10 results about "Activating function" patented technology

According to various embodiments, a wearable device comprises: a pair of ear cups for outputting sound, each of the pair of ear cups including an ear cover and an ear cushion; a rotating part coupled to at least one ear cup of the pair of ear cups, the rotating part including a support member disposed in the ear cover of the at least one ear cup and a rotating member coupled to be rotatable relative to the support member around a left-right axis; a rotating frame coupled to the rotating part so as to be rotatable relative to the ear cup around the left-right axis; at least one sensor for identifying the position of the rotating frame; and at least one processor, wherein, in a state in which the wearable device is worn on at least a part of the body of a user, rotation information related to the rotation of the rotating frame can be obtained using the at least one sensor, a change in a state of the rotating frame can be predicted on the basis of the rotation information, and activation or deactivation of at least one device connected to the at least one processor can be controlled on the basis of the prediction. Other various embodiments identified through the specification are possible.

Gaming systems and methods for determining which features, if any, to activate in association with one or more plays of a game based on a current value of a progressive award.

The invention provides an RFI classification method for pulsar data. The RFI classification method comprises the following steps: establishing a classification model MobileNetV2; a SENet attention mechanism module is added behind the 1 * 1 convolution layer of the dimension rise of the Bottleneck block; activating functions of a 3 * 3 deep convolutional layer and a dimension-reduced 1 * 1 convolutional layer of the Bottleneck block are changed into dynamic activating functions, the slope and the intercept of the dynamic activating functions are generated by an SENet attention mechanism module, and an RFI classification model is obtained; pulsar observation signals obtained in real time are processed into a visual spectrogram, the visual spectrogram is input into the RFI classification model, the RFI classification model is trained, and after training is completed, an RFI classification result of the pulsar observation signals is obtained. According to the RFI classification method disclosed by the invention, the feature learning ability of the model is enhanced by fusing an SENet attention mechanism through model adaptive learning, and the lightweight ability of the model is improved to the greatest extent by modifying an activation function.

A toy, in particular a toy doll, characterized in that the toy has a head section (1) and an actuating element (2), wherein the head section (1) can be attached to or is attached to a body section, wherein the actuating element (2) is arranged on the head section (1), preferably in a nose area of ​​the head section (1), wherein the actuating element (2) is connected to an AI module (3) via signal technology and is designed to transmit actuating signals to the AI ​​module (3), wherein the AI ​​module (3) is received in an interior area, in particular a cavity, of the body section and is designed to generate feedback signals based on input sound signals and / or the actuating signals transmitted by the actuating element (2).

Provided are a T-cell activating anti-CD3 antibody and a multispecific antibody comprising same, in particular a T-cell engager. Also provided are a composition comprising the antibody or multispecific antibody, and a therapeutic application thereof.

The invention relates to the technical field of pharmacology, in particular to application and method of gastrodin in preparation of an MT1 agonist with selectivity, and the application is as follows: the affinity of the gastrodin to an MT1 receptor exceeds 30 times that of the gastrodin to an MT2 receptor, and the MT1-mediated ERK1 / 2 activation function of the gastrodin is 3 times that of the gastrodin to the MT2; the gastrodin selectively activates MT1, inhibits generation of cAMP and activates a cAMP / ERK signal channel, so that the effects of promoting sleep, relieving pain, resisting inflammation, resisting tumors and the like are achieved, and the non-targeting effect related to MT2 is reduced. According to the invention, high affinity (Ki = 20.88 pM) of gastrodin to an MT1 receptor and low affinity to an MT2 receptor are determined, which indicates that gastrodin has more than 30 times of receptor selectivity. And the MT1 is selectively excited without influencing the MT2 receptor, so that the accurate regulation and control of an MT1 receptor mediated signal channel can be realized.

The invention relates to a channel clearing and collateral activating massage bed with a spine protection function, which comprises a bed frame, a bed surface, a sliding block, an elastic supporting plate, a traction rope, a movable rod, a self-rotating disc and a movable plate, and a switching assembly comprises a first rotating shaft rotationally connected with the edge of the self-rotating disc, a second rotating shaft movably matched with the self-rotating disc and a rotating plate. The threaded rod is manually rotated to drive the rectangular rod and the connecting plate to drive the rotating plate connected with the second rotating shaft to generate angle change, so that the first rotating shaft and the second rotating shaft at the eccentric point of the rotating disc are driven to get close to each other or get away from each other. When the first rotating shaft and the second rotating shaft get close to each other, the two sliding blocks move in the same direction in cooperation with the movable plate, and therefore movement of the position of the elastic supporting plate is achieved; when the first rotating shaft and the second rotating shaft are far away from each other, the two sliding blocks move close to each other or far away from each other, so that the arching radian of the elastic supporting plate is changed, and switching between a position adjusting mode and a radian adjusting mode is achieved.

A method comprises receiving an input signal for processing in one or more neurons of a neural network, wherein the neural network has zero bias neurons and includes a plurality of resistive processing unit (RPU) weights and each neuron has an activation function. The method also includes applying an arbitrary amplification factor to activation function outputs of the one or more neurons in the neural network, wherein the arbitrary amplification factor is based on a dynamic range of components in the neural network and compensates for conductance drift in values of the RPU weights. The method also includes performing a calculation with the neural network using the amplified activation function outputs of the one or more neurons.