Abstract — We consider the problem of automatically recognizing human faces from frontal views with varying expression and illumination, as well as occlusion and disguise. We cast the recognition problem as one of classifying among multiple linear regression models, and argue that new theory from sparse signal representation offers the key to addressing this problem. Based on a sparse representation computed by ℓ 1-minimization, we propose a general classification algorithm for (image-based) object recognition. This new framework provides new insights into two crucial issues in face recognition: feature extraction and robustness to occlusion. For feature extraction, we show that if sparsity in the recognition problem is properly harnessed, the choice of features is no longer critical. What is critical, however, is whether the number of features is sufficiently large and whether the sparse representation is correctly computed. Unconventional features such as downsampled images and random projections perform just as well as conventional features such as Eigenfaces and Laplacianfaces, as long as the dimension of the feature space surpasses certain threshold, predicted by the theory of sparse representation. This framework can handle errors due to occlusion and corruption uniformly, by exploiting the fact that these errors are often sparse w.r.t. to the standard (pixel) basis. The theory of sparse representation helps predict how much occlusion the recognition algorithm can handle and how to choose the training images to maximize robustness to occlusion. We conduct extensive experiments on publicly available databases to verify the efficacy of the proposed algorithm, and corroborate the above claims.

Many problems in information processing involve some form of dimensionality reduction. In this paper, we introduce Locality Preserving Projections (LPP). These are linear projective maps that arise by solving a variational problem that optimally preserves the neighborhood structure of the data set. LPP should be seen as an alternative to Principal Component Analysis (PCA) -- a classical linear technique that projects the data along the directions of maximal variance. When the high dimensional data lies on a low dimensional manifold embedded in the ambient space, the Locality Preserving Projections are obtained by finding the optimal linear approximations to the eigenfunctions of the Laplace Beltrami operator on the manifold. As a result, LPP shares many of the data representation properties of nonlinear techniques such as Laplacian Eigenmaps or Locally Linear Embedding. Yet LPP is linear and more crucially is defined everywhere in ambient space rather than just on the training data points. This is borne out by illustrative examples on some high dimensional data sets.

Complex data types—such as images, documents, DNA sequences, etc.—are becoming increasingly important in modern database applications. A typical query in many of these applications seeks to find objects that are similar to some target object, where (dis)similarity is defined by some distance function. Often, the cost of evaluating the distance between two objects is very high. Thus, the number of distance evaluations should be kept at a minimum, while (ideally) maintaining the quality of the result. One way to approach this goal is to embed the data objects in a vector space so that the distances of the embedded objects approximates the actual distances. Thus, queries can be performed (for the most part) on the embedded objects. In this paper, we are especially interested in examining the issue of whether or not the embedding methods will ensure that no relevant objects are left out (i.e., there are no false dismissals and, hence, the correct result is reported). Particular attention is paid to the SparseMap, FastMap, and MetricMap embedding methods. SparseMap is a variant of Lipschitz embeddings, while FastMap and MetricMap are inspired by dimension reduction methods for Euclidean spaces (using KLT or the related PCA and SVD). We show that, in general, none of these embedding methods guarantee that queries on the embedded objects have no false dismissals, while also demonstrating the limited cases in which the guarantee does hold. Moreover, we describe a variant of SparseMap that allows queries with no false dismissals. In addition, we show that with FastMap and MetricMap, the distances of the embedded objects can be much greater than the actual distances. This makes it impossible (or at least impractical) to modify FastMap and MetricMap to guarantee no false dismissals.

The exponential growth of data demands scalable infrastructures capable of indexing and searching rich content such as text, music, and images. A promising direction is to combine information retrieval with peer-to-peer technology for scalability, fault-tolerance, and low administration cost. One pioneering work along this direction is pSearch [32, 33]. pSearch places documents onto a peerto -peer overlay network according to semantic vectors produced using Latent Semantic Indexing (LSI). The search cost for a query is reduced since documents related to the query are likely to be co-located on a small number of nodes. Unfortunately, because of its reliance on LSI, pSearch also inherits the limitations of LSI. (1) When the corpus is large and heterogeneous, LSI's retrieval quality is inferior to methods such as Okapi. (2) The Singular Value Decomposition (SVD) used in LSI is unscalable in terms of both memory consumption and computation time.

Document representation and indexing is a key problem for document analysis and processing, such as clustering, classification and retrieval. Conventionally, Latent Semantic Indexing (LSI) is considered effective in deriving such an indexing. LSI essentially detects the most representative features for document representation rather than the most discriminative features. Therefore, LSI might not be optimal in discriminating documents with different semantics. In this paper, a novel algorithm called Locality Preserving Indexing (LPI) is proposed for document indexing. Each document is represented by a vector with low dimensionality. In contrast to LSI which discovers the global structure of the document space, LPI discovers the local structure and obtains a compact document representation subspace that best detects the essential semantic structure. We compare the proposed LPI approach with LSI on two standard databases. Experimental results show that LPI provides better representation in the sense of semantic structure.

We 1 develop Conditional Random Sampling (CRS), a technique particularly suitable for sparse data. In large-scale applications, the data are often highly sparse. CRS combines sketching and sampling in that it converts sketches of the data into conditional random samples online in the estimation stage, with the sample size determined retrospectively. This paper focuses on approximating pairwise l2 and l1 distances and comparing CRS with random projections. For boolean (0/1) data, CRS is provably better than random projections. We show using real-world data that CRS often outperforms random projections. This technique can be applied in learning, data mining, information retrieval, and database query optimizations. 1

In this paper we show that dimensionality reduction (i.e., Johnson-Lindenstrauss lemma) preserves not only the distances between static points, but also between moving points, and more generally between low-dimensional flats, polynomial curves, curves with low winding degree, and polynomial surfaces. We also show that surfaces with bounded doubling dimension can be embedded into low dimension with small additive error. Finally, we show that for points with polynomial motion, the radius of the smallest enclosing ball can be preserved under dimensionality reduction. 1

The intersection of tree transducer-based translation models with n-gram language models results in huge dynamic programs for machine translation decoding. We propose a multipass, coarse-to-fine approach in which the language model complexity is incrementally introduced. In contrast to previous orderbased bigram-to-trigram approaches, we focus on encoding-based methods, which use a clustered encoding of the target language. Across various encoding schemes, and for multiple language pairs, we show speed-ups of up to 50 times over single-pass decoding while improving BLEU score. Moreover, our entire decoding cascade for trigram language models is faster than the corresponding bigram pass alone of a bigram-to-trigram decoder. 1

We develop dynamic dimensionality reduction based on the approximation of the standard inner-product. The innerproduct, by itself, is used as a distance measure in a wide area of applications such as document databases, e.g. latent semantic indexing (LSI). A first order approximation to the inner-product is usually obtained from the Cauchy-Schwarz inequality. The method proposed in this paper refines such an approximation by using higher order power symmetric functions of the components of the vectors, which are powers of the p-norms of the vectors for p = 1; 2; : : : ; m. We show how to compute fixed coefficients that work as universal weights based on the moments of the probability density function assumed for the distribution of the components of the input vectors in the data set. Our experiments on synthetic and document data show that with this technique, the similarity between two objects in high dimensional space for certain applications can be accurately approximated by a signi...

This paper investigates the use of concept-based representations for text categorization. We introduce