@inproceedings{63157,
  abstract     = {{Three-phase cascaded H-bridge converters (CHBs) in star configuration require reliable current controllers to evenly charge the module DC-link capacitors. Conventionally, a current control in dq-coordinates is utilized. At steady state, the resulting calculated reference arm voltages are sinusoidal, have identical amplitudes and show a phase shift of 120 degree to each other. For balanced grid inductors, the resulting grid currents also have the same amplitude. However, own simulations show that unbalanced grid inductors always lead to different grid current amplitudes (4% difference in this case). As a result, the averaged charging module powers differ and the peak DC-link capacitor voltage rises as well. In the first step, an adaptation of an existing zero-sequence voltage injection is proposed. For balanced grid inductors, it converges to the 3rd harmonic voltage injection which can reduce the peak-to-peak DC-link voltage ripple up by to 50% and balances the power between the phases. However, unbalanced grid inductors still lead to the same unbalanced grid currents of 4%. Therefore, a new method with 4 integrators based on linear regression is proposed to achieve sinusoidal grid currents for unbalanced inductors. The proposed method has a similar transient dynamic as the conventional dq control, but balances the grid currents nearly ideally. Simulation results of a 1MW cascaded H bridge and a scaled-down prototype verify the proposed method.}},
  author       = {{Unruh, Roland and Böcker, Joachim and Schafmeister, Frank}},
  booktitle    = {{2025 Energy Conversion Congress & Expo Europe (ECCE Europe)}},
  keywords     = {{Cascaded H-Bridge, Current Control, dq Transformation, Linear Regression, Unbalanced Inductors}},
  location     = {{Birmingham, United Kingdom}},
  publisher    = {{IEEE}},
  title        = {{{Three-Phase Instantaneous Current Controller for Unbalanced Grid Inductors Without DQ Transform for Cascaded H-Bridge Converters}}},
  doi          = {{10.1109/ecce-europe62795.2025.11238538}},
  year         = {{2025}},
}

@inproceedings{27506,
  abstract     = {{Explainability for machine learning gets more and more important in high-stakes decisions like real estate appraisal. While traditional hedonic house pricing models are fed with hard information based on housing attributes, recently also soft information has been incorporated to increase the predictive performance. This soft information can be extracted from image data by complex models like Convolutional Neural Networks (CNNs). However, these are intransparent which excludes their use for high-stakes financial decisions. To overcome this limitation, we examine if a two-stage modeling approach can provide explainability. We combine visual interpretability by Regression Activation Maps (RAM) for the CNN and a linear regression for the overall prediction. Our experiments are based on 62.000 family homes in Philadelphia and the results indicate that the CNN learns aspects related to vegetation and quality aspects of the house from exterior images, improving the predictive accuracy of real estate appraisal by up to 5.4%.}},
  author       = {{Kucklick, Jan-Peter}},
  booktitle    = {{55th Annual Hawaii International Conference on System Sciences (HICSS-55)}},
  keywords     = {{Explainable Artificial Intelligence (XAI), Regression Activation Maps, Real Estate Appraisal, Convolutional Block Attention Module, Computer Vision}},
  location     = {{Virtual}},
  title        = {{{Visual Interpretability of Image-based Real Estate Appraisal}}},
  year         = {{2022}},
}

@inproceedings{27507,
  abstract     = {{Accurate real estate appraisal is essential in decision making processes of financial institutions, governments, and trending real estate platforms like Zillow. One of the most important factors of a property’s value is its location. However, creating accurate quantifications of location remains a challenge. While traditional approaches rely on Geographical Information Systems (GIS), recently unstructured data in form of images was incorporated in the appraisal process, but text data remains an untapped reservoir. Our study shows that using text data in form of geolocated Wikipedia articles can increase predictive performance over traditional GIS-based methods by 8.2% in spatial out-of-sample validation. A framework to automatically extract geographically weighted vector representations for text is established and used alongside traditional structural housing features to make predictions and to uncover local patterns on sale price for real estate transactions between 2015 and 2020 in Allegheny County, Pennsylvania.}},
  author       = {{Heuwinkel, Tim and Kucklick, Jan-Peter and Müller, Oliver}},
  booktitle    = {{55th Annual Hawaii International Conference on System Sciences (HICSS-55)}},
  keywords     = {{Real Estate Appraisal, Text Regression, Natural Language Processing (NLP), Location Intelligence, Wikipedia}},
  location     = {{Virtual}},
  title        = {{{Using Geolocated Text to Quantify Location in Real Estate Appraisal}}},
  year         = {{2022}},
}

@inproceedings{48849,
  abstract     = {{One-shot optimization tasks require to determine the set of solution candidates prior to their evaluation, i.e., without possibility for adaptive sampling. We consider two variants, classic one-shot optimization (where our aim is to find at least one solution of high quality) and one-shot regression (where the goal is to fit a model that resembles the true problem as well as possible). For both tasks it seems intuitive that well-distributed samples should perform better than uniform or grid-based samples, since they show a better coverage of the decision space. In practice, quasi-random designs such as Latin Hypercube Samples and low-discrepancy point sets are indeed very commonly used designs for one-shot optimization tasks. We study in this work how well low star discrepancy correlates with performance in one-shot optimization. Our results confirm an advantage of low-discrepancy designs, but also indicate the correlation between discrepancy values and overall performance is rather weak. We then demonstrate that commonly used designs may be far from optimal. More precisely, we evolve 24 very specific designs that each achieve good performance on one of our benchmark problems. Interestingly, we find that these specifically designed samples yield surprisingly good performance across the whole benchmark set. Our results therefore give strong indication that significant performance gains over state-of-the-art one-shot sampling techniques are possible, and that evolutionary algorithms can be an efficient means to evolve these.}},
  author       = {{Bossek, Jakob and Doerr, Carola and Kerschke, Pascal and Neumann, Aneta and Neumann, Frank}},
  booktitle    = {{Parallel Problem Solving from Nature (PPSN XVI)}},
  isbn         = {{978-3-030-58111-4}},
  keywords     = {{Continuous optimization, Fully parallel search, One-shot optimization, Regression, Surrogate-assisted optimization}},
  pages        = {{111–124}},
  publisher    = {{Springer-Verlag}},
  title        = {{{Evolving Sampling Strategies for One-Shot Optimization Tasks}}},
  doi          = {{10.1007/978-3-030-58112-1_8}},
  year         = {{2020}},
}

@article{11846,
  abstract     = {{In this paper, we present a new technique for automatic speech recognition (ASR) in reverberant environments. Our approach is aimed at the enhancement of the logarithmic Mel power spectrum, which is computed at an intermediate stage to obtain the widely used Mel frequency cepstral coefficients (MFCCs). Given the reverberant logarithmic Mel power spectral coefficients (LMPSCs), a minimum mean square error estimate of the clean LMPSCs is computed by carrying out Bayesian inference. We employ switching linear dynamical models as an a priori model for the dynamics of the clean LMPSCs. Further, we derive a stochastic observation model which relates the clean to the reverberant LMPSCs through a simplified model of the room impulse response (RIR). This model requires only two parameters, namely RIR energy and reverberation time, which can be estimated from the captured microphone signal. The performance of the proposed enhancement technique is studied on the AURORA5 database and compared to that of constrained maximum-likelihood linear regression (CMLLR). It is shown by experimental results that our approach significantly outperforms CMLLR and that up to 80\% of the errors caused by the reverberation are recovered. In addition to the fact that the approach is compatible with the standard MFCC feature vectors, it leaves the ASR back-end unchanged. It is of moderate computational complexity and suitable for real time applications.}},
  author       = {{Krueger, Alexander and Haeb-Umbach, Reinhold}},
  journal      = {{IEEE Transactions on Audio, Speech, and Language Processing}},
  keywords     = {{ASR, AURORA5 database, automatic speech recognition, Bayesian inference, belief networks, CMLLR, computational complexity, constrained maximum likelihood linear regression, least mean squares methods, LMPSC computation, logarithmic Mel power spectrum, maximum likelihood estimation, Mel frequency cepstral coefficients, MFCC feature vectors, microphone signal, minimum mean square error estimation, model-based feature enhancement, regression analysis, reverberant speech recognition, reverberation, RIR energy, room impulse response, speech recognition, stochastic observation model, stochastic processes}},
  number       = {{7}},
  pages        = {{1692--1707}},
  title        = {{{Model-Based Feature Enhancement for Reverberant Speech Recognition}}},
  doi          = {{10.1109/TASL.2010.2049684}},
  volume       = {{18}},
  year         = {{2010}},
}

@article{11778,
  abstract     = {{In this paper, it is shown that a correlation criterion is the appropriate criterion for bottom-up clustering to obtain broad phonetic class regression trees for maximum likelihood linear regression (MLLR)-based speaker adaptation. The correlation structure among speech units is estimated on the speaker-independent training data. In adaptation experiments the tree outperformed a regression tree obtained from clustering according to closeness in acoustic space and achieved results comparable with those of a manually designed broad phonetic class tree}},
  author       = {{Haeb-Umbach, Reinhold}},
  journal      = {{IEEE Transactions on Speech and Audio Processing}},
  keywords     = {{acoustic space, adaptation experiments, automatic generation, bottom-up clustering, broad phonetic class regression trees, correlation criterion, correlation methods, maximum likelihood estimation, maximum likelihood linear regression based speaker adaptation, MLLR adaptation, pattern clustering, phonetic regression class trees, speaker-independent training data, speech recognition, speech units, statistical analysis, trees (mathematics)}},
  number       = {{3}},
  pages        = {{299--302}},
  title        = {{{Automatic generation of phonetic regression class trees for MLLR adaptation}}},
  doi          = {{10.1109/89.906003}},
  volume       = {{9}},
  year         = {{2001}},
}