State-of-the-art frameworks for generating approximate circuits automatically explore the search space in an iterative process - often greedily. Synthesis and verification processes are invoked in each iteration to evaluate the found solutions and to guide the search algorithm. As a result, a large number of approximate circuits is subjected to analysis - leading to long runtimes - but only a few approximate circuits might form an acceptable solution. In this paper, we present our Jump Search (JS) method which seeks to reduce the runtime of an approximation process by reducing the number of expensive synthesis and verification steps. To reduce the runtime, JS computes impact factors for each approximation candidate in the circuit to create a selection of approximate circuits without invoking synthesis or verification processes. We denote the selection as path from which JS determines the final solution. In our experimental results, JS achieved speed-ups of up to 57x while area savings remain comparable to the reference search method, Simulated Annealing.

Multi-talker speech and moving speakers still pose a significant challenge to automatic speech recognition systems. Assuming an enrollment utterance of the target speakeris available, the so-called SpeakerBeam concept has been recently proposed to extract the target speaker from a speech mixture. If multi-channel input is available, spatial properties of the speaker can be exploited to support the source extraction. In this contribution we investigate different approaches to exploit such spatial information. In particular, we are interested in the question, how useful this information is if the target speaker changes his/her position. To this end, we present a SpeakerBeam-based source extraction network that is adapted to work on moving speakers by recursively updating the beamformer coefficients. Experimental results are presented on two data sets, one with articially created room impulse responses, and one with real room impulse responses and noise recorded in a conference room. Interestingly, spatial features turn out to be advantageous even if the speaker position changes.

Ranking plays a central role in a large number of applications driven by RDF knowledge graphs. Over the last years, many popular RDF knowledge graphs have grown so large that rankings for the facts they contain cannot be computed directly using the currently common 64-bit platforms. In this paper, we tackle two problems: Computing ranks on such large knowledge bases efficiently and incrementally. First, we present D-HARE, a distributed approach for computing ranks on very large knowledge graphs. D-HARE assumes the random surfer model and relies on data partitioning to compute matrix multiplications and transpositions on disk for matrices of arbitrary size. Moreover, the data partitioning underlying D-HARE allows the execution of most of its steps in parallel. As very large knowledge graphs are often updated periodically, we tackle the incremental computation of ranks on large knowledge bases as a second problem. We address this problem by presenting I-HARE, an approximation technique for calculating the overall ranking scores of a knowledge without the need to recalculate the ranking from scratch at each new revision. We evaluate our approaches by calculating ranks on the 3 × 10^9 and 2.4 × 10^9 triples from Wikidata resp. LinkedGeoData. Our evaluation demonstrates that D-HARE is the first holistic approach for computing ranks on very large RDF knowledge graphs. In addition, our incremental approach achieves a root mean squared error of less than 10E−7 in the best case. Both D-HARE and I-HARE are open-source and are available at: https://github.com/dice-group/incrementalHARE.

Existing tools for automated machine learning, such as Auto-WEKA, TPOT, auto-sklearn, and more recently ML-Plan, have shown impressive results for the tasks of single-label classification and regression. Yet, there is only little work on other types of machine learning problems so far. In particular, there is almost no work on automating the engineering of machine learning solutions for multi-label classification (MLC). We show how the scope of ML-Plan, an AutoML-tool for multi-class classification, can be extended towards MLC using MEKA, which is a multi-label extension of the well-known Java library WEKA. The resulting approach recursively refines MEKA's multi-label classifiers, nesting other multi-label classifiers for meta algorithms and single-label classifiers provided by WEKA as base learners. In our evaluation, we find that the proposed approach yields strong results and performs significantly better than a set of baselines we compare with.

In this paper we present our audio tagging system for the DCASE 2019 Challenge Task 2. We propose a model consisting of a convolutional front end using log-mel-energies as input features, a recurrent neural network sequence encoder and a fully connected classifier network outputting an activity probability for each of the 80 considered event classes. Due to the recurrent neural network, which encodes a whole sequence into a single vector, our model is able to process sequences of varying lengths. The model is trained with only little manually labeled training data and a larger amount of automatically labeled web data, which hence suffers from label noise. To efficiently train the model with the provided data we use various data augmentation to prevent overfitting and improve generalization. Our best submitted system achieves a label-weighted label-ranking average precision (lwlrap) of 75.5% on the private test set which is an absolute improvement of 21.7% over the baseline. This system scored the second place in the teams ranking of the DCASE 2019 Challenge Task 2 and the fifth place in the Kaggle competition “Freesound Audio Tagging 2019” with more than 400 participants. After the challenge ended we further improved performance to 76.5% lwlrap setting a new state-of-the-art on this dataset.

This paper deals with multi-channel speech recognition in scenarios with multiple speakers. Recently, the spectral characteristics of a target speaker, extracted from an adaptation utterance, have been used to guide a neural network mask estimator to focus on that speaker. In this work we present two variants of speakeraware neural networks, which exploit both spectral and spatial information to allow better discrimination between target and interfering speakers. Thus, we introduce either a spatial preprocessing prior to the mask estimation or a spatial plus spectral speaker characterization block whose output is directly fed into the neural mask estimator. The target speaker’s spectral and spatial signature is extracted from an adaptation utterance recorded at the beginning of a session. We further adapt the architecture for low-latency processing by means of block-online beamforming that recursively updates the signal statistics. Experimental results show that the additional spatial information clearly improves source extraction, in particular in the same-gender case, and that our proposal achieves state-of-the-art performance in terms of distortion reduction and recognition accuracy.

We present an unsupervised training approach for a neural network-based mask estimator in an acoustic beamforming application. The network is trained to maximize a likelihood criterion derived from a spatial mixture model of the observations. It is trained from scratch without requiring any parallel data consisting of degraded input and clean training targets. Thus, training can be carried out on real recordings of noisy speech rather than simulated ones. In contrast to previous work on unsupervised training of neural mask estimators, our approach avoids the need for a possibly pre-trained teacher model entirely. We demonstrate the effectiveness of our approach by speech recognition experiments on two different datasets: one mainly deteriorated by noise (CHiME 4) and one by reverberation (REVERB). The results show that the performance of the proposed system is on par with a supervised system using oracle target masks for training and with a system trained using a model-based teacher.

We propose a training scheme to train neural network-based source separation algorithms from scratch when parallel clean data is unavailable. In particular, we demonstrate that an unsupervised spatial clustering algorithm is sufficient to guide the training of a deep clustering system. We argue that previous work on deep clustering requires strong supervision and elaborate on why this is a limitation. We demonstrate that (a) the single-channel deep clustering system trained according to the proposed scheme alone is able to achieve a similar performance as the multi-channel teacher in terms of word error rates and (b) initializing the spatial clustering approach with the deep clustering result yields a relative word error rate reduction of 26% over the unsupervised teacher.

In this paper we consider human daily activity recognition using an acoustic sensor network (ASN) which consists of nodes distributed in a home environment. Assuming that the ASN is permanently recording, the vast majority of recordings is silence. Therefore, we propose to employ a computationally efficient two-stage sound recognition system, consisting of an initial sound activity detection (SAD) and a subsequent sound event classification (SEC), which is only activated once sound activity has been detected. We show how a low-latency activity detector with high temporal resolution can be trained from weak labels with low temporal resolution. We further demonstrate the advantage of using spatial features for the subsequent event classification task.

Despite the strong modeling power of neural network acoustic models, speech enhancement has been shown to deliver additional word error rate improvements if multi-channel data is available. However, there has been a longstanding debate whether enhancement should also be carried out on the ASR training data. In an extensive experimental evaluation on the acoustically very challenging CHiME-5 dinner party data we show that: (i) cleaning up the training data can lead to substantial error rate reductions, and (ii) enhancement in training is advisable as long as enhancement in test is at least as strong as in training. This approach stands in contrast and delivers larger gains than the common strategy reported in the literature to augment the training database with additional artificially degraded speech. Together with an acoustic model topology consisting of initial CNN layers followed by factorized TDNN layers we achieve with 41.6% and 43.2% WER on the DEV and EVAL test sets, respectively, a new single-system state-of-the-art result on the CHiME-5 data. This is a 8% relative improvement compared to the best word error rate published so far for a speech recognizer without system combination.

The weighted prediction error (WPE) algorithm has proven to be a very successful dereverberation method for the REVERB challenge. Likewise, neural network based mask estimation for beamforming demonstrated very good noise suppression in the CHiME 3 and CHiME 4 challenges. Recently, it has been shown that this estimator can also be trained to perform dereverberation and denoising jointly. However, up to now a comparison of a neural beamformer and WPE is still missing, so is an investigation into a combination of the two. Therefore, we here provide an extensive evaluation of both and consequently propose variants to integrate deep neural network based beamforming with WPE. For these integrated variants we identify a consistent word error rate (WER) reduction on two distinct databases. In particular, our study shows that deep learning based beamforming benefits from a model-based dereverberation technique (i.e. WPE) and vice versa. Our key findings are: (a) Neural beamforming yields the lower WERs in comparison to WPE the more channels and noise are present. (b) Integration of WPE and a neural beamformer consistently outperforms all stand-alone systems.

This contribution presents a speech enhancement system for the CHiME-5 Dinner Party Scenario. The front-end employs multi-channel linear time-variant filtering and achieves its gains without the use of a neural network. We present an adaptation of blind source separation techniques to the CHiME-5 database which we call Guided Source Separation (GSS). Using the baseline acoustic and language model, the combination of Weighted Prediction Error based dereverberation, guided source separation, and beamforming reduces the WER by 10:54% (relative) for the single array track and by 21:12% (relative) on the multiple array track.

NARA-WPE is a Python software package providing implementations of the weighted prediction error (WPE) dereverberation algorithm. WPE has been shown to be a highly effective tool for speech dereverberation, thus improving the perceptual quality of the signal and improving the recognition performance of downstream automatic speech recognition (ASR). It is suitable both for single-channel and multi-channel applications. The package consist of (1) a Numpy implementation which can easily be integrated into a custom Python toolchain, and (2) a TensorFlow implementation which allows integration into larger computational graphs and enables backpropagation through WPE to train more advanced front-ends. This package comprises of an iterative offline (batch) version, a block-online version, and a frame-online version which can be used in moderately low latency applications, e.g. digital speech assistants.

Arbitrary sampling rate conversion has already received considerable attention in the past, but still lacks an equivalent representation of the effective time-dilation process in the block frequency domain. Good sampling rate converters in the time domain have been known, for instance, in terms of time-varying 'Sinc' or fixed 'Farrow' polynomial filters. The former can deliver nearly exact conversion at high complexity, while the latter has pronounced computational efficiency with limited accuracy. Only recently, it was shown that a composite 'polyphase Farrow' form with high resampling precision can be implemented with quasi-fixed filters that operate at the input sampling rate. We therefore propose to capitalize from that fixed-filter architecture in that we translate the polyphase-Farrow filters into an equivalent FFT-based overlap-save form. Experimental evaluation and comparison with other state-of-the art frequency-domain approaches then proves currently the best price-performance ratio of the proposed algorithm. It is thus an ideal candidate for the new framework of acoustic sensor networks that critically rests upon fast and accurate alignment of autonomous sampling processes.

A framework for set‐oriented multiobjective optimal control of partial differential equations using reduced order modeling has recently been developed [1]. Following concepts from localized reduced bases methods, error estimators for the reduced cost functionals are utilized to construct a library of locally valid reduced order models. This way, a superset of the Pareto set can efficiently be computed while maintaining a prescribed error bound. In this article, this algorithm is applied to a problem with non‐smooth objective functionals. Using an academic example, we show that the extension to non‐smooth problems can be realized in a straightforward manner. We then discuss the implications on the numerical results.

In multinomial classification, reduction techniques are commonly used to decompose the original learning problem into several simpler problems. For example, by recursively bisecting the original set of classes, so-called nested dichotomies define a set of binary classification problems that are organized in the structure of a binary tree. In contrast to the existing one-shot heuristics for constructing nested dichotomies and motivated by recent work on algorithm configuration, we propose a genetic algorithm for optimizing the structure of such dichotomies. A key component of this approach is the proposed genetic representation that facilitates the application of standard genetic operators, while still supporting the exchange of partial solutions under recombination. We evaluate the approach in an extensive experimental study, showing that it yields classifiers with superior generalization performance.

We revisit algorithm selection for declarative programming solvers. We introduce two main ideas to improve cost-sensitive hierarchical clustering: First, to augment the portfolio builder with a self-configuration component. And second, we propose that the algorithm selector assesses the confidence level of its own prediction, so that a more defensive recourse action can be used to overturn the original recommendation.

In scientific computing, unstructured meshes are a crucial foundation for the simulation of real-world physical phenomena. Compared to regular grids, they allow resembling the computational domain with a much higher accuracy, which in turn leads to more efficient computations.<br />There exists a wealth of supporting libraries and frameworks that aid programmers with the implementation of applications working on such grids, each built on top of existing parallelization technologies. However, many approaches require the programmer to introduce a different programming paradigm into their application or provide different variants of the code. SYCL is a new programming standard providing a remedy to this dilemma by building on standard C ++17 with its so-called single-source approach: Programmers write standard C ++ code and expose parallelism using C++17 keywords. The application is<br />then transformed into a concrete implementation by the SYCL implementation. By encapsulating the OpenCL ecosystem, different SYCL implementations enable not only the programming of CPUs but also of heterogeneous platforms such as GPUs or other devices. For the first time, this paper showcases a SYCL-<br />based solver for the nodal Discontinuous Galerkin method for Maxwell’s equations on unstructured meshes. We compare our solution to a previous C-based implementation with respect to programmability and performance on heterogeneous platforms.<br