Random and Adversarial Bit Error Robustness: Energy-Efficient and Secure DNN Accelerators.

Deep neural network (DNN) accelerators received considerable attention in recent years due to the potential to save energy compared to mainstream hardware. Low-voltage operation of DNN accelerators allows to further reduce energy consumption, however, causes bit-level failures in the memory storing the quantized weights. Furthermore, DNN accelerators are vulnerable to adversarial attacks on voltage controllers or individual bits. In this paper, we show that a combination of robust fixed-point quantization, weight clipping, as well as random bit error training (RandBET) or adversarial bit error training (AdvBET) improves robustness against random or adversarial bit errors in quantized DNN weights significantly. This leads not only to high energy savings for low-voltage operation as well as low-precision quantization, but also improves security of DNN accelerators. In contrast to related work, our approach generalizes across operating voltages and accelerators and does not require hardware changes. Moreover, we present a novel adversarial bit error attack and are able to obtain robustness against both targeted and untargeted bit-level attacks. Without losing more than 0.8%/2% in test accuracy, we can reduce energy consumption on CIFAR10by 20%/30% for 8/4-bit quantization. Allowing up to 320 adversarial bit errors, we reduce test error from above 90% (chance level) to 26.22%.

K2S Challenge: From Undersampled K-Space to Automatic Segmentation.

Magnetic Resonance Imaging (MRI) offers strong soft tissue contrast but suffers from long acquisition times and requires tedious annotation from radiologists. Traditionally, these challenges have been addressed separately with reconstruction and image analysis algorithms. To see if performance could be improved by treating both as end-to-end, we hosted the K2S challenge, in which challenge participants segmented knee bones and cartilage from 8× undersampled k-space. We curated the 300-patient K2S dataset of multicoil raw k-space and radiologist quality-checked segmentations. 87 teams registered for the challenge and there were 12 submissions, varying in methodologies from serial reconstruction and segmentation to end-to-end networks to another that eschewed a reconstruction algorithm altogether. Four teams produced strong submissions, with the winner having a weighted Dice Similarity Coefficient of 0.910 ± 0.021 across knee bones and cartilage. Interestingly, there was no correlation between reconstruction and segmentation metrics. Further analysis showed the top four submissions were suitable for downstream biomarker analysis, largely preserving cartilage thicknesses and key bone shape features with respect to ground truth. K2S thus showed the value in considering reconstruction and image analysis as end-to-end tasks, as this leaves room for optimization while more realistically reflecting the long-term use case of tools being developed by the MR community.

An Analytical Temperature-Dependent Design Model for Contour-Mode MEMS Resonators and Oscillators Verified by Measurements.

The importance of micro-electromechanical systems (MEMS) for radio-frequency (RF) applications is rapidly growing. In RF mobile-communication systems, MEMS-based circuits enable a compact implementation, low power consumption and high RF performance, e.g., bulk-acoustic wave filters with low insertion loss and low noise or fast and reliable MEMS switches. However, the cross-hierarchical modelling of micro-electronic and micro-electromechanical constituents together in one multi-physical design process is still not as established as the design of integrated micro-electronic circuits, such as operational amplifiers. To close the gap between micro-electronics and micro-electromechanics, this paper presents an analytical approach towards the linear top-down design of MEMS resonators, based on their electrical specification, by the solution of the mechanical wave equation. In view of the central importance of thermal effects for the performance and stability of MEMS-based RF circuits, the temperature dependence was included in the model; the aim was to study the variations of the RF parameters of the resonators and to enable a temperature dependent MEMS oscillator simulation. The variations of the resonator parameters with respect to the ambient temperature were then verified by RF measurements in a vacuum chamber at temperatures between −35 ∘C and 85 ∘C. The systematic body of data revealed temperature coefficients of the resonant frequency between −26 ppm/K and −20 ppm/K, which are in good agreement with other data from the literature. Based on the MEMS resonator model derived, a MEMS oscillator was designed, simulated, and measured in a vacuum chamber yielding a measured temperature coefficient of the oscillation frequency of −26.3 ppm/K. The difference of the temperature coefficients of frequency of oscillator and resonator turned out to be mainly influenced by the limited Q-factor of the MEMS device. In both studies, the analytical model and the measurement showed very good agreement in terms of temperature dependence and the prediction of fabrication results of the resonators designed. This analytical modelling approach serves therefore as an important step towards the design and simulation of micro-electronics and micro-electromechanics in one uniform design process. Furthermore, temperature dependences of MEMS oscillators can now be studied by simulations instead of time-consuming and complex measurements.

Analysis and Optimization of Loss Functions for Multiclass, Top-k, and Multilabel Classification.

Top-k error is currently a popular performance measure on large scale image classification benchmarks such as ImageNet and Places. Despite its wide acceptance, our understanding of this metric is limited as most of the previous research is focused on its special case, the top-1 error. In this work, we explore two directions that shed more light on the top-k error. First, we provide an in-depth analysis of established and recently proposed single-label multiclass methods along with a detailed account of efficient optimization algorithms for them. Our results indicate that the softmax loss and the smooth multiclass SVM are surprisingly competitive in top-k error uniformly across all k, which can be explained by our analysis of multiclass top-k calibration. Further improvements for a specific k are possible with a number of proposed top-k loss functions. Second, we use the top-k methods to explore the transition from multiclass to multilabel learning. In particular, we find that it is possible to obtain effective multilabel classifiers on Pascal VOC using a single label per image for training, while the gap between multiclass and multilabel methods on MS COCO is more significant. Finally, our contribution of efficient algorithms for training with the considered top-k and multilabel loss functions is of independent interest.

MeDeCom: discovery and quantification of latent components of heterogeneous methylomes.

It is important for large-scale epigenomic studies to determine and explore the nature of hidden confounding variation, most importantly cell composition. We developed MeDeCom as a novel reference-free computational framework that allows the decomposition of complex DNA methylomes into latent methylation components and their proportions in each sample. MeDeCom is based on constrained non-negative matrix factorization with a new biologically motivated regularization function. It accurately recovers cell-type-specific latent methylation components and their proportions. MeDeCom is a new unsupervised tool for the exploratory study of the major sources of methylation variation, which should lead to a deeper understanding and better biological interpretation.

An Efficient Multilinear Optimization Framework for Hypergraph Matching.

Hypergraph matching has recently become a popular approach for solving correspondence problems in computer vision as it allows the use of higher-order geometric information. Hypergraph matching can be formulated as a third-order optimization problem subject to assignment constraints which turns out to be NP-hard. In recent work, we have proposed an algorithm for hypergraph matching which first lifts the third-order problem to a fourth-order problem and then solves the fourth-order problem via optimization of the corresponding multilinear form. This leads to a tensor block coordinate ascent scheme which has the guarantee of providing monotonic ascent in the original matching score function and leads to state-of-the-art performance both in terms of achieved matching score and accuracy. In this paper we show that the lifting step to a fourth-order problem can be avoided yielding a third-order scheme with the same guarantees and performance but being two times faster. Moreover, we introduce a homotopy type method which further improves the performance.

Learning using privileged information: SVM+ and weighted SVM.

Prior knowledge can be used to improve predictive performance of learning algorithms or reduce the amount of data required for training. The same goal is pursued within the learning using privileged information paradigm which was recently introduced by Vapnik et al. and is aimed at utilizing additional information available only at training time-a framework implemented by SVM+. We relate the privileged information to importance weighting and show that the prior knowledge expressible with privileged features can also be encoded by weights associated with every training example. We show that a weighted SVM can always replicate an SVM+ solution, while the converse is not true and we construct a counterexample highlighting the limitations of SVM+. Finally, we touch on the problem of choosing weights for weighted SVMs when privileged features are not available.

Isotope pattern deconvolution for peptide mass spectrometry by non-negative least squares/least absolute deviation template matching.

BACKGROUND: The robust identification of isotope patterns originating from peptides being analyzed through mass spectrometry (MS) is often significantly hampered by noise artifacts and the interference of overlapping patterns arising e.g. from post-translational modifications. As the classification of the recorded data points into either 'noise' or 'signal' lies at the very root of essentially every proteomic application, the quality of the automated processing of mass spectra can significantly influence the way the data might be interpreted within a given biological context. RESULTS: We propose non-negative least squares/non-negative least absolute deviation regression to fit a raw spectrum by templates imitating isotope patterns. In a carefully designed validation scheme, we show that the method exhibits excellent performance in pattern picking. It is demonstrated that the method is able to disentangle complicated overlaps of patterns. CONCLUSIONS: We find that regularization is not necessary to prevent overfitting and that thresholding is an effective and user-friendly way to perform feature selection. The proposed method avoids problems inherent in regularization-based approaches, comes with a set of well-interpretable parameters whose default configuration is shown to generalize well without the need for fine-tuning, and is applicable to spectra of different platforms. The R package IPPD implements the method and is available from the Bioconductor platform (http://bioconductor.fhcrc.org/help/bioc-views/devel/bioc/html/IPPD.html).

An integer linear programming approach for finding deregulated subgraphs in regulatory networks.

Deregulation of cell signaling pathways plays a crucial role in the development of tumors. The identification of such pathways requires effective analysis tools that facilitate the interpretation of expression differences. Here, we present a novel and highly efficient method for identifying deregulated subnetworks in a regulatory network. Given a score for each node that measures the degree of deregulation of the corresponding gene or protein, the algorithm computes the heaviest connected subnetwork of a specified size reachable from a designated root node. This root node can be interpreted as a molecular key player responsible for the observed deregulation. To demonstrate the potential of our approach, we analyzed three gene expression data sets. In one scenario, we compared expression profiles of non-malignant primary mammary epithelial cells derived from BRCA1 mutation carriers and of epithelial cells without BRCA1 mutation. Our results suggest that oxidative stress plays an important role in epithelial cells of BRCA1 mutation carriers and that the activation of stress proteins may result in avoidance of apoptosis leading to an increased overall survival of cells with genetic alterations. In summary, our approach opens new avenues for the elucidation of pathogenic mechanisms and for the detection of molecular key players.