Search

Constraint programming is known for being an efficient approach to solving combinatorial problems. Important design choices in a solver are the branching heuristics, designed to lead the search to the best solutions in a minimum amount of time. However, developing these heuristics is a time-consuming process that requires problem-specific expertise. This observation has motivated many efforts to use machine learning to automatically learn efficient heuristics without expert intervention. Although several generic variable-selection heuristics are available in the literature, the options for value-selection heuristics are more scarce. We propose to tackle this issue by introducing a generic learning procedure that can be used to obtain a value-selection heuristic inside a constraint programming solver. This has been achieved thanks to the combination of a deep Q-learning algorithm, a tailored reward signal, and a heterogeneous graph neural network. Experiments on graph coloring, maximum independent set, and maximum cut problems show that this framework competes with the well-known impact-based and activity-based search heuristics and can find solutions close to optimality without requiring a large number of backtracks.

Lamellar nanoporous gold thin films, constituted of a stack of very thin layers of porous gold, are synthesized by chemical etching from a stack of successively deposited nanolayers of copper and gold. The gold ligament size, the pore size and the distance between lamellas are tunable in the few tens nanometer range by controlling the initial thickness of the layers and the etching time. The SERS activity of these lamellar porous gold films is characterized by their SERS responses after adsorption of probe bipyridine and naphtalenethiol molecules. The SERS signal is investigated as a function of the bipyridine concentration from 10−14 mol L−1 to 10−3 mol L−1. The higher SERS response corresponds to an experimental detection limit down to 10−12 mol L−1. These performance is mainly attributed to the specific nanoporous gold architecture and the larger accessible surface to volume ratio. The lamellar nanoporous gold substrate is explored for sensitive SERS detection of dimethyl methylphosphonate (DMMP), a surrogate molecule of the highly toxic G-series nerve agents. The resultant nanostructure facilitates the diffusion of target molecules through the nanopores and their localization at the enhancing metallic surface leading to the unequivocal Raman signature of DMMP at a concentration of 5 parts per million.

Taps only consist of a small part of the manual input when inter-acting with touch-enabled surfaces. Indeed, how the hand behaves in the hovering space is informative of what the user intends to do. In this article, we present a data collection related to hand and eye motion. We tailored a kiosk-like system to record participants' gaze and hand movements. We specifically designed a memory game to detect the decision-making process users may face. Our data collection comprises of 177 trials from 71 participants. Based on a hand movement classification, we extracted 16588 hovers. We study the gaze behaviour during hovers, and we found out that the distance between gaze and hand depends on the target's location on the screen. We also showed how indecision can be deducted from this distance.

Prior work has shown gaze behaviour correlates with mouse movement in conventional desktop interfaces. As many devices now employ direct touch instead of a mouse, we have conducted a first study analysing how gaze correlates with touch input. We collected touch and gaze data on a tablet from 24 participants, on search, shopping, and ``link-following'' tasks representative of typical online use of tablets. Our analysis shows a gaze fixation typically leads touch input by 337 ms within 197 pixels. We observed a stronger correlation in the context of link-following, in comparison with touch on other HTML objects and virtual keyboard.

Zoë Zuilhof, Sandhaya Norris, Claude Blondeau, Pierre Tessier, Pierre Blier Department of Psychiatry, University of Ottawa, The Royal Ottawa Institute of Mental Health Research, Ottawa, ON, Canada Introduction: This study investigated if optimized dose regimens of escitalopram and bupropion combination from treatment initiation can be superior to either drug alone in speed of onset, remission rate, and maintenance of therapeutic efficacy.Methods: Patients from a single site (N=85) within a larger double-blind 12-week trial (N=245) showed a lower dropout rate (14% vs 40%) and used higher doses; therefore, this cohort was analyzed separately. Uniquely at this single site, after 12 weeks, non-remitters on a single drug received the other one in addition and combination non-remitters underwent a switch of escitalopram for duloxetine for a 6-week period. Escitalopram could be given up to 40 mg/day and bupropion up to 450 mg/day. A 6-month prolongation was then implemented in remitters, maintaining the double-blind design throughout. Remission was defined as ≤7 on the 17-item Hamilton Rating Scale for Depression, as in the initial publication.Results: At week 2, combination treatment was superior in remission rate (5/28) compared with both bupropion (0/26) and escitalopram monotherapies (0/31; P=0.03 and P=0.02, respectively). The week 12 remission rate of combination treatment showed a higher rate (15/28) relative to bupropion monotherapy (7/26; P=0.04), but not statistically different from escitalopram monotherapy (11/31; P=0.13). The 6-week augmentation produced remission in 7/21 monotherapy non-remitters and 0/6 in the switch group (P=0.13). Remission was sustained in 28/31 patients enrolled in the 6-month maintenance.Conclusion: These results suggest that combination of escitalopram and bupropion from treatment initiation is superior to either monotherapy in speed of onset. The addition of a second drug in non-remitters can lead to additional remissio