Given e-commerce scenarios that user profiles are invisible, sessionbased recommendation is proposed to generate recommendation results from short sessions. Previous work only considers the user's sequential behavior in the current session, whereas the user's main purpose in the current session is not emphasized. In this paper, we propose a novel neural networks framework, i.e., Neural Attentive Recommendation Machine (NARM), to tackle this problem. Specifically, we explore a hybrid encoder with an attention mechanism to model the user's sequential behavior and capture the user's main purpose in the current session, which are combined as a unified session representation later. We then compute the recommendation scores for each candidate item with a bi-linear matching scheme based on this unified session representation. We train NARM by jointly learning the item and session representations as well as their matchings. We carried out extensive experiments on two benchmark datasets. Our experimental results show that NARM outperforms state-of-the-art baselines on both datasets. Furthermore, we also find that NARM achieves a significant improvement on long sessions, which demonstrates its advantages in modeling the user's sequential behavior and main purpose simultaneously.

This paper describes the theory of an integrative optical imaging system and its application to the analysis of the diffusion of 3-, 10-, 40-, and 70-kDa fluorescent dextran molecules in agarose gel and brain extracellular microenvironment. The method uses a precisely defined source of fluorescent molecules pressure ejected from a micropipette, and a detailed theory of the intensity contributions from out-of-focus molecules in a three-dimensional medium to a two-dimensional image. Dextrans tagged with either tetramethylrhodamine or Texas Red were ejected into 0.3% agarose gel or rat cortical slices maintained in a perfused chamber at 34 degrees C and imaged using a compound epifluorescent microscope with a 10 x water-immersion objective. About 20 images were taken at 2-10-s intervals, recorded with a cooled CCD camera, then transferred to a 486 PC for quantitative analysis. The diffusion coefficient in agarose gel, D, and the apparent diffusion coefficient, D*, in brain tissue were determined by fitting an integral expression relating the measured two-dimensional image intensity to the theoretical three-dimensional dextran concentration. The measurements in dilute agarose gel provided a reference value of D and validated the method. Values of the tortuosity, lambda = (D/D*)1/2, for the 3- and 10-kDa dextrans were 1.70 and 1.63, respectively, which were consistent with previous values derived from tetramethylammonium measurements in cortex. Tortuosities for the 40- and 70-kDa dextrans had significantly larger values of 2.16 and 2.25, respectively. This suggests that the extracellular space may have local constrictions that hinder the diffusion of molecules above a critical size that lies in the range of many neurotrophic compounds.

No abstract

Somatodendritic release of dopamine (DA) in midbrain is, at least in part, nonsynaptic; moreover, midbrain DA receptors are predominantly extrasynaptic. Thus somatodendritic DA mediates volume transmission, with an efficacy regulated by the diffusion and uptake characteristics of the local extracellular microenvironment. Here, we quantitatively evaluated diffusion and uptake in substantia nigra pars compacta (SNc) and reticulata (SNr), ventral tegmental area (VTA), and cerebral cortex in guinea pig brain slices. The geometric parameters that govern diffusion, extracellular volume fraction (alpha) and tortuosity (lambda), together with linear uptake (k'), were determined for tetramethylammonium (TMA(+)), and for DA, using point-source diffusion combined with ion-selective and carbon-fiber microelectrodes. TMA(+)-diffusion measurements revealed a large alpha of 30% in SNc, SNr, and VTA, which was significantly higher than the 22% in cortex. Values for lambda and k' for TMA(+) were similar among regions. Point-source DA-diffusion curves fitted theory well with linear uptake, with significantly higher values of k' for DA in SNc and VTA (0.08--0.09 s(-1)) than in SNr (0.006 s(-1)), where DA processes are sparser. Inhibition of DA uptake by GBR-12909 caused a greater decrease in k' in SNc than in VTA. In addition, DA uptake was slightly decreased by the norepinephrine transport inhibitor, desipramine in both regions, although this was statistically significant only in VTA. We used these data to model the radius of influence of DA in midbrain. Simulated release from a 20-vesicle point source produced DA concentrations sufficient for receptor activation up to 20 microm away with a DA half-life at this distance of several hundred milliseconds. Most importantly, this model showed that diffusion rather than uptake was the most important determinant of DA time course in midbrain, which contrasts strikingly with the striatum where uptake dominates. The issues considered here, while specific for DA in midbrain, illustrate fundamental biophysical properties relevant for all extracellular communication.