The Data Science for Pavement Challenge (DSPC) seeks to accelerate the research and development of automated vision systems for pavement condition monitoring and evaluation by providing a platform with benchmarked datasets and codes for teams to innovate and develop machine learning algorithms that are practice-ready for use by industry. The first edition of the competition attracted 22 teams from 8 countries. Participants were required to automatically detect and classify different types of pavement distresses present in images captured from multiple sources, and under different conditions. The competition was data-centric: teams were tasked to increase the accuracy of a predefined model architecture by utilizing various data modification methods such as cleaning, labeling and augmentation. A real-time, online evaluation system was developed to rank teams based on the F1 score. Leaderboard results showed the promise and challenges of machine for advancing automation in pavement monitoring and evaluation. This paper summarizes the solutions from the top 5 teams. These teams proposed innovations in the areas of data cleaning, annotation, augmentation, and detection parameter tuning. The F1 score for the top-ranked team was approximately 0.9. The paper concludes with a review of different experiments that worked well for the current challenge and those that did not yield any significant improvement in model accuracy.
Federated learning provides a privacy guarantee for generating good deep learning models on distributed clients with different kinds of data. Nevertheless, dealing with non-IID data is one of the most challenging problems for federated learning. Researchers have proposed a variety of methods to eliminate the negative influence of non-IIDness. However, they only focus on the non-IID data provided that the universal class distribution is balanced. In many real-world applications, the universal class distribution is long-tailed, which causes the model seriously biased. Therefore, this paper studies the joint problem of non-IID and long-tailed data in federated learning and proposes a corresponding solution called Federated Ensemble Distillation with Imbalance Calibration (FEDIC). To deal with non-IID data, FEDIC uses model ensemble to take advantage of the diversity of models trained on non-IID data. Then, a new distillation method with logit adjustment and calibration gating network is proposed to solve the long-tail problem effectively. We evaluate FEDIC on CIFAR-10-LT, CIFAR-100-LT, and ImageNet-LT with a highly non-IID experimental setting, in comparison with the state-of-the-art methods of federated learning and long-tail learning. Our code is available at https://github.com/shangxinyi/FEDIC.
Federated learning (FL) provides a privacy-preserving solution for distributed machine learning tasks. One challenging problem that severely damages the performance of FL models is the co-occurrence of data heterogeneity and long-tail distribution, which frequently appears in real FL applications. In this paper, we reveal an intriguing fact that the biased classifier is the primary factor leading to the poor performance of the global model. Motivated by the above finding, we propose a novel and privacy-preserving FL method for heterogeneous and long-tailed data via Classifier Re-training with Federated Features (CReFF). The classifier re-trained on federated features can produce comparable performance as the one re-trained on real data in a privacy-preserving manner without information leakage of local data or class distribution. Experiments on several benchmark datasets show that the proposed CReFF is an effective solution to obtain a promising FL model under heterogeneous and long-tailed data. Comparative results with the state-of-the-art FL methods also validate the superiority of CReFF. Our code is available at https://github.com/shangxinyi/CReFF-FL.
Due to the sequential nature of the successive-cancellation (SC) algorithm, the decoding of polar codes suffers from significant decoding latencies. Fast SC decoding is able to speed up the SC decoding process, by implementing parallel decoders at the intermediate levels of the SC decoding tree for some special nodes with specific information and frozen bit patterns. To further improve the parallelism of SC decoding, this paper present a new class of special node composed of a sequence of rate one or single-parity-check (SR1/SPC) nodes, which envelops a wide variety of existing special node types. Then, we analyse the parity constraints caused by the frozen bits in each descendant node, such that the decoding performance of the SR1/SPC node can be preserved once the parity constraints are satisfied. Finally, a generalized fast SC decoding algorithm is proposed for the SR1/SPC node, where the corresponding parity constraints are taken into consideration. Simulation results show that compared with the state-of-the-art fast SC decoding algorithms, the proposed decoding algorithm achieves a higher degree of parallelism, especially for high-rate polar codes, without tangibly altering the error-correction performance.
Process mining is a relatively new subject which builds a bridge between traditional process modelling and data mining. Process discovery is one of the most critical parts of process mining which aims at discovering process models automatically from event logs. The performance of existing process discovery algorithms can be affected when there are missing activity labels in event logs. Several methods have been proposed to repair missing activity labels, but their accuracy can drop when a large number of activity labels are missing. In this paper, we propose a LSTM-based prediction model to predict the missing activity labels in event logs. The proposed model takes both the prefix and suffix sequences of the events with missing activity labels as input. Additional attributes of event logs are also utilised to improve the performance. Our evaluation on several publicly available datasets show that the proposed method performed consistently better than existing methods to repair missing activity labels in event logs.
With the increase of structure complexity, convolutional neural networks (CNNs) take a fair amount of computation cost. Meanwhile, existing research reveals the salient parameter redundancy in CNNs. The current pruning methods can compress CNNs with little performance drop, but when the pruning ratio increases, the accuracy loss is more serious. Moreover, some iterative pruning methods are difficult to accurately identify and delete unimportant parameters due to the accuracy drop during pruning. We propose a novel adversarial iterative pruning method (AIP) for CNNs based on knowledge transfer. The original network is regarded as the teacher while the compressed network is the student. We apply attention maps and output features to transfer information from the teacher to the student. Then, a shallow fully-connected network is designed as the discriminator to allow the output of two networks to play an adversarial game, thereby it can quickly recover the pruned accuracy among pruning intervals. Finally, an iterative pruning scheme based on the importance of channels is proposed. We conduct extensive experiments on the image classification tasks CIFAR-10, CIFAR-100, and ILSVRC-2012 to verify our pruning method can achieve efficient compression for CNNs even without accuracy loss. On the ILSVRC-2012, when removing 36.78% parameters and 45.55% floating-point operations (FLOPs) of ResNet-18, the Top-1 accuracy drop are only 0.66%. Our method is superior to some state-of-the-art pruning schemes in terms of compressing rate and accuracy. Moreover, we further demonstrate that AIP has good generalization on the object detection task PASCAL VOC.
Process mining is a relatively new subject which builds a bridge between process modelling and data mining. An exclusive choice in a process model usually splits the process into different branches. However, in some processes, it is possible to switch from one branch to another. The inductive miner guarantees to return sound process models, but fails to return a precise model when there are switch behaviours between different exclusive choice branches due to the limitation of process trees. In this paper, we present a novel extension to the process tree model to support switch behaviours between different branches of the exclusive choice operator and propose a novel extension to the inductive miner to discover sound process models with switch behaviours. The proposed discovery technique utilizes the theory of a previous study to detect possible switch behaviours. We apply both artificial and publicly-available datasets to evaluate our approach. Our results show that our approach can improve the precision of discovered models by 36% while maintaining high fitness values compared to the original inductive miner.
The insights revealed from process mining heavily rely on the quality of event logs. Activities extracted from different data sources or the free-text nature within the same system may lead to inconsistent labels. Such inconsistency would then lead to redundancy of activity labels, which refer to labels that have different syntax but share the same behaviours. The identifications of these labels from data-driven process discovery are difficult and would rely heavily on human intervention. In this paper, we propose an approach to detect redundant activity labels using control-flow relations and data values from event logs. We have evaluated our approach using two publicly available logs and also a case study using the MIMIC-III data set. The results demonstrate that our approach can detect redundant activity labels even with low occurrence frequencies. This approach can value-add to the preprocessing step to generate more representative event logs for process mining tasks.
Process mining acts as a valuable tool to analyse the behaviour of an organisation by offering techniques to discover, monitor and enhance real processes. The key to process mining is to discovery understandable process models. However, real-life logs can be complex with redundant activities, which share similar behaviour but have different syntax. We show that the existence of such redundant activities heavily affects the quality of discovered process models. Existing approaches filter activities by frequency, which cannot solve problems caused by redundant activities. In this paper, we propose first to discover redundant activities in the log level and, then, use the discovery results to simplify event logs. Two publicly available data sets are used to evaluate the usability of our approach in real-life processes. Our approach can be adopted as a preprocessing step before applying any discovery algorithms to produce simplify models.
Business processes are continuously evolving in order to adapt to changes due to various factors. One type of process changes are branching frequency changes, which are related to changes in frequencies between different options when there is an exclusive choice. Existing methods either cannot detect such changes or cannot provide accurate and comprehensive results. In this paper, we propose a method which takes both event logs and process models as input and generates a choice sequence for each exclusive choice in the process model. The method then identifies change points based on the choice sequences. We evaluate our method on a real-life event log. Results show that our method can identify branching frequency changes in process models and provide comprehensive results to users.