OMOle J. Mengshoel

Ole J. Mengshoel

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Ole J. Mengshoel has a total of 45 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    SemNet: Learning semantic attributes for human activity recognition with deep belief networks

    by Shanmuga Venkatachalam, Harideep Nair, Ming Zeng, Cathy Shunwen Tan, Ole J. Mengshoel, John Paul Shen

    Human Activity Recognition (HAR) is a prominent application in mobile computing and Internet of Things (IoT) that aims to detect human activities based on multimodal sensor signals generated as a result of diverse body movements. Human physical activities are typically composed of simple actions (such as “arm up”, “arm down”, “arm curl”, etc.), referred to assemanticfeatures. Such abstract semantic features, in contrast to high-level activities (“walking”, “sitting”, etc.) and low-level signals (raw sensor readings), can be developed manually to assist activity recognition. Although effective, this manual approach relies heavily on human domain expertise and is not scalable. In this paper, we address this limitation by proposing a machine learning method, SemNet, based on deep belief networks. SemNet automatically constructs semantic features representative of the axial bodily movements. Experimental results show that SemNet outperforms baseline approaches and is capable of learning features that highly correlate with manually defined semantic attributes. Furthermore, our experiments using a different model, namely deep convolutional LSTM, on household activities illustrate the broader applicability of semantic attribute interpretation to diverse deep neural network approaches. These empirical results not only demonstrate that such a deep learning technique is semantically meaningful and superior to its handcrafted counterpart, but also provides a better understanding of the deep learning methods that are used for Human Activity Recognition.


  • Open Access

    Customized Graph Embedding: Tailoring Embedding Vectors to different Applications

    by Bitan Hou, Yujing Wang, Ming Zeng, Shan Jiang, Ole J. Mengshoel, Yunhai Tong, Jing Bai

    Graph is a natural representation of data for a variety of real-word applications, such as knowledge graph mining, social network analysis and biological network comparison. For these applications, graph embedding is crucial as it provides vector representations of the graph. One limitation of existing graph embedding methods is that their embedding optimization procedures are disconnected from the target application. In this paper, we propose a novel approach, namely Customized Graph Embedding (CGE) to tackle this problem. The CGE algorithm learns customized vector representations of graph nodes by differentiating the importance of distinct graph paths automatically for a specific application. Extensive experiments were carried out on a diverse set of node classification datasets, which demonstrate strong performances of CGE and provide deep insights into the model.

  • Renewable Energy Integration: Bayesian Networks for Probabilistic State Estimation

    by Ole J. Mengshoel, Priya K. Sundararajan, Erik Reed, Dongzhen Piao, Briana Johnson


  • Open Access

    SpectralLeader: Online Spectral Learning for Single Topic Models

    by Tong Yu, Branislav Kveton, Zheng Wen, Hung Bui, Ole J. Mengshoel

    We study the problem of learning a latent variable model from a stream of data. Latent variable models are popular in practice because they can explain observed data in terms of unobserved concepts. These models have been traditionally studied in the offline setting. In the online setting, on the other hand, the online EM is arguably the most popular algorithm for learning latent variable models. Although the online EM is computationally efficient, it typically converges to a local optimum. In this work, we develop a new online learning algorithm for latent variable models, which we call SpectralLeader. SpectralLeader always converges to the global optimum, and we derive a sublinear upper bound on its $n$-step regret in the bag-of-words model. In both synthetic and real-world experiments, we show that SpectralLeader performs similarly to or better than the online EM with tuned hyper-parameters.

  • Open Access

    CADDeLaG: Framework for distributed anomaly detection in large dense graph sequences

    by Aniruddha Basak, Kamalika Das, Ole J. Mengshoel

    Random walk based distance measures for graphs such as commute-time distance are useful in a variety of graph algorithms, such as clustering, anomaly detection, and creating low dimensional embeddings. Since such measures hinge on the spectral decomposition of the graph, the computation becomes a bottleneck for large graphs and do not scale easily to graphs that cannot be loaded in memory. Most existing graph mining libraries for large graphs either resort to sampling or exploit the sparsity structure of such graphs for spectral analysis. However, such methods do not work for dense graphs constructed for studying pairwise relationships among entities in a data set. Examples of such studies include analyzing pairwise locations in gridded climate data for discovering long distance climate phenomena. These graphs representations are fully connected by construction and cannot be sparsified without loss of meaningful information. In this paper we describe CADDeLaG, a framework for scalable computation of commute-time distance based anomaly detection in large dense graphs without the need to load the entire graph in memory. The framework relies on Apache Spark's memory-centric cluster-computing infrastructure and consists of two building blocks: a decomposable algorithm for commute time distance computation and a distributed linear system solver. We illustrate the scalability of CADDeLaG and its dependency on various factors using both synthetic and real world data sets. We demonstrate the usefulness of CADDeLaG in identifying anomalies in a climate graph sequence, that have been historically missed due to ad hoc graph sparsification and on an election donation data set.

  • Open Access

    Semi-Supervised Convolutional Neural Networks for Human Activity Recognition

    by Ming Zeng, Tong Yu, Xiao Wang, Le T. Nguyen, Ole J. Mengshoel, Ian Lane

    Labeled data used for training activity recognition classifiers are usually limited in terms of size and diversity. Thus, the learned model may not generalize well when used in real-world use cases. Semi-supervised learning augments labeled examples with unlabeled examples, often resulting in improved performance. However, the semi-supervised methods studied in the activity recognition literatures assume that feature engineering is already done. In this paper, we lift this assumption and present two semi-supervised methods based on convolutional neural networks (CNNs) to learn discriminative hidden features. Our semi-supervised CNNs learn from both labeled and unlabeled data while also performing feature learning on raw sensor data. In experiments on three real world datasets, we show that our CNNs outperform supervised methods and traditional semi-supervised learning methods by up to 18% in mean F1-score (Fm).

  • Open Access

    Customized Nonlinear Bandits for Online Response Selection in Neural Conversation Models

    by Bing Liu, Tong Yu, Ian Lane, Ole J. Mengshoel

    Dialog response selection is an important step towards natural response generation in conversational agents. Existing work on neural conversational models mainly focuses on offline supervised learning using a large set of context-response pairs. In this paper, we focus on online learning of response selection in retrieval-based dialog systems. We propose a contextual multi-armed bandit model with a nonlinear reward function that uses distributed representation of text for online response selection. A bidirectional LSTM is used to produce the distributed representations of dialog context and responses, which serve as the input to a contextual bandit. In learning the bandit, we propose a customized Thompson sampling method that is applied to a polynomial feature space in approximating the reward. Experimental results on the Ubuntu Dialogue Corpus demonstrate significant performance gains of the proposed method over conventional linear contextual bandits. Moreover, we report encouraging response selection performance of the proposed neural bandit model using the Recall@k metric for a small set of online training samples.

  • Open Access

    Self-Awareness of Cloud Applications

    by Alexandru Iosup, Xiaoyun Zhu, Arif Merchant, Eva Kalyvianaki, Martina Maggio, Simon Spinner, Tarek Abdelzaher, Ole Mengshoel, Sara Bouchenak

    Cloud applications today deliver an increasingly larger portion of the Information and Communication Technology (ICT) services. To address the scale, growth, and reliability of cloud applications, self-aware management and scheduling are becoming commonplace. How are they used in practice? In this chapter, we propose a conceptual framework for analyzing state-of-the-art self-awareness approaches used in the context of cloud applications. We map important applications corresponding to popular and emerging application domains to this conceptual framework, and compare the practical characteristics, benefits, and drawbacks of self-awareness approaches. Last, we propose a roadmap for addressing open challenges in self-aware cloud and datacenter applications.

  • Open Access

    Towards Real-time, On-board, Hardware-supported Sensor and Software Health Management for Unmanned Aerial Systems

    by Johann Schumann, Kristin Y. Rozier, Thomas Reinbacher, Ole J. Mengshoel, Timmy Mbaya, Corey Ippolito

    Unmanned aerial systems (UASs) can only be deployed if they can effectively complete their missions and respond to failures and uncertain environmental conditions while maintaining safety with respect to other aircraft as well as humen and property on the ground. In this paper, we design a real-time, on-board system health management (SHM) capability to continuously monitor sensors, software, and hardware components for detection and diagnosis of failures and violations of safety or performance rules during the flight of a UAS. Our approach to SHM is three-pronged, providing: (1) real-time monitoring of sensor and/or software signals; (2) signal analysis, preprocessing, and advanced on- the-fly temporal and Bayesian probabilistic fault diagnosis; (3) an unobtrusive, lightweight, read-only, low-power realization using Field Programmable Gate Arrays (FPGAs) that avoids overburdening limited computing resources or costly re-certification of flight software due to instrumentation.Our implementation provides a novel approach of combin- ing modular building blocks, integrating responsive runtime monitoring of temporal logic system safety requirements with model-based diagnosis and Bayesian network-based probabilistic analysis. We demonstrate this approach using actual data from the NASA Swift UAS, an experimental all-electric aircraft.


  • Open Access

    Integrating Probabilistic Reasoning and Statistical Quality Control Techniques for Fault Diagnosis in Hybrid Domains

    by Brian Ricks, Craig Harrison, Ole J. Mengshoel

    Bayesian networks, which may be compiled to arithmetic circuits in the interest of speed and predictability, provide a probabilistic method for system fault diagnosis. Currently, there is a limitation in arithmetic circuits in that they can only represent discrete random variables, while important fault types such as drift and offset faults are continuous and induce continuous sensor data. In this paper, we investigate how to handle continuous behavior while using discrete random variables with a small number of states. Central in our approach is the novel integration of a method from statistical quality control, known as cumulative sum (CUSUM), with probabilistic reasoning using static arithmetic circuits compiled from static Bayesian networks. Experimentally, an arithmetic circuit model of the ADAPT Electrical Power System (EPS), a real- world EPS located at the NASA Ames Research Center, is considered. We report on the validation of this approach using PRODIAGNOSE, which had the best performance in three of four industrial track competitions at the International Workshop on Principles of Diagnosis in 2009 and 2010 (DXC-09 and DXC-10). We demonstrate that PRODIAGNOSE, augmented with the CUSUM technique, is successful in diagnosing faults that are small in magnitude (offset faults) or drift linearly from a nominal state (drift faults). In one of these experiments, detection accuracy dramatically improved when CUSUM was used, jumping from 46.15% (CUSUM disabled) to 92.31% (CUSUM enabled).