Abstract: A method may comprise, for each one of a plurality of destination users, computing a score using a first function based on a probability of a source user performing a source action directed towards the destination user, a second function based on a probability of the destination user performing a destination action in response to the source action, and a third function based on a measure of interaction by the destination user with an online service to result from the destination action being performed by the destination user. The score for inactive users may be boosted using an optimization algorithm with a first constraint comprising a maximum threshold number of the inactive users to display as recommendations to the source user and a second constraint comprising a minimum threshold number of the inactive users for which the source user to perform the source action.

Abstract: An online connection server is configured to more accurately predict connections for a viewing member of an online connection network. The online connection server may implement a machine-learning model that uses prior interactions by the viewing member to determine those connections that are likely to lead to more substantial interactions with the viewing member. The machine-learning model may be implemented using a reinforcement learning technique, such as a Deep Q network. The online connection server may further implement a state representation module that generates a state from a graph-based embedding of the viewing member profile, where the state is used to train the machine-learning model and determine an optimal candidate to recommend as a connection for the viewing member.

Abstract: An online connection server is configured to more accurately predict connections for a viewing member of an online connection network. The online connection server may implement a machine-learning model that uses prior interactions by the viewing member to determine those connections that are likely to lead to more substantial interactions with the viewing member. The machine-learning model may be implemented using a reinforcement learning technique, such as a Deep Q network. The online connection server may further implement a state representation module that generates a state from a graph-based embedding of the viewing member profile, where the state is used to train the machine-learning model and determine an optimal candidate to recommend as a connection for the viewing member.

Abstract: A stable Crystalline Form II of non-solvate of Apremilast (Formula I), methods of making Form II, pharmaceutical compositions comprising Form II, and their uses are disclosed. Also discloses are mixed crystals comprising Form Hand Form B and methods of making the same. The crystalline forms are characterized using X-ray powder diffractometry (XRPD), infrared spectroscopy (IR), differential scanning calorimetry (DSC), and thermal gravimetric analysis (TG). As compared with Forms A, B, C, D, E, F, and G reported in prior art references, Apremilast Form II of the present invention is more stable to temperature, light, and humidity, and is more suitable for long term storage; the crystallization solvents are safe and can be easily removed; the Form II has a white or off white appearance, and can be directly used in preparation processing; the preparation methods are simple and easy to reproduce, and are suitable for industrialized production.

Abstract: A stable Crystalline Form II of non-solvate of Apremilast (Formula I), methods of making Form II, pharmaceutical compositions comprising Form II, and their uses are disclosed. Also discloses are mixed crystals comprising Form Hand Form B and methods of making the same. The crystalline forms are characterized using X-ray powder diffractometry (XRPD), infrared spectroscopy (IR), differential scanning calorimetry (DSC), and thermal gravimetric analysis (TG). As compared with Forms A, B, C, D, E, F, and G reported in prior art references, Apremilast Form II of the present invention is more stable to temperature, light, and humidity, and is more suitable for long term storage; the crystallization solvents are safe and can be easily removed; the Form II has a white or off white appearance, and can be directly used in preparation processing; the preparation methods are simple and easy to reproduce, and are suitable for industrialized production.

Abstract: An apparatus for managing power in an electric vehicle includes a control circuit and a first switch. The control circuit is configured to generate a first control signal based on a current of a battery operable for powering the electric vehicle, and to generate a second control signal based on a voltage of the battery. The first switch is configured to control connection of the battery to a power source and a load in the electric vehicle according to the first control signal. The first control signal controls a voltage at a terminal of the first switch to maintain the current of the battery to be substantially equal to a current setting, and the second control signal controls the battery to switch between a first state and a second state.

Abstract: A power supply apparatus for delivering power to electrical devices in an electrical vehicle. The power supply apparatus is connected to a single power source and capable of delivering multiple levels of output voltage current. The power supply apparatus has two power supplies and a controller. The controller enables a primary power supply and disables a secondary power supply if the primary power supply is connected to the external power source. The controller enables the secondary power supply and disables the primary power supply if the primary power supply is not connected to the external power source.