No abstract

Session-based recommender systems (SBRSs) have shown superior performance over conventional methods. However, they show limited scalability on large-scale industrial datasets since most models learn one embedding per item. This leads to a large memory requirement (of storing one vector per item) and poor performance on sparse sessions with cold-start or unpopular items. Using one public and one large industrial dataset, we experimentally show that state-of-the-art SBRSs have low performance on sparse sessions with sparse items. We propose M2TRec, a Metadata-aware Multitask Transformer model for session-based recommendations. Our proposed method learns a transformation function from item metadata to embeddings, and is thus, item-ID free (i.e., does not need to learn one embedding per item). It integrates item metadata to learn shared representations of diverse item attributes. During inference, new or unpopular items will be assigned identical representations for the attributes they share with items previously observed during training, and thus will have similar representations with those items, enabling recommendations of even cold-start and sparse items. Additionally, M2TRec is trained in a multi-task setting to predict the next item in the session along with its primary category and subcategories. Our multi-task strategy makes the model converge faster and significantly improves the overall performance. Experimental results show significant performance gains using our proposed approach on sparse items on the two datasets.

Fossil fuels play a significant role in the automobile industry, as well as marine vessel systems, with a lot of benefits like high-density and low-cost power supply, which is relatively easy to reserve, apply and carry. However, fossil fuel combustion produces several emissions such as CO 2 , which become greenhouse gas emissions and harmful to human health. One of the most favorable emission-free modern technologies is fuel cells, which can be applied to supply power to marine vessel propulsion systems. In such a most electrified ship, the whole of the shipboard grid can be regarded as a direct current (DC) stand-alone microgrid (MG) configuration with linear resistive and nonlinear constant power loads (CPLs). The main challenge in this new configuration of the shipboard MG is stabilizing the system's currents and voltages, subjected to stochastic disturbances arising from the effect of external winds and waves. Hence, the key objective of this research is to introduce a new modified robust adaptive stochastic backstepping controller, which is equipped with an artificial neural network, for stabilizing the current and voltage of the DC MG. The developed approach is robust against uncertainties and disturbances, has a systematic design procedure, and offers a low computational burden compared to the other complicated nonlinear controllers. In the end, a simulation is run for investigating the performance of the proposed controller over the state-of-the-art methods.

Here, an optimal linear parameter varying (LPV) controller is developed for non‐linear proton exchange membrane fuel cell (PEMFC) systems. Two performance criteria are involved in the optimal controller, including the membrane pressure and the power of the PEMFC. The former should be minimized; and, the latter should be maximized. Since the goal is to derive sufficient conditions in terms of linear matrix inequality (LMI) constraints, both performances are re‐formulated and augmented in a unique convex optimization problem. Further, the non‐linear PEMFC with the equilibrium profile dependent on time‐varying parameters, that is, current and temperature, is represented by an LPV model. Using the LMI constraints and the obtained LPV model, the gains of the controller are derived. Comparing with the state‐of‐the‐art methods, the proposed approach offers an optimal control approach with a systematic design method, in which both practical issues are treated, simultaneously. To show the advantages of the developed approach, six typical controllers, including the proposed and existing approaches, are considered and several numerical simulation results are conducted.

No abstract