Joint Cross-correlation Research Articles

Deep discriminative image feature learning for cross-modal semantics understanding

Deep hashing image feature learning methods has attracted attentions of cross-modal semantics understanding researchers due to its low storage costs and efficient query speed. Usually heterogeneous cross-modal data are embedded in semantic space, then they may be converted to their own corresponding binary hash code through the learning hash function. But when mapping heterogeneous data to common Hamming space, some works ignore joint cross correlation that contributes to interactively explore the latent semantic information between different modalities, resulting in sub-optimal feature. To address the above issues, we present a novel deep discriminative feature learning methods for cross-modal semantics understanding named Deep Discriminant Semantic Joint Hashing(DDSJH). In order to maximize joint cross correlation, we use mutual information which contribute to semantics understanding. Features in semantic space are exchanged with the pairwise features to calculate loss between semantic space and Hamming space. Thus, the corresponding information in cross-modal data are collaboratively utilized to realize exploration of underlying mutual joint semantic correlation. Hash codes of similar categories are as close as possible, but it should be considered that hash codes of different categories data should be as discriminative as possible. So we harness linear classifier to learn discriminative hash code. Extensive experiments on two image–text cross-modal datasets show that our proposed approach achieves better accuracy than several state-of-the-art methods.

Joint cross-correlation analysis reveals complex, time-dependent functional relationship between cortical neurons and arm electromyograms.

Correlation between cortical activity and electromyographic (EMG) activity of limb muscles has long been a subject of neurophysiological studies, especially in terms of corticospinal connectivity. Interest in this issue has recently increased due to the development of brain-machine interfaces with output signals that mimic muscle force. For this study, three monkeys were implanted with multielectrode arrays in multiple cortical areas. One monkey performed self-timed touch pad presses, whereas the other two executed arm reaching movements. We analyzed the dynamic relationship between cortical neuronal activity and arm EMGs using a joint cross-correlation (JCC) analysis that evaluated trial-by-trial correlation as a function of time intervals within a trial. JCCs revealed transient correlations between the EMGs of multiple muscles and neural activity in motor, premotor and somatosensory cortical areas. Matching results were obtained using spike-triggered averages corrected by subtracting trial-shuffled data. Compared with spike-triggered averages, JCCs more readily revealed dynamic changes in cortico-EMG correlations. JCCs showed that correlation peaks often sharpened around movement times and broadened during delay intervals. Furthermore, JCC patterns were directionally selective for the arm-reaching task. We propose that such highly dynamic, task-dependent and distributed relationships between cortical activity and EMGs should be taken into consideration for future brain-machine interfaces that generate EMG-like signals.

Spectral indications of thermal Sunyaev-Zel'dovich effect in ARCHEOPS and WMAP data

In this paper, we present a joint cross-correlation analysis of the Archeops CMB maps at 143 and 217 GHz and the WMAP CMB maps at 41, 61 and 94 GHz with sky templates of projected galaxy density constructed from the 2MASS Extended Source catalog. These templates have been divided in patches sorted in decreasing galaxy density with a fixed number of pixels (we considered patches having 32, 64 and 128 pixels) and the cross correlation has been performed independently for each of these patches. We find that the densest patch shows a strong temperature decrement in the Q, V, W bands of WMAP and in the 143 GHz channel of Archeops, but not in the 217 GHz channel. Furthermore, we find that the spectral behavior of the amplitude of this decrement is compatible with that expected for the non-relativistic thermal Sunyaev-Zel'dovich effect, and is incompatible (at 4.5 sigma level) with the null hypothesis of having only CMB, noise and a dust component (nu^2) in those pixels. We find that the first 32-pixel sized patch samples the cores of 11 known massive nearby galaxy clusters. Under the assumption that the decrement found in this patch is due entirely to the thermal Sunyaev-Zel'dovich effect, we obtain an average comptonization parameter for those sources of y = (0.41 +- 0.08) x 10^(-4) at 13 arcmin angular scales. This value is compatible at 1 sigma with the expectation, y = 0.49 x 10^(-4), from a model of the cluster flux number counts based on the standard Lambda-CDM cosmology. The observed value of y is slightly diluted when computed for the first patch of 64 and 128 pixels, presumably due to the inclusion of less massive clusters, and the dilution factor inferred is also compatible with the quoted model.