2results about How to "Loss will not" patented technology

The invention provides a hardware self-locking power-on and power-off control system and method.The hardware self-locking power-on and power-off control system comprises a battery, a key switch circuit, a self-locking circuit and a load circuit, the key switch circuit comprises a power-on switch and a power-off switch, and the positive electrode of the battery is connected with the input end of the power-on switch and the input end of the power-off switch; the self-locking circuit is respectively connected with the load circuit, the output end of the power-on switch and the output end of the power-off switch, when the power-on switch is pressed down firstly and then loosened, the load circuit is controlled to be powered on and maintain a power-on state through the self-locking circuit, and when the power-off switch is pressed down, the load circuit is controlled to be powered off through the self-locking circuit. Hardware startup and shutdown control of the load circuit can be realized through the key switch circuit and the self-locking circuit, zero current after shutdown of the load circuit can also be realized, and the problems of static loss and software runaway risk in an existing startup and shutdown control mode can be relieved.

The invention relates to the technical field of deep learning and neural network model compression, and discloses a model filter compression method based on gradient guidance and terminal equipment in order to solve the technical problems of high reasoning cost and difficult deployment of an FSMN model caused by a high-order FIR filter. A trained FSMN model is obtained, the model comprises at least one high-order finite impulse response filter layer, the weight of the high-order finite impulse response filter layer is defined as the step b, and the gradient of a final loss function of the FSMN model relative to the gradient is determined and calculated; and step c, searching and determining an infinite impulse response filter of which the order is lower than that of the FIR filter based on the guidance of the gradient, and defining the weight of the infinite impulse response filter as step d, and replacing the generated compressed FSMN model with the infinite impulse response filter. The problem that the FSMN model is high in reasoning cost is solved, and the parameter quantity and the calculation complexity of the model are remarkably reduced.