldwide are confronted with the question of how to shape their power system and how to establish alternative technologies to reduce harmful CO2 emissions. The German government plans that even before the year 2050, all electricity generated and consumed in Germany should be greenhouse gas neutral [1]. To successfully integrate renewable energies, a future energy system must be able to handle the intermittent nature of renewable energy sources such as wind and solar. One important means to address such electricity production variability is demand-side flexibility. Here, industry plays a major role in responding to variable electricity supply with adequate flexibility. This is where the Kopernikus project SynErgie comes in with more than 80 project partners from academia, industry, governmental, and non-governmental organizations as well as energy suppliers and network operators. The Kopernikus project SynErgie investigates how to best leverage demand-side flexibility in the German industry. The current electricity market design in Germany is not well suited to deal with increasing levels of re- newable energy, and it does not embrace demand-side flexibility. Almost 6 GW of curtailed power in 2019 provide evidence that changes are needed with respect to the rules governing electricity markets. These rules were designed at a time when electricity generation was concentrated on a few large and dispatchable conventional power plants and demand was considered inelastic. The SynErgie Cluster IV investigates how a future-proof electricity market design should be organized. The corresponding Work Package IV.3.1 more specifically deals with analyzing and designing allocation and pricing rules on electricity spot markets. The resulting design must be well suited to accommodate demand-side flexibility and address the intermittent nature of important renewable energy sources. This whitepaper is the result of a fruitful collaboration among the partners involved in SynErgie Cluster IV which include Germany’s leading research organizations and practitioners in the field. The collaboration led to an expert workshop in October 2020 with participation from a number of international energy market experts such as Mette Bjørndal (NHH), Endre Bjørndal (NHH), Peter Cramton (University of Maryland and University of Cologne), and Raphael Heffron (University of Dundee). The whitepaper details the key recommendations from this workshop. In particular, the whitepaper recommends a move to a locational, marginal price-based system together with new bidding formats allowing to better express flexibility. We argue in favor of a one-step introduction of locational, marginal prices instead of repeatedly splitting existing zones. Frequent zone splitting involves recurring political debates as well as short- and long-run instabilities affecting the basis for financial con- tracts, for example. Importantly, the definition of stable prize zones is very challenging with increasing levels of distributed and renewable energy sources. The recommendation is the outcome of an intense debate about advantages and downsides of different policy alternatives. However, such a transition to locational, marginal prices is not without challenges, and it is a call to arms for the research community, policymak- ers, and practitioners to develop concepts on how to best facilitate the transition and ensure a reliable and efficient electricity market of the future. We’d like to thank all the project partners and are grateful for the financial support from the Federal Ministry of Education and Research as well as the Project Management Jülich. Hans Ulrich Buhl (Cluster Lead) Martin Bichler (Work Package Lead) [less ▲]

Blockchain Won’t Kill the Banks: Why Disintermediation Doesn’t Work in International Trade Finance

in Communications of the Association for Information Systems (2021), 49

In the financial services industry, many people assume blockchain to have significant impacts. From research and practice, we observe two main paradigms of how organizations interact with blockchain ... [more ▼]

In the financial services industry, many people assume blockchain to have significant impacts. From research and practice, we observe two main paradigms of how organizations interact with blockchain. First, organizations use blockchain to optimize existing processes (blockchain-based business process optimization (BPO)). Second, organizations use blockchain to disrupt existing processes, foster disintermediation, and enable disruptive business models (blockchain-based business process disruption (BPD)). However, we lack scientific research that evaluates its de facto potential. We bridge this gap by following a design science research approach to design blockchain-based business process re-engineering (BPRE) for a letter of credit that combines the advantages of BPO and BPD. We conduct three design cycles and develop three artefacts: a BPO, a BPD, and a BPRE approach. Our BPRE approach combines the advantages of partial disintermediation (i.e., increased efficiency and transparency) with the advantages of intermediaries (i.e., process flexibility, liquidity provision, and dispute mediation). [less ▲]

Nanoscale interfacial engineering enables highly stable and efficient perovskite photovoltaics

in Energy and Environmental Science (2021)

Chapter 17 - Luxembourg

in Griller, Stefan; Lentsch, Elisabeth (Eds.) EMU Integration and Member States' Constituztions (2021)

Written Evidence Submitted to UK Parliament to Influencer Culture Inquiry

Diverse speeches and writings (2021)

The Blockchain Effect: From Inter-Ecosystem to Intra-Ecosystem Competition

in Proceedings of the Twenty-Ninth European Conference on Information Systems (ECIS) (2021)

Blockchains enable distributed operation, decentralized control, and token-based representations of tangible and intangible assets. Organizations commonly use blockchain technology to foster collaboration ... [more ▼]

Blockchains enable distributed operation, decentralized control, and token-based representations of tangible and intangible assets. Organizations commonly use blockchain technology to foster collaboration. In this paper, we investigate the use of blockchain to foster competition. We conduct a single-case study of Germany’s mobility-as-a-service community and its efforts to use blockchain as a technical backbone for mobility ecosystems. The community views blockchain as a technology that embodies organizing principles of empowerment and equality. These principles motivated the community to rethink ecosystem structure. In particular, the community began to question the exclusive, non-adversarial position of mobility service aggregators. We find that rethinking this position might shift their competitive focus from the inter- to the intra-ecosystem level and enables the creation of a larger ecosystem. As a second-order effect, the community began to rethink ecosystem governance. Specifically, it began to explore options for effectively distributed decision making while safeguarding efficiency. [less ▲]

The supply of hours worked and fluctuations between growth regimes

Scientific Conference (2021)

Declining hours of work per worker in conjunction with a growing work force may give rise to fluctuations between growth regimes. This is shown in an overlapping generations model with two-period lived ... [more ▼]

Declining hours of work per worker in conjunction with a growing work force may give rise to fluctuations between growth regimes. This is shown in an overlapping generations model with two-period lived individuals endowed with Boppart-Krusell preferences (Boppart and Krusell (2020)). On the supply side, economic growth is due to the expansion of consumption-good varieties through endogenous research. A sufficiently negative equilibrium elasticity of the individual supply of hours worked to an expansion in the set of consumption-good varieties destabilizes the steady state so that equilibrium trajectories may fluctuate between two growth regimes, one with and the other without an active research sector. Fluctuations affect intergenerational welfare, the evolution of GDP, and the functional income distribution. A stabilization policy can shift the economy onto its steady-state path. Fluctuations arise for empirically reasonable parameter constellations. The economics of fluctuations between growth regimes is linked to the intergenerational trade of shares and their pricing in the asset market. [less ▲]

Power and Bandwidth Minimization for Demand-Aware GEO Satellite Systems

Scientific Conference (2021)

Smart radio resource allocation combined with the recent advances of digital payloads will allow to control the transmit power and bandwidth of the satellites depending on the demand and the channel ... [more ▼]

Smart radio resource allocation combined with the recent advances of digital payloads will allow to control the transmit power and bandwidth of the satellites depending on the demand and the channel conditions of users. The system flexibility is important not only to handle divergent demand requirements but also to efficiently utilize the limited and expensive satellite resources. In this paper, we propose a demand-aware smart radio resource allocation technique, where the transmit power and the bandwidth of the GEO satellite are minimized while satisfying the user demand. The formulated optimization problem is non-convex mixed-integer nonlinear program which is difficult to solve. Hence, we apply a quadratic transform to solve the problem iteratively. The numerical results showed that the proposed scheme outperforms the benchmark schemes in terms of bandwidth utilization while accurately providing capacity-ondemand. [less ▲]

On performance characterization of cascaded multiwire-PLC/MIMO-RF communication system

in 2021 29th European Signal Processing Conference (EUSIPCO) (2021)