scholarly journals Query-Driven Multi-Instance Learning

2020 ◽  
Vol 34 (04) ◽  
pp. 4158-4165
Author(s):  
Yen-Chi Hsu ◽  
Cheng-Yao Hong ◽  
Ming-Sui Lee ◽  
Tyng-Luh Liu
Keyword(s):  
Experimental Results ◽  
Weighted Sum ◽  
Class Label ◽  
Action Classification ◽  
Video Clips ◽  
A New Technique ◽  
Class Labels ◽  
The Given ◽  
Network Component ◽  
Generalized Compatibility

We introduce a query-driven approach (qMIL) to multi-instance learning where the queries aim to uncover the class labels embodied in a given bag of instances. Specifically, it solves a multi-instance multi-label learning (MIML) problem with a more challenging setting than the conventional one. Each MIML bag in our formulation is annotated only with a binary label indicating whether the bag contains the instance of a certain class and the query is specified by the word2vec of a class label/name. To learn a deep-net model for qMIL, we construct a network component that achieves a generalized compatibility measure for query-visual co-embedding and yields proper instance attentions to the given query. The bag representation is then formed as the attention-weighted sum of the instances' weights, and passed to the classification layer at the end of the network. In addition, the qMIL formulation is flexible for extending the network to classify unseen class labels, leading to a new technique to solve the zero-shot MIML task through an iterative querying process. Experimental results on action classification over video clips and three MIML datasets from MNIST, CIFAR10 and Scene are provided to demonstrate the effectiveness of our method.

scholarly journals Hierarchical Classification Based on Label Distribution Learning

2019 ◽  
Vol 33 ◽  
pp. 5533-5540 ◽  
Author(s):  
Changdong Xu ◽  
Xin Geng
Keyword(s):  
Hierarchical Classification ◽  
Experimental Results ◽  
Challenging Problem ◽  
Training Set ◽  
True Label ◽  
Major Bottleneck ◽  
Novel Method ◽  
Class Labels ◽  
Label Distribution ◽  
The Given

Hierarchical classification is a challenging problem where the class labels are organized in a predefined hierarchy. One primary challenge in hierarchical classification is the small training set issue of the local module. The local classifiers in the previous hierarchical classification approaches are prone to over-fitting, which becomes a major bottleneck of hierarchical classification. Fortunately, the labels in the local module are correlated, and the siblings of the true label can provide additional supervision information for the instance. This paper proposes a novel method to deal with the small training set issue. The key idea of the method is to represent the correlation among the labels by the label distribution. It generates a label distribution that contains the supervision information of each label for the given instance, and then learns a mapping from the instance to the label distribution. Experimental results on several hierarchical classification datasets show that our method significantly outperforms other state-of-theart hierarchical classification approaches.

An approach for outlier and novelty detection for text data based on classifier confidence

2020 ◽  
pp. 1-15
Author(s):  
Nikola Pižurica ◽  
Savo Tomović
Keyword(s):  
Anomaly Detection ◽  
Novelty Detection ◽  
Experimental Results ◽  
Classification Model ◽  
Training Dataset ◽  
Training Phase ◽  
High Confidence ◽  
Text Data ◽  
Class Labels ◽  
The Given

In this paper we present an approach for novelty detection in text data. The approach can also be considered as semi-supervised anomaly detection because it operates with the training dataset containing labelled instances for the known classes only. During the training phase the classification model is learned. It is assumed that at least two known classes exist in the available training dataset. In the testing phase instances are classified as normal or anomalous based on the classifier confidence. In other words, if the classifier cannot assign any of the known class labels to the given instance with sufficiently high confidence (probability), the instance will be declared as novelty (anomaly). We propose two procedures to objectively measure the classifier confidence. Experimental results show that the proposed approach is comparable to methods known in the literature.

BiLabel-Specific Features for Multi-Label Classification

2021 ◽  
Vol 16 (1) ◽  
pp. 1-23
Author(s):  
Min-Ling Zhang ◽  
Jun-Peng Fang ◽  
Yi-Bo Wang
Keyword(s):  
Predictive Models ◽  
Comparative Studies ◽  
State Of The Art ◽  
Classification Model ◽  
Generation Process ◽  
Prototype Selection ◽  
Class Label ◽  
Benchmark Datasets ◽  
Label Correlations ◽  
Class Labels

In multi-label classification, the task is to induce predictive models which can assign a set of relevant labels for the unseen instance. The strategy of label-specific features has been widely employed in learning from multi-label examples, where the classification model for predicting the relevancy of each class label is induced based on its tailored features rather than the original features. Existing approaches work by generating a group of tailored features for each class label independently, where label correlations are not fully considered in the label-specific features generation process. In this article, we extend existing strategy by proposing a simple yet effective approach based on BiLabel-specific features. Specifically, a group of tailored features is generated for a pair of class labels with heuristic prototype selection and embedding. Thereafter, predictions of classifiers induced by BiLabel-specific features are ensembled to determine the relevancy of each class label for unseen instance. To thoroughly evaluate the BiLabel-specific features strategy, extensive experiments are conducted over a total of 35 benchmark datasets. Comparative studies against state-of-the-art label-specific features techniques clearly validate the superiority of utilizing BiLabel-specific features to yield stronger generalization performance for multi-label classification.

A Technique for the Laboratory Determination of Recirculation in Single Needle Dialysis

1993 ◽  
Vol 16 (2) ◽  
pp. 63-70 ◽  
Author(s):  
N.A. Hoenich ◽  
P.T. Smirthwaite ◽  
C. Woffindin ◽  
P. Lancaster ◽  
T.H. Frost ◽  
...  
Keyword(s):  
Experimental Data ◽  
Flow Rate ◽  
Experimental Results ◽  
New Technique ◽  
Treatment Efficiency ◽  
Single Lumen ◽  
Theoretical Predictions ◽  
Lumen Catheter ◽  
A New Technique

Recirculation is an important factor in single needle dialysis and, if high, can compromise treatment efficiency. To provide information regarding recirculation characteristics of access devices used in single needle dialysis, we have developed a new technique to characterise recirculation and have used this to measure the recirculation of a Terumo 15G fistula needle and a VasCath SC2300 single lumen catheter. The experimentally obtained results agreed well with those established clinically (8.5 ± 2.4% and 18.4 ± 3.4%). The experimental results have also demonstrated a dependence on access type, pump speeds and fistula flow rate. A comparison of experimental data with theoretical predictions showed that the latter exceeded those measured with the largest contribution being due to the experimental fistula.

IMPROVED TIME-MULTIPLEXED FPGA ARCHITECTURE AND ALGORITHM FOR MINIMIZING COMMUNICATION COST DESIGNS

2013 ◽  
Vol 22 (05) ◽  
pp. 1350033
Author(s):  
CHI-CHOU KAO ◽  
YEN-TAI LAI
Keyword(s):  
Greedy Algorithm ◽  
Sequential Circuits ◽  
Precedence Constraints ◽  
Communication Cost ◽  
Experimental Results ◽  
Time Sharing ◽  
Fpga Architecture ◽  
The Given

The Time-Multiplexed FPGA (TMFPGA) architecture can improve dramatically logic utilization by time-sharing logic but it needs a large amount of registers among sub-circuits for partitioning the given sequential circuits. In this paper, we propose an improved TMFPGA architecture to simplify the precedence constraints so that the number of the registers among sub-circuits can be reduced for sequential circuits partitioning. To demonstrate the practicability of the architecture, we also present a greedy algorithm to minimize the maximum number of the registers. Experimental results demonstrate the effectives of the algorithm.

Experimental Investigations of the Machinability of Titanium Alloy TC4 with Different Hydrogen Contents

2011 ◽  
Vol 418-420 ◽  
pp. 1307-1311
Author(s):  
Jun Hu ◽  
Yong Jie Bao ◽  
Hang Gao ◽  
Ke Xin Wang
Keyword(s):  
Surface Roughness ◽  
Titanium Alloy ◽  
Cutting Force ◽  
Hydrogen Content ◽  
Experimental Results ◽  
Cutting Conditions ◽  
Experimental Investigations ◽  
Cutting Condition ◽  
The Given

The experiments were carried out in the paper to investigate the effect of adding hydrogen in titanium alloy TC4 on its machinability. The hydrogen contents selected were 0, 0.25%, 0.49%, 0.63%, 0.89% and 1.32%, respectively. Experiments with varing hydrogen contents and cutting conditions concurrently. Experimental results showed that the cutting force of the titanium alloy can be obviously reduced and the surface roughness can be improved by adding appropriate hydrogen in the material. In the given cutting condition, the titanium alloy TC4 with 0.49% hydrogen content showed better machinability.

Mining Allocating Patterns in Investment Portfolios

2009 ◽  
pp. 110-135
Author(s):  
Yanbo J. Wang ◽  
Xinwei Zheng ◽  
Frans Coenen
Keyword(s):  
Data Mining ◽  
Portfolio Management ◽  
Association Rule ◽  
Common Type ◽  
Experimental Results ◽  
Transaction Database ◽  
Investment Portfolios ◽  
The Given

An association rule (AR) is a common type of mined knowledge in data mining that describes an implicative co-occurring relationship between two sets of binary-valued transaction-database attributes, expressed in the form of an ? rule. A variation of ARs is the (WARs), which addresses the weighting issue in ARs. In this chapter, the authors introduce the concept of “one-sum” WAR and name such WARs as allocating patterns (ALPs). An algorithm is proposed to extract hidden and interesting ALPs from data. The authors further indicate that ALPs can be applied in portfolio management. Firstly by modelling a collection of investment portfolios as a one-sum weighted transaction- database that contains hidden ALPs. Secondly the authors show that ALPs, mined from the given portfolio-data, can be applied to guide future investment activities. The experimental results show good performance that demonstrates the effectiveness of using ALPs in the proposed application.

A Fuzzy Adaptive K-SVD Dictionary Algorithm for Face Recogntion

2013 ◽  
Vol 347-350 ◽  
pp. 3797-3803 ◽  
Author(s):  
Xiao Ning Song ◽  
Zi Liu
Keyword(s):  
Iterative Method ◽  
Fuzzy Set ◽  
Sparse Coding ◽  
Sparse Representations ◽  
Experimental Results ◽  
Sparse Signal ◽  
Signal Representations ◽  
The Given ◽  
Class Assignment ◽  
Fuzzy Adaptive

Sparse representations using overcomplete dictionaries has concentrated mainly on the study of pursuit algorithms that decompose signals with respect to a given dictionary. Designing dictionaries to better fit the above model can be done by either selecting one from a prespecified set of linear transforms or adapting the dictionary to a set of training signals. The K-SVD algorithm is an iterative method that alternates between sparse coding of the examples based on the current dictionary and a process of updating the dictionary atoms to better fit the data. However, the existing K-SVD algorithm is employed to dwell on the concept of a binary class assignment meaning that the multi-classes samples are assigned to the given classes definitely. The work proposed in this paper provides a novel fuzzy adaptive way to adapting dictionaries in order to achieve the fuzzy sparse signal representations, the update of the dictionary columns is combined with an update of the sparse representations by incorporated a new mechanism of fuzzy set, which is called fuzzy K-SVD. Experimental results conducted on the ORL and Yale face databases demonstrate the effectiveness of the proposed method.

scholarly journals Actions as a Basis for Online Embodied Concepts

Languages ◽  
2019 ◽  
Vol 4 (1) ◽  
pp. 16
Author(s):  
Holly Keily
Keyword(s):  
Embodied Cognition ◽  
Experimental Results ◽  
Matching Task ◽  
Video Clips ◽  
Motor Information ◽  
Physical Attributes ◽  
Main Action ◽  
Embodied Action ◽  
Embodied Representation ◽  
Short Event

In co-speech gesture research, embodied cognition implies that concepts are associated with haptic and motor information that provides a framework for a gestural plan. When speakers access concepts, embodied action images are automatically activated. This study considers situations in which speakers need to create online concepts of events to investigate the aspect of the event that forms the basis of a new concept. Speakers watched short event video clips with familiar or unfamiliar attributes. They described those clips to partners who had to perform a matching task. Experimental results show that speakers gestured less and produced shorter gestures when relaying longer event descriptions. Speakers were more likely to produce gesture when some aspect of the event was unfamiliar, and they were most sensitive to the familiarity of the event’s main action. Further, when speakers did gesture, they were most likely to gesture to represent the action of the event over the physical attributes of it (the instrument used to enact or the object acted upon). These findings suggest that in creating an embodied concept for something unfamiliar, the motion of the event acts as a basis for their online embodied representation of the concept.

scholarly journals Aspect-augmented Adversarial Networks for Domain Adaptation

2017 ◽  
Vol 5 ◽  
pp. 515-528 ◽  
Author(s):  
Yuan Zhang ◽  
Regina Barzilay ◽  
Tommi Jaakkola
Keyword(s):  
Transfer Learning ◽  
Domain Adaptation ◽  
Experimental Results ◽  
Dependent Manner ◽  
Target Classification ◽  
Adversarial Networks ◽  
Adversarial Training ◽  
Classification Tasks ◽  
Class Labels

We introduce a neural method for transfer learning between two (source and target) classification tasks or aspects over the same domain. Rather than training on target labels, we use a few keywords pertaining to source and target aspects indicating sentence relevance instead of document class labels. Documents are encoded by learning to embed and softly select relevant sentences in an aspect-dependent manner. A shared classifier is trained on the source encoded documents and labels, and applied to target encoded documents. We ensure transfer through aspect-adversarial training so that encoded documents are, as sets, aspect-invariant. Experimental results demonstrate that our approach outperforms different baselines and model variants on two datasets, yielding an improvement of 27% on a pathology dataset and 5% on a review dataset.

Sign in / Sign up

Export Citation Format

Share Document