No description available.

Multi-instance (MI) learning is a variant of supervised machine learning, where each learning example contains a bag of instances instead of just a single feature vector. MI learning has applications in areas such as drug activity prediction, fruit disease management and image classification. This thesis investigates the case where each instance has a weight value determining the level of influence that it has on its bag's class label. This is a more general assumption than most existing approaches use, and thus is more widely applicable. The challenge is to accurately estimate these weights in order to make predictions at the bag level. An existing approach known as MILES is retroactively identified as an algorithm that uses instance weights for MI learning, and is evaluated using a variety of base learners on benchmark problems. New algorithms for learning instance weights for MI learning are also proposed and rigorously evaluated on both artificial and real-world datasets. The new algorithms are shown to achieve better root mean squared error rates than existing approaches on artificial data generated according to the algorithms' underlying assumptions. Experimental results also demonstrate that the new algorithms are competitive with existing approaches on real-world problems.

The implementation of data mining techniques in the medical area has generated great interest because of its potential for more efficient, economic and robust performance when compared to physicians. In this paper, we focus on the implementation of Multiple-Instance Learning (MIL) in the area of medical image mining, particularly to hard exudates detection in retinal images from diabetic patients. Our proposed approach deals with the highly noisy images that are common in the medical area, improving the detection specificity while keeping the sensitivity as high as possible. We have also investigated the effect of feature selection on system performance. We describe how we implement the idea of MIL on the problem of retinal image mining, discuss the issues that are characteristic of retinal images as well as issues common to other medical image mining problems, and report the results of initial experiments. / Singapore-MIT Alliance (SMA)

The objective of the present work is to propose a novel method to extract a stable feature set representative of image content. Each image is represented by a linear combination of fractal orthonormal basis vectors. The mapping coefficients of an image projected onto each orthonormal basis constitute the feature vector. The set of orthonormal basis vectors are generated by utilizing fractal iterative function through target and domain blocks mapping. The distance measure remains consistent, i.e., isometric embedded, between any image pairs before and after the projection onto orthonormal axes. Not only similar images generate points close to each other in the feature space, but also dissimilar ones produce feature points far apart. The above statements are logically equivalent to that distant feature points are guaranteed to map to images with dissimilar contents, while close feature points correspond to similar images. In this paper, we adapt the Multiple Instance Learning paradigm using the Diverse Density algorithm as a way of modeling the ambiguity in images in order to learning concepts used to classify images. A user labels an image as positive if the image contains the concepts, as negative if the image far from the concepts. Each example image is a bag of blocks where only the bag is labeled. The User selects positive and negative image examples to train the concepts in feature space. From a small collection of positive and negative examples, the system learns the concepts using them to retrieve images that contain the concepts from database. Each concept having similar blocks becomes the group in each image. According groups¡¦ location distribution, variation and spatial relations computes positive examples and database images similarity.

During the Digital Period, the more people using these digital video. When there are more and more users and amount of video data, the management of video data becomes a significant dimension during development. Therefore, there are more and more studying of accomplishing video database system, which provide users to search and get them. In this paper, a novel method for Video Scene Change Detection and video database retrieval is proposed. Uses Fractal orthonormal bases to guarantee the similar index has the similar image the characteristic union support vector clustering, splits a video into a sequence of shots, extracts a few representative frames(key-frames) to take the video database index from each shot. When image search compared to, according to MIL to pick up the characteristic, which images pursues the video database to have the similar characteristic, computation similar, makes the place output according to this.

No description available.

Prostate cancer is one of the most common types of cancer occurring in men. Several types of research have been done using deep learning methods for the classification/prediction of cancer grades. In this thesis, the results of prostate cancer risk prediction, based only on the images of cells from the prostate tissues, have been analyzed. Cell images from the prostate tissues were extracted using a deep learning based segmentation model. These cell images were then used in a Multiple Instance Learning model for cancer risk prediction. An attention mechanism was used to visualize the regions in the tissue to which the model paid more attention. The results suggest that the Multiple Instance Learning (MIL) model achieves an Area Under the receiver Operating Characteristics (AUROC) of 0.641 ± 0.013, which is better than a random model for low-risk vs. high-risk cancer prediction. The model’s prediction was made on cell images, with the glandular information destroyed. The MIL model, however, performs worse than a model which gets to see the glandular architecture of the cells in the prostate tissues. / Prostatacancer är en av de vanligaste typerna av cancer som förekommer ho smän. Flera typer av forskning har gjorts med metoder för djupinlärning förklassificering/förutsägelse av cancerns malignitetsgrad. I detta examensarbete harresultaten av prostatacancerriskprediktion, baserad enbart på bilder av celler från prostatavävnaderna, analyserats. Cellbilder från prostatavävnaderna extraherades med hjälp av en djupinlärningsbaserad segmenteringsmodell. Dessa cellbilder användes sedan i en Multiple Instance Learning-modell för förutsägelse av cancerrisk. En uppmärksamhetsmekanism användes för att visualisera de regioner i vävnaden som modellen ägnade mer uppmärksamhet åt. Resultaten tyder på att Multiple Instance Learning-modellen uppnår en AUROC på 0.641 ± 0.013, vilket är bättre än en slumpmässig modell för förutsägelse av lågrisk kontra högrisk cancer. Modellens förutsägelse gjordes på cellbilder, med körtelinformationen förstörd. MIL-modellen presterar dock sämre än en modell som får se körtelarkitekturen hos cellerna i prostatavävnaderna.

The objective of the present work is to employ a multiple-instance learning image retrieval system by incorporating a spatial similarity measure. Multiple-Instance learning is a way of modeling ambiguity in supervised learning given multiple examples. From a small collection of positive and negative example images, semantically relevant concepts can be derived automatically and employed to retrieve images from an image database. The degree of similarity between two spatial relations is linked to the distance between the associated nodes in an Interval Neighbor Group (ING). The shorter the distance, the higher degree of similarity, while a longer one, a lower degree of similarity. Once all the pairwise similarity values are derived, an ensemble similarity measure will then integrate these pairwise similarity assessments and give an overall similarity value between two images. Therefore, images in a database can be quantitatively ranked according to the degree of ensemble similarity with the query image. Similarity retrieval method evaluates the ensemble similarity based on the spatial relations and common objects present in the maximum common subimage between the query and a database image are considered. Therefore, reliable spatial relation features extracted from the image, combined with a multiple-instance learning paradigm to derive relevant concepts, can produce desirable retrieval results that better match user¡¦s expectation. In order to demonstrate the feasibility of the proposed approach, two sets of test for querying an image database are performed, namely, the proposed RSS-ING scheme v.s. 2D Be-string similarity method, and single-instance vs. multiple-instance learning. The performance in terms of similarity curves, execution time and memory space requirement show favorably for the proposed multiple-instance spatial similarity-based approach.

abstract: Computer Vision as a eld has gone through signicant changes in the last decade. The eld has seen tremendous success in designing learning systems with hand-crafted features and in using representation learning to extract better features. In this dissertation some novel approaches to representation learning and task learning are studied. Multiple-instance learning which is generalization of supervised learning, is one example of task learning that is discussed. In particular, a novel non-parametric k- NN-based multiple-instance learning is proposed, which is shown to outperform other existing approaches. This solution is applied to a diabetic retinopathy pathology detection problem eectively. In cases of representation learning, generality of neural features are investigated rst. This investigation leads to some critical understanding and results in feature generality among datasets. The possibility of learning from a mentor network instead of from labels is then investigated. Distillation of dark knowledge is used to eciently mentor a small network from a pre-trained large mentor network. These studies help in understanding representation learning with smaller and compressed networks. / Dissertation/Thesis / Doctoral Dissertation Computer Science 2017

No description available.