Research

2024

1. ICRA

   

   Safe Table Tennis Swing Stroke with Low-Cost Hardware

   Francesco Cursi, Marcus Kalander, Shuang Wu, Xidi Xue, Yu Tian, Guangjian Tian, Xingyue Quan, and Jianye Hao

   In 2024 IEEE International Conference on Robotics and Automation (ICRA), 2024

   Abs HTML

   Playing table tennis with a human player is a challenging robotic task due to its dynamic nature. Despite a number of researches being devoted to developing robotic table tennis systems, most of the works have demanding hardware requirements and ignore safety measures when generating the swing stoke. To address these issues, we propose a safe motion planning framework that fully pushes the robotic hardware performance limits to play table tennis. In particular, we propose a pipeline to generate manipulator joint trajectories with environmental safety constraints and scale the trajectories to satisfy joint movement limitations. We use three different agents to validate the planning algorithm with our handmade robot platform in both simulation and real-world environments.

2. Machine Learning

   

   Exploiting Counter-Examples for Active Learning with Partial labels

   Fei Zhang, Yunjie Ye, Lei Feng, Zhongwen Rao , Jieming Zhu, Marcus Kalander, Chen Gong, Jianye Hao, and Bo Han

   Machine Learning, 2024

   Abs arXiv Code

   This paper studies a new problem, active learning with partial labels (ALPL). In this setting, an oracle annotates the query samples with partial labels, relaxing the oracle from the demanding accurate labeling process. To address ALPL, we first build an intuitive baseline that can be seamlessly incorporated into existing AL frameworks. Though effective, this baseline is still susceptible to the overfitting, and falls short of the representative partial-label-based samples during the query process. Drawing inspiration from human inference in cognitive science, where accurate inferences can be explicitly derived from counter-examples (CEs), our objective is to leverage this human-like learning pattern to tackle the overfitting while enhancing the process of selecting representative samples in ALPL. Specifically, we construct CEs by reversing the partial labels for each instance, and then we propose a simple but effective WorseNet to directly learn from this complementary pattern. By leveraging the distribution gap between WorseNet and the predictor, this adversarial evaluation manner could enhance both the performance of the predictor itself and the sample selection process, allowing the predictor to capture more accurate patterns in the data. Experimental results on five real-world datasets and four benchmark datasets show that our proposed method achieves comprehensive improvements over ten representative AL frameworks, highlighting the superiority of WorseNet.

2023

1. ICLR

   

   Out-of-distribution Detection with Implicit Outlier Transformation

   Qizhou Wang, Junjie Ye , Feng Liu, Quanyu Dai, Marcus Kalander, Tongliang Liu, Jianye Hao, and Bo Han

   The Eleventh International Conference on Learning Representations (ICLR), 2023

   Abs arXiv Code

   Outlier exposure (OE) is powerful in out-of-distribution (OOD) detection, enhancing detection capability via model fine-tuning with surrogate OOD data. However, surrogate data typically deviate from test OOD data. Thus, the performance of OE, when facing unseen OOD data, can be weakened. To address this issue, we propose a novel OE-based approach that makes the model perform well for unseen OOD situations, even for unseen OOD cases. It leads to a min-max learning scheme – searching to synthesize OOD data that leads to worst judgments and learning from such OOD data for uniform performance in OOD detection. In our realization, these worst OOD data are synthesized by transforming original surrogate ones. Specifically, the associated transform functions are learned implicitly based on our novel insight that model perturbation leads to data transformation. Our methodology offers an efficient way of synthesizing OOD data, which can further benefit the detection model, besides the surrogate OOD data. We conduct extensive experiments under various OOD detection setups, demonstrating the effectiveness of our method against its advanced counterparts.

2. arXiv

   

   Exploit CAM by itself: Complementary Learning System for Weakly Supervised Semantic Segmentation

   Jiren Mai, Fei Zhang, Junjie Ye, Marcus Kalander, Xian Zhang , WanKou Yang, Tongliang Liu, and Bo Han

   arXiv preprint arXiv:2303.02449, 2023

   Abs arXiv

   Weakly Supervised Semantic Segmentation (WSSS) with image-level labels has long been suffering from fragmentary object regions led by Class Activation Map (CAM), which is incapable of generating fine-grained masks for semantic segmentation. To guide CAM to find more non-discriminating object patterns, this paper turns to an interesting working mechanism in agent learning named Complementary Learning System (CLS). CLS holds that the neocortex builds a sensation of general knowledge, while the hippocampus specially learns specific details, completing the learned patterns. Motivated by this simple but effective learning pattern, we propose a General-Specific Learning Mechanism (GSLM) to explicitly drive a coarse-grained CAM to a fine-grained pseudo mask. Specifically, GSLM develops a General Learning Module (GLM) and a Specific Learning Module (SLM). The GLM is trained with image-level supervision to extract coarse and general localization representations from CAM. Based on the general knowledge in the GLM, the SLM progressively exploits the specific spatial knowledge from the localization representations, expanding the CAM in an explicit way. To this end, we propose the Seed Reactivation to help SLM reactivate non-discriminating regions by setting a boundary for activation values, which successively identifies more regions of CAM. Without extra refinement processes, our method is able to achieve breakthrough improvements for CAM of over 20.0% mIoU on PASCAL VOC 2012 and 10.0% mIoU on MS COCO 2014 datasets, representing a new state-of-the-art among existing WSSS methods.

2022

1. arXiv

   

   RiskLoc: Localization of Multi-dimensional Root Causes by Weighted Risk

   Marcus Kalander

   arXiv preprint arXiv:2205.10004, 2022

   Abs arXiv Code

   Failures and anomalies in large-scale software systems are unavoidable incidents. When an issue is detected, operators need to quickly and correctly identify its location to facilitate a swift repair. In this work, we consider the problem of identifying the root cause set that best explains an anomaly in multi-dimensional time series with categorical attributes. The huge search space is the main challenge, even for a small number of attributes and small value sets, the number of theoretical combinations is too large to brute force. Previous approaches have thus focused on reducing the search space, but they all suffer from various issues, requiring extensive manual parameter tuning, being too slow and thus impractical, or being incapable of finding more complex root causes. We propose RiskLoc to solve the problem of multidimensional root cause localization. RiskLoc applies a 2-way partitioning scheme and assigns element weights that linearly increase with the distance from the partitioning point. A risk score is assigned to each element that integrates two factors, 1) its weighted proportion within the abnormal partition, and 2) the relative change in the deviation score adjusted for the ripple effect property. Extensive experiments on multiple datasets verify the effectiveness and efficiency of RiskLoc, and for a comprehensive evaluation, we introduce three synthetically generated datasets that complement existing datasets. We demonstrate that RiskLoc consistently outperforms state-of-the-art baselines, especially in more challenging root cause scenarios, with gains in F1-score up to 57% over the second-best approach with comparable running times.

2. KDD Cup

   

   Wind Power Forecasting with Deep Learning: Team didadida_hualahuala

   Marcus Kalander, Zhongwen Rao, and Chengzhi Zhang

   KDD Cup 2022, 2022

   Abs HTML Video Code

   Wind energy is an effective supplement to traditional energy sources. However, wind power generation is strongly weather-dependent and thus not only unsustainable in terms of production, but also highly volatile. This complex variability poses a huge challenge to the smooth operation of the grid system. In this scenario, a unique Spatial Dynamic Wind Power Forecasting dataset from Longyuan Power Group Corp. Ltd (SDWPF) is used for modern wind turbine power forecasting. In this work, we propose a framework for accurate wind power generation forecasting (WPF) based on deep learning. To obtain a more generalizable learner, we use ensemble learning to build our framework, which consists of the following two parts: a) a modified version of the DLinear model, which uses time series decomposition and linear layers for information aggregation, and b) an extreme time-gated network that adaptively captures finegrained information and re-aggregates information by spatial location in the inference stage to obtain higher forecasting accuracy. The results indicate that our proposed combined model framework can capture long-term time series information well. Our code is available at https://github. com/shaido987/KDD_wind_power_forecast.

3. DASC

   

   Contrastive Representation based Active Learning for Time Series

   Lujia Pan, Marcus Kalander, Yuchao Zhang , and Pinghui Wang

   In 2022 IEEE Intl Conf on Dependable, Autonomic and Secure Computing (DASC), 2022

   Abs HTML

   Active Learning designs query strategies to select the most representative samples to be labeled by an oracle in an attempt to maximize the model’s performance while minimizing the labeling workload. We propose REAL, a new pooling-based active learning algorithm for time series data that learns the query strategy and optimizes the representation model in a contrastive manner. To initialize the process, a cluster module is employed to select the first sample set for labeling. Subsequent samples are selected through a contrastive loss function from three complementary perspectives, self-consistency, attraction to similar samples, and repulsion of disparate samples. Concurrently, the contrastive loss is also used to update the representation model. We evaluate our method on various time series classification tasks against state-of-the-art algorithms and demonstrate gains or comparable performance for an equal number of labeled samples.

4. JIFS

   

   LDAAD: An effective label de-noising approach for anomaly detection

   Lujia Pan, Marcus Kalander , and Pinghui Wang

   Journal of Intelligent & Fuzzy Systems, 2022

   Abs PDF

   Classification algorithms are widely applied to predict failures and detect anomalies in various application areas. It is common to assume that the data and labels are correct when training, but this is challenging to guarantee in the real world. If there are erroneous labels in the training data, a model can easily overfit to these, resulting in poor performance. How to handle label noise has been previously researched, however, few works focus on label noise in anomaly detection. In this work, we propose LDAAD, a novel algorithm framework for label de-noising for anomaly detection that combines unsupervised learning and semi-supervised learning methods. Specifically, we apply anomaly detection to partition the training data into low-risk and high-risk sets. We subsequently build upon ideas from cross-validation and train multiple classification models on segments of the low-risk data. The models are used both to relabel the samples in the high-risk set and to filter the low-risk samples. Finally, we merge the two sets to obtain a final sample set with more confident labels. We evaluate LDAAD on multiple real-world datasets and show that LDAAD achieves robust results that outperform the benchmark methods. Specifically, LDAAD achieves a 5% accuracy improvement over the second-best method for symmetric noise while having a minimal detrimental impact when no label noise is present.

2021

1. KDD

   

   Discrete-time Temporal Network Embedding via Implicit Hierarchical Learning in Hyperbolic Space

   Menglin Yang, Min Zhou, Marcus Kalander, Zengfeng Huang, and Irwin King

   In Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining, 2021

   Abs arXiv Code

   Representation learning over temporal networks has drawn considerable attention in recent years. Efforts are mainly focused on modeling structural dependencies and temporal evolving regularities in Euclidean space which, however, underestimates the inherent complex and hierarchical properties in many real-world temporal networks, leading to sub-optimal embeddings. To explore these properties of a complex temporal network, we propose a hyperbolic temporal graph network (HTGN) that fully takes advantage of the exponential capacity and hierarchical awareness of hyperbolic geometry. More specially, HTGN maps the temporal graph into hyperbolic space, and incorporates hyperbolic graph neural network and hyperbolic gated recurrent neural network, to capture the evolving behaviors and implicitly preserve hierarchical information simultaneously. Furthermore, in the hyperbolic space, we propose two important modules that enable HTGN to successfully model temporal networks: (1) hyperbolic temporal contextual self-attention (HTA) module to attend to historical states and (2) hyperbolic temporal consistency (HTC) module to ensure stability and generalization. Experimental results on multiple real-world datasets demonstrate the superiority of HTGN for temporal graph embedding, as it consistently outperforms competing methods by significant margins in various temporal link prediction tasks. Specifically, HTGN achieves AUC improvement up to 9.98% for link prediction and 11.4% for new link prediction. Moreover, the ablation study further validates the representational ability of hyperbolic geometry and the effectiveness of the proposed HTA and HTC modules. Code is publicly available at https://github.com/marlin-codes/HTGN.

2. arXiv

   

   gCastle: A Python Toolbox for Causal Discovery

   Keli Zhang, Shengyu Zhu, Marcus Kalander, Ignavier Ng, Junjian Ye, Zhitang Chen, and Lujia Pan

   arXiv preprint arXiv:2111.15155, 2021

   Abs arXiv Code

   gCastle is an end-to-end Python toolbox for causal structure learning. It provides functionalities of generating data from either simulator or real-world dataset, learning causal structure from the data, and evaluating the learned graph, together with useful practices such as prior knowledge insertion, preliminary neighborhood selection, and post-processing to remove false discoveries. Compared with related packages, gCastle includes many recently developed gradient-based causal discovery methods with optional GPU acceleration. gCastle brings convenience to researchers who may directly experiment with the code as well as practitioners with graphical user interference. Three real-world datasets in telecommunications are also provided in the current version. gCastle is available under Apache License 2.0 at \urlhttps://github.com/huawei-noah/trustworthyAI/tree/master/gcastle.

3. IJCAI WS

   

   An Ensemble Noise-Robust K-fold Cross-Validation Selection Method for Noisy Labels

   Yong Wen^*, Marcus Kalander^*, Chanfei Su, and Lujia Pan

   Weakly Supervised Representation Learning Workshop @ IJCAI, 2021

   Abs arXiv

   We consider the problem of training robust and accurate deep neural networks (DNNs) when subject to various proportions of noisy labels. Large-scale datasets tend to contain mislabeled samples that can be memorized by DNNs, impeding the performance. With appropriate handling, this degradation can be alleviated. There are two problems to consider: how to distinguish clean samples and how to deal with noisy samples. In this paper, we present Ensemble Noise-robust K-fold Cross-Validation Selection (E-NKCVS) to effectively select clean samples from noisy data, solving the first problem. For the second problem, we create a new pseudo label for any sample determined to have an uncertain or likely corrupt label. E-NKCVS obtains multiple predicted labels for each sample and the entropy of these labels is used to tune the weight given to the pseudo label and the given label. Theoretical analysis and extensive verification of the algorithms in the noisy label setting are provided. We evaluate our approach on various image and text classification tasks where the labels have been manually corrupted with different noise ratios. Additionally, two large real-world noisy datasets are also used, Clothing-1M and WebVision. E-NKCVS is empirically shown to be highly tolerant to considerable proportions of label noise and has a consistent improvement over state-of-the-art methods. Especially on more difficult datasets with higher noise ratios, we can achieve a significant improvement over the second-best model. Moreover, our proposed approach can easily be integrated into existing DNN methods to improve their robustness against label noise.

2020

1. ICSOC

   

   An Influence-based Approach for Root Cause Alarm Discovery in Telecom Networks

   Keli Zhang^*, Marcus Kalander^*, Min Zhou, Xi Zhang, and Junjian Ye

   In International Conference on Service-Oriented Computing, 2020

   Abs arXiv Video Code

   Alarm root cause analysis is a significant component in the day-to-day telecommunication network maintenance, and it is critical for efficient and accurate fault localization and failure recovery. In practice, accurate and self-adjustable alarm root cause analysis is a great challenge due to network complexity and vast amounts of alarms. A popular approach for failure root cause identification is to construct a graph with approximate edges, commonly based on either event co-occurrences or conditional independence tests. However, considerable expert knowledge is typically required for edge pruning. We propose a novel data-driven framework for root cause alarm localization, combining both causal inference and network embedding techniques. In this framework, we design a hybrid causal graph learning method (HPCI), which combines Hawkes Process with Conditional Independence tests, as well as propose a novel Causal Propagation-Based Embedding algorithm (CPBE) to infer edge weights. We subsequently discover root cause alarms in a real-time data stream by applying an influence maximization algorithm on the weighted graph. We evaluate our method on artificial data and real-world telecom data, showing a significant improvement over the best baselines.

2. arXiv

   

   Spatio-Temporal Hybrid Graph Convolutional Network for Traffic Forecasting in Telecommunication Networks

   Marcus Kalander, Min Zhou, Chengzhi Zhang, Hanling Yi, and Lujia Pan

   arXiv preprint arXiv:2009.09849, 2020

   Abs arXiv

   Telecommunication networks play a critical role in modern society. With the arrival of 5G networks, these systems are becoming even more diversified, integrated, and intelligent. Traffic forecasting is one of the key components in such a system, however, it is particularly challenging due to the complex spatial-temporal dependency. In this work, we consider this problem from the aspect of a cellular network and the interactions among its base stations. We thoroughly investigate the characteristics of cellular network traffic and shed light on the dependency complexities based on data collected from a densely populated metropolis area. Specifically, we observe that the traffic shows both dynamic and static spatial dependencies as well as diverse cyclic temporal patterns. To address these complexities, we propose an effective deep-learning-based approach, namely, Spatio-Temporal Hybrid Graph Convolutional Network (STHGCN). It employs GRUs to model the temporal dependency, while capturing the complex spatial dependency through a hybrid-GCN from three perspectives: spatial proximity, functional similarity, and recent trend similarity. We conduct extensive experiments on real-world traffic datasets collected from telecommunication networks. Our experimental results demonstrate the superiority of the proposed model in that it consistently outperforms both classical methods and state-of-the-art deep learning models, while being more robust and stable.

3. Computer Networks

   

   Proactive Microwave Link Anomaly Detection in Cellular Data Networks

   Lujia Pan, Jianfeng Zhang, Patrick PC Lee, Marcus Kalander, Junjian Ye , and Pinghui Wang

   Computer Networks, 2020

   Abs HTML

   Microwave links are widely used in cellular networks for large-scale data transmission. From the network operators’ perspective, it is critical to quickly and accurately detect microwave link failures before they actually happen, thereby maintaining the robustness of the data transmissions. We present PMADS, a machine-learning-based proactive microwave link anomaly detection system that exploits both performance data and network topological information to detect microwave link anomalies that may eventually lead to actual failures. Our key observation is that anomalous links often share similar network topological properties, thereby enabling us to further improve the detection accuracy. To this end, PMADS adopts a network-embedding-based approach to encode topological information into features. It further adopts a novel active learning algorithm, ADAL, to continuously update the detection model at low cost by first applying unsupervised learning to separate anomalies as outliers from the training set. We evaluate PMADS on a real-world dataset collected from 2142 microwave links in a production LTE network during a one-month period, and show that PMADS achieves a precision of 94.4% and a recall of 87.1%. Furthermore, using the active learning feedback loop, only 7% of the training data is required to achieve comparable results. PMADS is currently deployed in a metropolitan LTE network that serves around four million subscribers in a Middle Eastern country.

2016

1. M.Sc.

   

   A natural language processing approach for identifying driving styles in curves

   Eric McNabb^*, and Marcus Kalander^*

   2016

   Abs HTML

   A machine able to autonomously recognise driving styles has numerous applications, of which the most straightforward is to recognise risky behaviour. Such knowledge can be used to teach new drivers with the goal of reducing accidents in the future and increasing traffic safety for all road users. Furthermore, insurance companies can incentivise safe driving with lower premiums, which in turn can motivate a more careful driving style. Another application is within the field of autonomous vehicles where learning about driving styles is imperative for autonomous vehicles to be able to interact with other drivers in traffic. The first step towards identifying different driving styles is being able to recognise and distinguish between them. The aim of this thesis is to identify the indicators of aggressive driving in curves from a large amount of naturalistic driving data. The first step was finding curve sections to analyse within trips and the second step was reducing the data to become more manageable. Symbolic representations were used for the second preprocessing step, which in turn allowed the use of Natural Language Processing techniques for the analysis. We categorise drivers into different groups depending on their perceived tendency towards aggressive driving styles. This categorisation is used to compare the drivers and their driving style with each other. The tendencies used were Speeding, Braking, Jerky curve handling and Rough curve handling. Some general trends among the analysed drivers are also identified. It is possible to reuse the categorisation to include more drivers in the future or to use what we have learned about the features and drivers for further research.

2014

1. B.Sc.

   

   Chalmers oanvända datorkraft-Distribuering av arbete och energihantering med HTCondor

   Daniel Bergqvist, Marcus Kalander, Oliver Andersson, Pontus Johansson Berg, and Rurik Högfeldt

   2014

   Abs HTML

   Chalmers University of Technology today have numerous computers which are never powered down, at the same time there is a need for more computation power for researchers. Hence, we have investigated the possibility to use the computers for computing power and secondly, if there is no work to be done, to put them into power saving mode. We have made a thorough study where we compared different distributing systems and in the end HTCondor was chosen as the best to implement. HTCondor is an excellent system for opportunistic use of computing power, i.e. make use of computers that no one else is currently using. The system is used at several universities around the world with good results and would be excellent at Chalmers, where there is a need for such a system. Our implementation shows that HTCondor is well capable of handling unused computing power. HTCondor can handle most file types that may be run on the system. The system can also put computers in sleep mode when they are not in use and turn them on when needed. To simplify for the users of HTCondor we have created a user interface that have all necessary functions required to make use of the system as no good such interface was available. There are doubts from Chalmers regarding the idea of shutting down computers, they argue that the computers always must be available and that problems may arise when turning them on, for example programs that do not start. Our investigation shows that there are no bigger problems with turning on and off computers, which is shown in a energy study in which sleep mode is determined to be the best energy saving option for Chalmers.