Deep Learning with Discriminative Margin Loss for Cross-Domain Consumer-to-Shop Clothes Retrieval.

Consumer-to-shop clothes retrieval refers to the problem of matching photos taken by customers with their counterparts in the shop. Due to some problems, such as a large number of clothing categories, different appearances of clothing items due to different camera angles and shooting conditions, different background environments, and different body postures, the retrieval accuracy of traditional consumer-to-shop models is always low. With advances in convolutional neural networks (CNNs), the accuracy of garment retrieval has been significantly improved. Most approaches addressing this problem use single CNNs in conjunction with a softmax loss function to extract discriminative features. In the fashion domain, negative pairs can have small or large visual differences that make it difficult to minimize intraclass variance and maximize interclass variance with softmax. Margin-based softmax losses such as Additive Margin-Softmax (aka CosFace) improve the discriminative power of the original softmax loss, but since they consider the same margin for the positive and negative pairs, they are not suitable for cross-domain fashion search. In this work, we introduce the cross-domain discriminative margin loss (DML) to deal with the large variability of negative pairs in fashion. DML learns two different margins for positive and negative pairs such that the negative margin is larger than the positive margin, which provides stronger intraclass reduction for negative pairs. The experiments conducted on publicly available fashion datasets DARN and two benchmarks of the DeepFashion dataset-(1) Consumer-to-Shop Clothes Retrieval and (2) InShop Clothes Retrieval-confirm that the proposed loss function not only outperforms the existing loss functions but also achieves the best performance.

Potential Involvement of Impaired BKCa Channel Function in Sensory Defensiveness and Some Behavioral Disturbances Induced by Unfamiliar Environment in a Mouse Model of Fragile X Syndrome.

In fragile X syndrome (FXS), sensory hypersensitivity and impaired habituation is thought to result in attention overload and various behavioral abnormalities in reaction to the excessive and remanent salience of environment features that would normally be ignored. This phenomenon, termed sensory defensiveness, has been proposed as the potential cause of hyperactivity, hyperarousal, and negative reactions to changes in routine that are often deleterious for FXS patients. However, the lack of tools for manipulating sensory hypersensitivity has not allowed the experimental testing required to evaluate the relevance of this hypothesis. Recent work has shown that BMS-204352, a BKCa channel agonist, was efficient to reverse cortical hyperexcitability and related sensory hypersensitivity in the Fmr1-KO mouse model of FXS. In the present study, we report that exposing Fmr1-KO mice to novel or unfamiliar environments resulted in multiple behavioral perturbations, such as hyperactivity, impaired nest building and excessive grooming of the back. Reversing sensory hypersensitivity with the BKCa channel agonist BMS-204352 prevented these behavioral abnormalities in Fmr1-KO mice. These results are in support of the sensory defensiveness hypothesis, and confirm BKCa as a potentially relevant molecular target for the development of drug medication against FXS/ASD.

Energy demands of diverse spiking cells from the neocortex, hippocampus, and thalamus.

It has long been known that neurons in the brain are not physiologically homogeneous. In response to current stimulus, they can fire several distinct patterns of action potentials that are associated with different physiological classes ranging from regular-spiking cells, fast-spiking cells, intrinsically bursting cells, and low-threshold cells. In this work we show that the high degree of variability in firing characteristics of action potentials among these cells is accompanied with a significant variability in the energy demands required to restore the concentration gradients after an action potential. The values of the metabolic energy were calculated for a wide range of cell temperatures and stimulus intensities following two different approaches. The first one is based on the amount of Na(+) load crossing the membrane during a single action potential, while the second one focuses on the electrochemical energy functions deduced from the dynamics of the computational neuron models. The results show that the thalamocortical relay neuron is the most energy-efficient cell consuming between 7 and 18 nJ/cm(2) for each spike generated, while both the regular and fast spiking cells from somatosensory cortex and the intrinsically-bursting cell from a cat visual cortex are the least energy-efficient, and can consume up to 100 nJ/cm(2) per spike. The lowest values of these energy demands were achieved at higher temperatures and high external stimuli.

Metabolic efficiency with fast spiking in the squid axon.

Fundamentally, action potentials in the squid axon are consequence of the entrance of sodium ions during the depolarization of the rising phase of the spike mediated by the outflow of potassium ions during the hyperpolarization of the falling phase. Perfect metabolic efficiency with a minimum charge needed for the change in voltage during the action potential would confine sodium entry to the rising phase and potassium efflux to the falling phase. However, because sodium channels remain open to a significant extent during the falling phase, a certain overlap of inward and outward currents is observed. In this work we investigate the impact of ion overlap on the number of the adenosine triphosphate (ATP) molecules and energy cost required per action potential as a function of the temperature in a Hodgkin-Huxley model. Based on a recent approach to computing the energy cost of neuronal action potential generation not based on ion counting, we show that increased firing frequencies induced by higher temperatures imply more efficient use of sodium entry, and then a decrease in the metabolic energy cost required to restore the concentration gradients after an action potential. Also, we determine values of sodium conductance at which the hydrolysis efficiency presents a clear minimum.

Energy efficiency of information transmission by electrically coupled neurons.

The generation of spikes by neurons is energetically a costly process. This paper studies the consumption of energy and the information entropy in the signalling activity of a model neuron both when it is supposed isolated and when it is coupled to another neuron by an electrical synapse. The neuron has been modelled by a four-dimensional Hindmarsh-Rose type kinetic model for which an energy function has been deduced. For the isolated neuron values of energy consumption and information entropy at different signalling regimes have been computed. For two neurons coupled by a gap junction we have analyzed the roles of the membrane and synapse in the contribution of the energy that is required for their organized signalling. Computational results are provided for cases of identical and nonidentical neurons coupled by unidirectional and bidirectional gap junctions. One relevant result is that there are values of the coupling strength at which the organized signalling of two neurons induced by the gap junction takes place at relatively low values of energy consumption and the ratio of mutual information to energy consumption is relatively high. Therefore, communicating at these coupling values could be energetically the most efficient option.