Discriminative and consistent similarities in instance-level Multiple Instance Learning

@article{Rastegari2015DiscriminativeAC,
  title={Discriminative and consistent similarities in instance-level Multiple Instance Learning},
  author={Mohammad Rastegari and Hannaneh Hajishirzi and Ali Farhadi},
  journal={2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2015},
  pages={740-748}
}
In this paper we present a bottom-up method to instance-level Multiple Instance Learning (MIL) that learns to discover positive instances with globally constrained reasoning about local pairwise similarities. We discover positive instances by optimizing for a ranking such that positive (top rank) instances are highly and consistently similar to each other and dissimilar to negative instances. Our approach takes advantage of a discriminative notion of pairwise similarity coupled with a… 

Figures and Tables from this paper

Sparse multiple instance learning as document classification

The proposed methods achieve significantly higher accuracies and AUC than the state-of-the-art in a large number of sparse MIL problems, and the document classification analogy explains their efficacy in sparse Mil problems.

Modeling and Optimization of Classifiers with Latent Variables

A new optimization framework, called Generalized MajorizationMinimization (G-MM), is introduced that extends existing approaches to non-convex optimization such as Expectation Maximization and Concave Convex Procedure (CCCP), and does not require bounds to be tight, making it very flexible.

Generalized Majorization-Minimization

This work derives G-MM algorithms for several latent variable models and shows empirically that they consistently outperform their MM counterparts in optimizing non-convex objectives, and appears to be less sensitive to initialization.

Aligning Sentences from Standard Wikipedia to Simple Wikipedia

This work improves monolingual sentence alignment for text simplification, specifically for text in standard and simple Wikipedia by using a greedy search over the document and a word-level semantic similarity score based on Wiktionary that also accounts for structural similarity through syntactic dependencies.

cvpaper.challenge in 2015 - A review of CVPR2015 and DeepSurvey

This review focused on reading the ALL 602 conference papers presented at the CVPR2015, the premier annual computer vision event held in June 2015, and proposed "DeepSurvey" as a mechanism embodying the entire process from the reading through all the papers, the generation of ideas, and to the writing of paper.

cvpaper.challenge in 2016: Futuristic Computer Vision through 1, 600 Papers Survey

The paper gives futuristic challenges disscussed in the cvpaper.challenge. In 2015 and 2016, we thoroughly study 1,600+ papers in several conferences/journals such as CVPR/ICCV/ECCV/NIPS/PAMI/IJCV.

PIGMIL: Positive Instance Detection via Graph Updating for Multiple Instance Learning

A positive instance detection via graph updating for multiple instance learning, called PIGMIL, is proposed, to detect TPI accurately and its excellent performance compared to classic baseline MIL methods is demonstrated.

cvpaper.challenge in CVPR2015 -- A review of CVPR2015

This challenge aims to simultaneously read papers and create documents for easy understanding top conference papers in Japanese in the fields of computer vision, image processing, pattern recognition and machine learning.

References

SHOWING 1-10 OF 48 REFERENCES

SMILE: A Similarity-Based Approach for Multiple Instance Learning

A novel MIL method named SMILE (Similarity-based Multiple Instance LEarning), which introduces a similarity weight to each instance in positive bag, which represents the instance similarity towards the positive and negative classes.

Instance-level Semisupervised Multiple Instance Learning

This paper proposes a new graph-based semi-supervised learning approach for multiple instance learning by defining an instance-level graph on the data, and empirically shows that this method outperforms state-of-the-art MIL algorithms on several real-world data sets.

MILES: Multiple-Instance Learning via Embedded Instance Selection

This work proposes a learning method, MILES (multiple-instance learning via embedded instance selection), which converts the multiple- instance learning problem to a standard supervised learning problem that does not impose the assumption relating instance labels to bag labels.

Multi-instance learning by treating instances as non-I.I.D. samples

This paper explicitly map every bag to an undirected graph and design a graph kernel for distinguishing the positive and negative bags and implicitly construct graphs by deriving affinity matrices and propose an efficient graph kernel considering the clique information.

Multiple Instance Learning by Discriminative Training of Markov Networks

Experimental results verify that encoding or learning the degree of ambiguity can improve classification performance and a discriminative max-margin learning algorithm leveraging efficient inference for cardinality-based cliques is proposed.

Convex Multiple-Instance Learning by Estimating Likelihood Ratio

An approach to multiple-instance learning is proposed that reformulates the problem as a convex optimization on the likelihood ratio between the positive and the negative class for each training instance, and shows that likelihood ratio estimation is generally a good surrogate for the 0-1 loss.

FAMER: Making Multi-Instance Learning Better and Faster

FAMER constructs a Locally Sensitive Hashing based similarity measure for multi-instance framework, and represents each bag as a histogram by embedding instances within the bag into an auxiliary space, which captures the correspondence information between two bags.

A Conditional Random Field for Multiple-Instance Learning

MI-CRF models bags as nodes in a CRF with instances as their states and combines discriminative unary instance classifiers and pairwise dissimilarity measures to improve the classification performance.

A Framework for Multiple-Instance Learning

A new general framework, called Diverse Density, is described, which is applied to learn a simple description of a person from a series of images containing that person, to a stock selection problem, and to the drug activity prediction problem.

Ensemble of exemplar-SVMs for object detection and beyond

This paper proposes a conceptually simple but surprisingly powerful method which combines the effectiveness of a discriminative object detector with the explicit correspondence offered by a