Multiple Instance Learning Method Research Articles

Cost-sensitive multiple-instance learning method with dynamic transactional data for personal credit scoring

We study how to assess an applicant’s credit risk with dynamic transactional data. The problem arises when an applicant applies for loans from financial institutions. A traditional credit-risk assessment model utilizes individual demographic and loan information from an application form. Nevertheless, dynamic transactional data is good indicators of an applicant’s credit risk. However, the lack of available data and the preexisting limitations of conventional approaches limit the use of the dynamic transactional data. In this study, we propose a cost-sensitive multiple-instance learning (MIL) approach to evaluate applicants’ credit scores that incorporate their dynamic transactional data and static individual information. Traditionally, MIL approaches can handle the variable number of input instances. However, to facilitate the implementation of MIL into credit scoring, we extend the MIL to consider the dynamic transactional data and cost-sensitive problem simultaneously. We compare our model with several benchmark MIL models by testing them on real-world data sets. Experimental results show that our model outperforms most benchmarks in many widely used criteria.

A similarity-based two-view multiple instance learning method for classification

Multiple instance learning (MIL) has been proposed to classify the bag of instances. In practice, we may meet the problems which have more than one view data. For example, in the image classification, textual information is always used to describe the image, which can be considered as two-view data. In this paper, we propose a new similarity-based two-view multi-instance learning (STMIL) method that can incorporate two-view data into learning so as to improve classification accuracy of MIL. In order to obtain the predictive classifier, we first convert the proposed model into a convex optimization problem, and then propose a new alternative framework to solve the proposed method. We then analyze the convergence of the proposed STMIL method. The experiments have been conducted to compare the performance of our proposed method and the previous methods. The results show that our method can deliver superior performance than other methods.

Weakly Labelled AudioSet Tagging With Attention Neural Networks

Audio tagging is the task of predicting the presence or absence of sound classes within an audio clip. Previous work in audio tagging focused on relatively small datasets limited to recognising a small number of sound classes. We investigate audio tagging on AudioSet, which is a dataset consisting of over 2 million audio clips and 527 classes. AudioSet is weakly labelled, in that only the presence or absence of sound classes is known for each clip, while the onset and offset times are unknown. To address the weakly-labelled audio tagging problem, we propose attention neural networks as a way to attend the most salient parts of an audio clip. We bridge the connection between attention neural networks and multiple instance learning (MIL) methods, and propose decision-level and feature-level attention neural networks for audio tagging. We investigate attention neural networks modeled by different functions, depths and widths. Experiments on AudioSet show that the feature-level attention neural network achieves a state-of-the-art mean average precision (mAP) of 0.369, outperforming the best multiple instance learning (MIL) method of 0.317 and Google's deep neural network baseline of 0.314. In addition, we discover that the audio tagging performance on AudioSet embedding features has a weak correlation with the number of training samples and the quality of labels of each sound class.

Deep multiple instance learning for automatic detection of diabetic retinopathy in retinal images

As a weakly supervised learning technique, multiple instance learning (MIL) has shown an advantage over supervised learning methods for automatic detection of diabetic retinopathy (DR): only the image-level annotation is needed to achieve both detection of DR images and DR lesions, making more graded and de-identified retinal images available for learning. However, the performance of existing studies on this technique is limited by the use of handcrafted features. The authors propose a deep MIL method for DR detection, which jointly learns features and classifiers from data and achieves a significant improvement on detecting DR images and their inside lesions. Specifically, a pre-trained convolutional neural network is adapted to achieve the patch-level DR estimation, and then global aggregation is used to make the classification of DR images. Further, the authors propose an end-to-end multi-scale scheme to better deal with the irregular DR lesions. For detection of DR images, they achieve an area under the ROC curve of 0.925 on a subset of a Kaggle dataset, and 0.960 on Messidor. For detection of DR lesions, they achieve an F1-score of 0.924 with sensitivity 0.995 and precision 0.863 on DIARETDB1 using the connected component-level validation.

Sub-Category Classifiers for Multiple-Instance Learning and its Application to Retinal Nerve Fiber Layer Visibility Classification.

We propose a novel multiple instance learning method to assess the visibility (visible/not visible) of the retinal nerve fiber layer (RNFL) in fundus camera images. Using only image-level labels, our approach learns to classify the images as well as to localize the RNFL visible regions. We transform the original feature space to a discriminative subspace, and learn a region-level classifier in that subspace. We propose a margin-based loss function to jointly learn this subspace and the region-level classifier. Experiments with a RNFL dataset containing 884 images annotated by two ophthalmologists give a system-annotator agreement (kappa values) of 0:73 and 0:72 respectively, with an inter-annotator agreement of 0:73. Our system agrees better with the more experienced annotator. Comparative tests with three public datasets (MESSIDOR and DR for diabetic retinopathy, UCSB for breast cancer) show that our novel MIL approach improves performance over the state-of-the-art. Our Matlab code is publicly available at https://github.com/ManiShiyam/Sub-category-classifiersfor- Multiple-Instance-Learning/wiki.

Identifying User-Specific Facial Affects from Spontaneous Expressions with Minimal Annotation

This paper presents Personalized Affect Detection with Minimal Annotation (PADMA), a user-dependent approach for identifying affective states from spontaneous facial expressions without the need for expert annotation. The conventional approach relies on the use of key frames in recorded affect sequences and requires an expert observer to identify and annotate the frames. It is susceptible to user variability and accommodating individual differences is difficult. The alternative is a user-dependent approach, but it would be prohibitively expensive to collect and annotate data for each user. PADMA uses a novel Association-based Multiple Instance Learning (AMIL) method, which learns a personal facial affect model through expression frequency analysis, and does not need expert input or frame-based annotation. PADMA involves a training/calibration phase in which the user watches short video segments and reports the affect that best describes his/her overall feeling throughout the segment. The most indicative facial gestures are identified and extracted from the facial response video, and the association between gesture and affect labels is determined by the distribution of the gesture over all reported affects. Hence both the geometric deformation and distribution of key facial gestures are specially adapted for each user. We show results that demonstrate the feasibility, effectiveness and extensibility of our approach.

Heart beat characterization from ballistocardiogram signals using extended functions of multiple instances.

A multiple instance learning (MIL) method, extended Function of Multiple Instances (eFUMI), is applied to ballistocardiogram (BCG) signals produced by a hydraulic bed sensor. The goal of this approach is to learn a personalized heartbeat "concept" for an individual. This heartbeat concept is a prototype (or "signature") that characterizes the heartbeat pattern for an individual in ballistocardiogram data. The eFUMI method models the problem of learning a heartbeat concept from a BCG signal as a MIL problem. This approach elegantly addresses the uncertainty inherent in a BCG signal (e. g., misalignment between training data and ground truth, mis-collection of heartbeat by some transducers, etc.). Given a BCG training signal coupled with a ground truth signal (e.g., a pulse finger sensor), training "bags" labeled with only binary labels denoting if a training bag contains a heartbeat signal or not can be generated. Then, using these bags, eFUMI learns a personalized concept of heartbeat for a subject as well as several non-heartbeat background concepts. After learning the heartbeat concept, heartbeat detection and heart rate estimation can be applied to test data. Experimental results show that the estimated heartbeat concept found by eFUMI is more representative and a more discriminative prototype of the heartbeat signals than those found by comparison MIL methods in the literature.

Learning Discriminative Pattern for Real-Time Car Brand Recognition

In this paper, we study the problem of recognizing car brands in surveillance videos, cast it as an image classification problem, and propose a novel multiple instance learning method, named Spatially Coherent Discriminative Pattern Learning, to discover the most discriminative patterns in car images. The learned discriminative patterns can effectively distinguish cars of different brands with high accuracy and efficiency. The experimental results demonstrate that our method is significantly superior to recent image classification methods on this problem. The proposed method is able to deliver an end-to-end real-time car recognition system for video surveillance. Moreover, we construct a large and challenging car image data set, consisting of 37 195 real-world car images from 30 brands, which could serve as a standard benchmark in this field and be used in various related research communities.

Multiple instance learning for computer aided detection and diagnosis of gastric cancer with dual-energy CT imaging

Multiple instance learning algorithms have been increasingly utilized in computer aided detection and diagnosis field. In this study, we propose a novel multiple instance learning method for the identification of tumor invasion depth of gastric cancer with dual-energy CT imaging. In the proposed scheme, two level features, bag-level features and instance-level features are extracted for subsequent processing and classification work. For instance-level features, there is some ambiguity in assigning labels to selected patches. An improved Citation-KNN method is presented to solve this problem. Compared with benchmarking state-of-the-art multiple instance learning algorithms using the same clinical dataset, the proposed algorithm can achieve improved results. The experimental evaluation is performed using leave-one-out cross validation with the total accuracy of 0.7692. The proposed multiple instance learning algorithm serves as an alternative method for computer aided diagnosis and identification of tumor invasion depth of gastric cancer with dual-energy CT imaging techniques.

Human Action Recognition Based on Improved MIL

According to the problem that achieves robust human actions recognition from image sequences in computer vision, using the Iterative Querying Heuristic algorithm as a guide, a improved Multiple Instance Learning (MIL) method is proposed for human action recognition in video image sequences. Experiments show that the new method can quickly recognize human actions and achieve high recognition rates, and on the Weizmann database validate our analysis.

A Novel Classification Method for Syndrome Differentiation of Patients with AIDS.

We consider the analysis of an AIDS dataset where each patient is characterized by a list of symptoms and is labeled with one or more TCM syndromes. The task is to build a classifier that maps symptoms to TCM syndromes. We use the minimum reference set-based multiple instance learning (MRS-MIL) method. The method identifies a list of representative symptoms for each syndrome and builds a Gaussian mixture model based on them. The models for all syndromes are then used for classification via Bayes rule. By relying on a subset of key symptoms for classification, MRS-MIL can produce reliable and high quality classification rules even on datasets with small sample size. On the AIDS dataset, it achieves average precision and recall 0.7736 and 0.7111, respectively. Those are superior to results achieved by alternative methods.

A novel multiple instance learning method based on extreme learning machine.

Since real-world data sets usually contain large instances, it is meaningful to develop efficient and effective multiple instance learning (MIL) algorithm. As a learning paradigm, MIL is different from traditional supervised learning that handles the classification of bags comprising unlabeled instances. In this paper, a novel efficient method based on extreme learning machine (ELM) is proposed to address MIL problem. First, the most qualified instance is selected in each bag through a single hidden layer feedforward network (SLFN) whose input and output weights are both initialed randomly, and the single selected instance is used to represent every bag. Second, the modified ELM model is trained by using the selected instances to update the output weights. Experiments on several benchmark data sets and multiple instance regression data sets show that the ELM-MIL achieves good performance; moreover, it runs several times or even hundreds of times faster than other similar MIL algorithms.

Nighttime Motion Vehicle Detection Based on Multiple Instance Learning

In this paper, we propose an effective approach for detecting moving vehicles in nighttime traffic scenes. We use Multiple Instance Learning method to automatically detect vehicle from video sequences by constructing the Multiple Instance Learning model at nighttime. At first, we extract SIFT feature using SIFT feature extraction algorithm, which is used to characterize moving vehicles at nighttime. Then Multiple Instance Learning model is used for the on-road detection of vehicles at nighttime, in order to improve the detection accuracy, the class label information was used for the learning of the Multiple Instance Learning model. Final experiments were performed and evaluate the proposed method at nighttime under urban traffic condition, the experiment results show that the average detection accuracy is over 96.2%, which validates that the proposed vehicle detection approach is feasible and effective for the on-road detection of vehicles at nighttime and identification in various nighttime environments.

Multiple instance learning for classification of dementia in brain MRI.

Machine learning techniques have been widely used to detect morphological abnormalities from structural brain magnetic resonance imaging data and to support the diagnosis of neurological diseases such as dementia. In this paper, we propose to use a multiple instance learning (MIL) method in an application for the detection of Alzheimer's disease (AD) and its prodromal stage mild cognitive impairment (MCI). In our work, local intensity patches are extracted as features. However, not all the patches extracted from patients with dementia are equally affected by the disease and some of them may not be characteristic of morphology associated with the disease. Therefore, there is some ambiguity in assigning disease labels to these patches. The problem of the ambiguous training labels can be addressed by weakly supervised learning techniques such as MIL. A graph is built for each image to exploit the relationships among the patches and then to solve the MIL problem. The constructed graphs contain information about the appearances of patches and the relationships among them, which can reflect the inherent structures of images and aids the classification. Using the baseline MR images of 834 subjects from the ADNI study, the proposed method can achieve a classification accuracy of 89% between AD patients and healthy controls, and 70% between patients defined as stable MCI and progressive MCI in a leave-one-out cross validation. Compared with two state-of-the-art methods using the same dataset, the proposed method can achieve similar or improved results, providing an alternative framework for the detection and prediction of neurodegenerative diseases.

Visual Tracking Based on Particle Filter and Weighted Multiple Instance Learning

Visual tracking is one of fundamental tasks in the field of computer vision. In recent years, detection algorithms have been more and more interested in the use of discriminative classifiers for tracking system. Template drift is the major shortcoming of the detection algorithms because of the online self-learning mechanism of the visual tracker. Multiple Instance Learning (MIL) method has been applied to target tracking, which can alleviate the drift to some extent. However, the MIL tracker cannot discriminatively consider the importance of sample and instance in its learning procedure. Particle filter is employed to make the use of the learned classifier and to help generating a better representative set of training samples for the online learning. In this paper, we present a tracking method built by particle filter and weighted MIL tracker, which integrates the sample and instance importance into an efficient online learning procedure when the classifier is being trained. At the end, the experimental results on various videos verify that the proposed method has a satisfaction performance in real-time object tracking.

Combining example selection with instance selection to speed up multiple-instance learning

Recently, several instance selection-based methods have been presented to solve the multiple-instance learning (MIL) problem. The basic idea is converting MIL into standard supervised learning by selecting some representative instance prototypes from the training set. However, training examples are not single instances but bags composed of one or more instances in MIL, so the computational complexity is often very high. Previous methods consider this issue only from the perspective of instance selection not from that of example selection. In this paper, we try to address this issue via combining example selection with instance selection. Three general example selection methods are derived by adapting three immune-inspired algorithms to MIL. Additionally, we propose a simple instance selection method for MIL based on the probability that an instance is positive given a set of negative instances. Our example selection methods are combined with the new MIL method and other previous instance selection-based ones as a preprocessing step. The theoretical analysis and empirical results show that our MIL method is competitive to the state-of-the-art and the proposed example selection methods could significantly speed up various instance selection-based MIL methods with slightly weakening their performance or even strengthening it.

Image Annotation by Multiple-Instance Learning With Discriminative Feature Mapping and Selection

Multiple-instance learning (MIL) has been widely investigated in image annotation for its capability of exploring region-level visual information of images. Recent studies show that, by performing feature mapping, MIL can be cast to a single-instance learning problem and, thus, can be solved by traditional supervised learning methods. However, the approaches for feature mapping usually overlook the discriminative ability and the noises of the generated features. In this paper, we propose an MIL method with discriminative feature mapping and feature selection, aiming at solving this problem. Our method is able to explore both the positive and negative concept correlations. It can also select the effective features from a large and diverse set of low-level features for each concept under MIL settings. Experimental results and comparison with other methods demonstrate the effectiveness of our approach.

Multiple Instance Learning of Visual Event Models

In this paper we propose a powerful visual event pattern learning method to address the issue of high-level video understanding. We first model the deformable temporal structure of the action event in videos by a temporal composition of several primitive motions. Moreover, we describe each action class by multiple temporal models to deal with the significant intra-class variability. We implement a multiple instance learning method to train the models in the weakly supervised setting. We have conducted experiments on three major benchmarks. The results are comparative to the state-of-the-arts.

Nonparallel Support Vector Machines for Multiple-Instance Learning

In this paper, we proposed a new multiple-instance learning (MIL) method based on nonparallel support vector machines (called MI-NPSVM). For the linear case, MI-NPSVM constructs two nonparallel hyperplanes by solving two SVM-type prob- lems, which is different from many other maximum margin SVM-based MIL methods. For the nonlinear case, kernel functions can be easily applied to extend the linear case, which is different from other nonparallel SVM-based MIL methods. Further- more, compared with the existing MIL method based on nonparallel SVM – MI-TSVM, MI-NPSVM has two main advantages. Firstly the method enforces the structural risk minimization; secondly it does not need to solve a bilevel programming prob- lem anymore, but to solve a series of standard Quadratic Programming Problems (QPPs). All experimental results on public datasets show that our method is superior to the traditional MIL methods like MI-SVM, MI-TSVM etc.

Real-time visual tracking via online weighted multiple instance learning

Adaptive tracking-by-detection methods have been widely studied with promising results. These methods first train a classifier in an online manner. Then, a sliding window is used to extract some samples from the local regions surrounding the former object location at the new frame. The classifier is then applied to these samples where the location of sample with maximum classifier score is the new object location. However, such classifier may be inaccurate when the training samples are imprecise which causes drift. Multiple instance learning (MIL) method is recently introduced into the tracking task, which can alleviate drift to some extent. However, the MIL tracker may detect the positive sample that is less important because it does not discriminatively consider the sample importance in its learning procedure. In this paper, we present a novel online weighted MIL (WMIL) tracker. The WMIL tracker integrates the sample importance into an efficient online learning procedure by assuming the most important sample (i.e., the tracking result in current frame) is known when training the classifier. A new bag probability function combining the weighted instance probability is proposed via which the sample importance is considered. Then, an efficient online approach is proposed to approximately maximize the bag likelihood function, leading to a more robust and much faster tracker. Experimental results on various benchmark video sequences demonstrate the superior performance of our algorithm to state-of-the-art tracking algorithms.