State-of-the-art image classification methods require an intensive learning/training stage (using SVM, Boosting, etc.) In contrast, non-parametric nearest-neighbor (NN) based image classifiers require no training time and have other favorable properties. However, the large performance gap between these two families of approaches rendered NN-based image classifiers useless. We claim that the effectiveness of non-parametric NN-based image classification has been considerably undervalued. We argue that two practices commonly used in image classification methods, have led to the inferior performance of NN-based image classifiers: (i) Quantization of local image descriptors (used to generate "bags-of-words ", codebooks). (ii) Computation of 'image-to-image' distance, instead of 'image-to-class' distance. We propose a trivial NN-based classifier - NBNN, (Naive-Bayes nearest-neighbor), which employs NN- distances in the space of the local image descriptors (and not in the space of images). NBNN computes direct 'image- to-class' distances without descriptor quantization. We further show that under the Naive-Bayes assumption, the theoretically optimal image classifier can be accurately approximated by NBNN. Although NBNN is extremely simple, efficient, and requires no learning/training phase, its performance ranks among the top leading learning-based image classifiers. Empirical comparisons are shown on several challenging databases (Caltech-101 ,Caltech-256 and Graz-01).

We propose a new approach for measuring similarity between two signals, which is applicable to many machine learning tasks, and to many signal types. We say that a signal S1 is “similar” to a signal S2 if it is “easy” to compose S1 from few large contiguous chunks of S2. Obviously, if we use small enough pieces, then any signal can be composed of any other. Therefore, the larger those pieces are, the more similar S1 is to S2. This induces a local similarity score at every point in the signal, based on the size of its supported surrounding region. These local scores can in turn be accumulated in a principled information-theoretic way into a global similarity score of the entire S1 to S2. “Similarity by Composition” can be applied between pairs of signals, between groups of signals, and also between dif- ferent portions of the same signal. It can therefore be employed in a wide variety of machine learning problems (clustering, classification, retrieval, segmentation, attention, saliency, labelling, etc.), and can be applied to a wide range of signal types (images, video, audio, biological data, etc.) We show a few such examples.

We address the problem of detecting irregularities in visual data, e.g., detecting suspicious behaviors in video sequences, or identifying salient patterns in images. The term "irregular" depends on the context in which the "regular" or "valid" are defined. Yet, it is not realistic to expect explicit definition of all possible valid configurations for a given context. We pose the problem of determining the validity of visual data as a process of constructing a puzzle: We try to compose a new observed image region or a new video segment ("the query") using chunks of data ("pieces of puzzle") extracted from previous visual examples ("the database "). Regions in the observed data which can be composed using large contiguous chunks of data from the database are considered very likely, whereas regions in the observed data which cannot be composed from the database (or can be composed, but only using small fragmented pieces) are regarded as unlikely/suspicious. The problem is posed as an inference process in a probabilistic graphical model. We show applications of this approach to identifying saliency in images and video, and for suspicious behavior recognition.