Nested dictionary learning for hierarchical organization of imagery and text

A tree-based dictionary learning model is developed for joint analysis of imagery and associated text. The dictionary learning may be applied directly to the imagery from patches, or to general feature vectors extracted from patches or superpixels (using any existing method for image feature extraction). Each image is associated with a path through the tree (from root to a leaf), and each of the multiple patches in a given image is associated with one node in that path. Nodes near the tree root are shared between multiple paths, representing image characteristics that are common among different types of images. Moving toward the leaves, nodes become specialized, representing details in image classes. If available, words (text) are also jointly modeled, with a path-dependent probability over words. The tree structure is inferred via a nested Dirichlet process, and a retrospective stick-breaking sampler is used to infer the tree depth and width.

Cross-Domain Multitask Learning with Latent Probit Models

Learning multiple tasks across heterogeneous domains is a challenging problem since the feature space may not be the same for different tasks. We assume the data in multiple tasks are generated from a latent common domain via sparse domain transforms and propose a latent probit model (LPM) to jointly learn the domain transforms, and the shared probit classifier in the common domain. To learn meaningful task relatedness and avoid over-fitting in classification, we introduce sparsity in the domain transforms matrices, as well as in the common classifier. We derive theoretical bounds for the estimation error of the classifier in terms of the sparsity of domain transforms. An expectation-maximization algorithm is derived for learning the LPM. The effectiveness of the approach is demonstrated on several real datasets.

Communications inspired linear discriminant analysis

We study the problem of supervised linear dimensionality reduction, taking an information-theoretic viewpoint. The linear projection matrix is designed by maximizing the mutual information between the projected signal and the class label. By harnessing a recent theoretical result on the gradient of mutual information, the above optimization problem can be solved directly using gradient descent, without requiring simplification of the objective function. Theoretical analysis and empirical comparison are made between the proposed method and two closely related methods, and comparisons are also made with a method in which Rényi entropy is used to define the mutual information (in this case the gradient may be computed simply, under a special parameter setting). Relative to these alternative approaches, the proposed method achieves promising results on real datasets. Copyright 2012 by the author(s)/owner(s).

Scoup-SMT: Scalable Coupled Sparse Matrix-Tensor Factorization

How can we correlate neural activity in the human brain as it responds to words, with behavioral data expressed as answers to questions about these same words? In short, we want to find latent variables, that explain both the brain activity, as well as the behavioral responses. We show that this is an instance of the Coupled Matrix-Tensor Factorization (CMTF) problem. We propose Scoup-SMT, a novel, fast, and parallel algorithm that solves the CMTF problem and produces a sparse latent low-rank subspace of the data. In our experiments, we find that Scoup-SMT is 50-100 times faster than a state-of-the-art algorithm for CMTF, along with a 5 fold increase in sparsity. Moreover, we extend Scoup-SMT to handle missing data without degradation of performance. We apply Scoup-SMT to BrainQ, a dataset consisting of a (nouns, brain voxels, human subjects) tensor and a (nouns, properties) matrix, with coupling along the nouns dimension. Scoup-SMT is able to find meaningful latent variables, as well as to predict brain activity with competitive accuracy. Finally, we demonstrate the generality of Scoup-SMT, by applying it on a Facebook dataset (users, friends, wall-postings); there, Scoup-SMT spots spammer-like anomalies.

En enkätundersökning om lärares syn på hur de arbetar med mål i grundskolan : A survey of teacher´s views on how they work with the goal in elementary school

Detta är en uppsats om lärares syn på målstyrningen vid två skolor samt hur de uppfattar att eleverna uppnår de uppsatta målen. Undersökningen utfördes som en kvalitativ enkätstudie. Resultatet visar att många lärare uppfattar att målen är otydliga vilket resulterar i att de tolkas olika. Angående hur man mäter att eleverna uppnår målen blev resultatet att det finns en mängd olika metoder att tillgå som till exempel loggbok och observation. En tydlig skillnad fanns i svaren bland dem som hade lång arbetslivserfarenhet gentemot den som hade minst arbetslivserfarenhet. Skillnaderna kan bland annat bottna i osäkerhet på sin yrkesroll samt bristande kunskaper i och med kort arbetslivserfarenhet.

A Bias in ML Estimates of Branch Lengths in the Presence of Multiple Signals

Sequence data often have competing signals that are detected by network programs or Lento plots. Such data can be formed by generating sequences on more than one tree, and combining the results, a mixture model. We report that with such mixture models, the estimates of edge (branch) lengths from maximum likelihood (ML) methods that assume a single tree are biased. Based on the observed number of competing signals in real data, such a bias of ML is expected to occur frequently. Because network methods can recover competing signals more accurately, there is a need for ML methods allowing a network. A fundamental problem is that mixture models can have more parameters than can be recovered from the data, so that some mixtures are not, in principle, identifiable. We recommend that network programs be incorporated into best practice analysis, along with ML and Bayesian trees.

Single-Symbol ML Decodable Distributed STBCs for Partially-Coherent Cooperative Networks

A relay network with N relays and a single source-destination pair is called a partially-coherent relay channel (PCRC) if the destination has perfect channel state information (CSI) of all the channels and the relays have only the phase information of the source-to-relay channels. In this paper, first, a new set of necessary and sufficient conditions for a space-time block code (STBC) to be single-symbol decodable (SSD) for colocated multiple antenna communication is obtained. Then, this is extended to a set of necessary and sufficient conditions for a distributed STBC (DSTBC) to be SSD for. a PCRC. Using this, several SSD DSTBCs for PCRC are identified. It is proved that even if a SSD STBC for a co-located MIMO channel does not satisfy the additional conditions for the code to be SSD for a PCRC, single-symbol decoding of it in a PCRC gives full-diversity and only coding gain is lost. It is shown that when a DSTBC is SSD for a PCRC, then arbitrary coordinate interleaving of the in-phase and quadrature-phase components of the variables does not disturb its SSD property for PCRC. Finally, it is shown that the possibility of channel phase compensation operation at the relay nodes using partial CSI at the relays increases the possible rate of SSD DSTBCs from (2)/(N) when the relays do not have CSI to(1)/(2), which is independent of N.

Infection-induced IL-10 and Jak-Stat: A review of the molecular circuitry controlling immune hyperactivity in response to pathogenic microbes

Generation of effective immune responses against pathogenic microbes depends on a fine balance between pro- and anti-inflammatory responses. Interleukin-10 (IL-10) is essential in regulating this balance and has garnered renewed interest recently as a modulator of the response to infection at the JAK-STAT signaling axis of host responses. Here, we examine how IL-10 functions as the “master regulator” of immune responses through JAK-STAT, and provide a perspective from recent insights on bacterial, protozoan, and viral infection model systems. Pattern recognition and subsequent molecular events that drive activation of IL-10-associated JAK-STAT circuitry are reviewed and the implications for microbial pathogenesis are discussed.

Nuclear Import Mechanisms of STAT and NF-kB Transcription Factors

The eukaryotic cell nucleoplasm is separated from the cytoplasm by the nuclear envelope. This compartmentation of eukaryotic cells requires that all nuclear proteins must be transported from the cytoplasm into the nucleus. Transport of macromolecules between the nucleus and the cytoplasm occurs through nuclear pore complexes (NPCs). Proteins to be targeted into the nucleus by the classical nuclear import system contain nuclear localization signals (NLSs), which are recognized by importin alpha, the NLS receptor. Importin alpha binds to importin beta, which docks the importin-cargo complex on the cytoplasmic side of the NPC and mediates the movement of the complex into the nucleus. Presently six human importin alpha isoforms have been identified. Transcription factors are among the most important regulators of gene expression in eukaryotic organisms. Transcription factors bind to specific DNA sequences on target genes and modulate the activity of the target gene. Many transcription factors, including signal transducers and activators of transcription (STAT) and nuclear factor kB (NF-kB), reside in the cytoplasm in an inactive form, and upon activation they are rapidly transported into the nucleus. In the nucleus STATs and NF-kB regulate the activity of genes whose products are critical in controlling numerous cellular and organismal processes, such as inflammatory and immune responses, cell growth, differentiation and survival. The aim of this study was to investigate the nuclear import mechanisms of STAT and NF-kB transcription factors. This work shows that STAT1 homodimers and STAT1/STAT2 heterodimers bind specifically and directly to importin alpha5 molecule via unconventional dimer-specific NLSs. Importin alpha molecules have two regions, which have been shown to directly interact with the amino acids in the NLS of the cargo molecule. The Arm repeats 2-4 comprise the N-terminal NLS binding site and Arm repeats 7-8 the C-terminal NLS binding site. In this work it is shown that the binding site for STAT1 homodimers and STAT1/STAT2 heterodimers is composed of Arm repeats 8 and 9 of importin alpha5 molecule. This work demonstrates that all NF-kB proteins are transported into the nucleus by importin alpha molecules. In addition, NLS was identified in RelB protein. The interactions between NF-kB proteins and importin alpha molecules were found to be directly mediated by the NLSs of NF-kB proteins. Moreover, we found that p50 binds to the N-terminal and p65 to the C-terminal NLS binding site of importin alpha3. The results from this thesis work identify previously uncharacterized mechanisms in nuclear import of STAT and NF-kB. These findings provide new insights into the molecular mechanisms regulating the signalling cascades of these important transcription factors from the cytoplasm into the nucleus to the target genes.

Low ML-Decoding Complexity, Large Coding Gain, Full-Rate, Full-Diversity STBCs for 2 x 2 and 4 x 2 MIMO Systems

This paper deals with low maximum-likelihood (ML)-decoding complexity, full-rate and full-diversity space-time block codes (STBCs), which also offer large coding gain, for the 2 transmit antenna, 2 receive antenna (2 x 2) and the 4 transmit antenna, 2 receive antenna (4 x 2) MIMO systems. Presently, the best known STBC for the 2 2 system is the Golden code and that for the 4 x 2 system is the DjABBA code. Following the approach by Biglieri, Hong, and Viterbo, a new STBC is presented in this paper for the 2 x 2 system. This code matches the Golden code in performance and ML-decoding complexity for square QAM constellations while it has lower ML-decoding complexity with the same performance for non-rectangular QAM constellations. This code is also shown to be information-lossless and diversity-multiplexing gain (DMG) tradeoff optimal. This design procedure is then extended to the 4 x 2 system and a code, which outperforms the DjABBA code for QAM constellations with lower ML-decoding complexity, is presented. So far, the Golden code has been reported to have an ML-decoding complexity of the order of for square QAM of size. In this paper, a scheme that reduces its ML-decoding complexity to M-2 root M is presented.

A Low-Complexity Near-ML Performance Achieving Algorithm for Large MIMO Detection

In this paper, we present a low-complexity, near maximum-likelihood (ML) performance achieving detector for large MIMO systems having tens of transmit and receive antennas. Such large MIMO systems are of interest because of the high spectral efficiencies possible in such systems. The proposed detection algorithm, termed as multistage likelihood-ascent search (M-LAS) algorithm, is rooted in Hopfield neural networks, and is shown to possess excellent performance as well as complexity attributes. In terms of performance, in a 64 x 64 V-BLAST system with 4-QAM, the proposed algorithm achieves an uncoded BER of 10(-3) at an SNR of just about 1 dB away from AWGN-only SISO performance given by Q(root SNR). In terms of coded BER, with a rate-3/4 turbo code at a spectral efficiency of 96 bps/Hz the algorithm performs close to within about 4.5 dB from theoretical capacity, which is remarkable in terms of both high spectral efficiency as well as nearness to theoretical capacity. Our simulation results show that the above performance is achieved with a complexity of just O(NtNt) per symbol, where N-t and N-tau denote the number of transmit and receive antennas.

Algebraic distributed space-time codes with low ML decoding complexity

"Extended Clifford algebras" are introduced as a means to obtain low ML decoding complexity space-time block codes. Using left regular matrix representations of two specific classes of extended Clifford algebras, two systematic algebraic constructions of full diversity Distributed Space-Time Codes (DSTCs) are provided for any power of two number of relays. The left regular matrix representation has been shown to naturally result in space-time codes meeting the additional constraints required for DSTCs. The DSTCs so constructed have the salient feature of reduced Maximum Likelihood (ML) decoding complexity. In particular, the ML decoding of these codes can be performed by applying the lattice decoder algorithm on a lattice of four times lesser dimension than what is required in general. Moreover these codes have a uniform distribution of power among the relays and in time, thus leading to a low Peak to Average Power Ratio at the relays.