150results about How to "Achieve acceleration" patented technology

The invention provides a neural network compression and acceleration method, storage equipment and a terminal. The method comprises the steps that the original neural network is pruned; clustering quantification is carried out on a network weight of the pruned original network, and the original network after clustering quantification is trained to obtain a target neural network; a sparse matrix isused to store the target neural network; an input feature map is converted into an input matrix; and the sparse matrix is multiplied by the input matrix to obtain an output feature map correspondingto the input feature map. Thus, the computing resource and storage space of the neural network are reduced, and thus, the computing cost is reduced.

The invention discloses a depth convolution neural network acceleration and compression method based on parameter quantification, and the method comprises the steps: carrying out the quantification of parameters of a depth convolution neural network, and obtaining a plurality of sub-codebooks and indexes corresponding to the sub-codebooks; and obtaining an output characteristic graph of the depth convolution neural network according to the plurality of sub-codebooks and indexes corresponding to the sub-codebooks. The method can achieve the acceleration and compression of the depth convolution neural network.

The invention discloses a micro-accelerometer and a preparation method thereof, belonging to the technical field of processing micro inertia devices. The micro-accelerometer is arranged on a package substrate which is formed by stacking an upper surface plate, a lower surface plate and a plurality of middle substrates, wherein a signal detection circuit of the micro-accelerometer is attached on the upper surface plate; a sensitive element of the micro-accelerometer is embedded in the middle substrates, i.e. the middle substrates with cavities form a frame of the sensitive element; a flexible hanger and a sensitive mass block of the sensitive element are arranged in the cavities, and one end of the flexible hanger is connected with the sensitive element, while the other end is fixed on the frame; metal electrodes are respectively sputtered on the sensitive mass block and the surface of the frame corresponding to the sensitive mass block so as to form a flat plate sensitive capacitor, or a metal piezoresistive thick-film pattern is deposited at a connecting part of the flexible hanger and the inner side surface of the frame so as to form a metal piezoresistive strain gauge. The micro-accelerometer has high sensitivity and high-temperature resistance and can be fused into a whole with a system level package substrate, thus the micro-accelerometer has low processing difficulty and cost.

An embodiment of the invention discloses a congestion control method and a device thereof. The method comprises the steps of determining a current network bandwidth estimated value based on the data volume which are transmitted in a preset time length before the current time; determining whether network congestion occurs; if yes, based on the RTT value at the current time and the current network bandwidth estimated value, updating a slow starting threshold; using the updated slow starting threshold as the initial value of a congestion window size, adjusting the size of the congestion window according to a preset congestion window increase algorithm; and transmitting the data packet based on the adjusted congestion window size. According to the congestion control method and the device thereof, when network congestion occurs, a transmitting end adjusts the congestion window and the slow starting threshold based on the RTT value of the current time and the current network bandwidth estimated value. The current network bandwidth is sufficiently considered in adjusting the slow starting threshold and the congestion window, and furthermore the current network bandwidth resource is sufficiently used, thereby realizing TCP acceleration.

The invention provides an ARM (Advanced RISC Machine) platform based Android startup acceleration method. By using the technology, the Android startup procedure is accelerated, the BLCR (Berkeley Lab Checkpoint/Restart) technology is adopted to perform breakpoint settings on the more time-consuming pre-loaded class and pre-loaded resource procedure process of a Zygote process in the Android startup procedure, then the current status and information of the process are stored into the next image file when the process is started up to execute breakpoints, and the process is directly started up from the image file when being started up at the next time, so that the time-consuming pre-loaded class and pre-loaded resource process is skipped to achieve the purpose of accelerating startup. Through the Android acceleration realized by the BLCR technology, the startup time of Android is shortened to 13 seconds, the startup speed of Android is increased by nearly 30% to 40%, and the method has no effect on the subsequent performance of Android.

The invention provides a compression method based on layer-by-layer network binarization. The compression method based on layer-by-layer network binarization includes the steps: constructing a floating point type deep convolutional neural network; according to the opposite sequence of the hierarchy depth of the deep convolutional neural network, performing layer-by-layer binarization on the parameters in the network from deep to shallow until binarizing all the hierarchies in the deep convolutional neural network, and obtaining a binarized deep convolutional neural network; and performing pedestrian detection through the binarized deep convolutional neural network. Therefore, the compression method based on layer-by-layer network binarization realizes compression and acceleration of the network, and can effectively solve the problem of great precision loss caused by network quantification.

The invention discloses a method for accelerating RNA secondary structure prediction based on stochastic context-free grammar (SCFG), aiming at accelerating the speed of RNA secondary structure predication by using the SCFG. The method comprises the following steps of: firstly establishing a heterogeneous computer system comprising a host computer and a reconfigurable algorithm accelerator, then transmitting a formatted CM model and an encoded RNA sequence into the reconfigurable algorithm accelerator through the host computer, and executing a non-backtrace CYK/inside algorithm calculation by a PE array of the reconfigurable algorithm accelerator, wherein task division strategies of region-dependent segmentation and layered column-dependent parallel processing are adopted in the calculation so as to realize fine-grained parallel calculation, and n numbers of PEs simultaneously calculate n numbers of data positioned different columns of a matrix by adopting an SPMD way, but different calculation sequences are adopted in the calculation according to different state types. The invention realizes the application acceleration of the RNA sequence secondary structure prediction based on the SCFG model and has high acceleration ratio and low cost.

The invention discloses a multi-step typed ion jet device and a method, belonging to the fields of aerospace technology and microelectronic technique; the device comprises a first accelerating electrode, a second accelerating electrode, and an end hollow electrode; the first accelerating electrode and the second accelerating electrode are alternatively arranged so as to form a multi-step electrode structure; an insulation layer is arranged between every two adjacent first accelerating electrode and second accelerating electrode so as to lead the first accelerating electrode and the second accelerating electrode to be insulated with each other; the method utilizes the electric field convergence effect of polarized electrode array tip part, reinforces the electric field strength of region adjacent to the electrode, thus ionizing the gas molecular of adjacent area; the multi-step ion jet device structure is adopted and voltage is applied on the structure, thus forming ion jet. The device and the method can improve the ionization efficiency, reduce the working voltage, and work under the environment of multi-gas-plasma.

The present invention discloses a convolutional neural network-based image processing method and device and a mobile terminal. The method is suitable for being executed in the mobile terminal having a graphic program interface, wherein a convolutional neural network comprises a plurality of processing layers and a plurality of data layers, and the storage parameters of the data layers corresponding to the graphic program interface are stored in the mobile terminal and comprise the characteristic graph sizes, the texture transverse number and the texture longitudinal number. The method comprises the steps of taking a to-be-processed picture as the first data layer to input in the convolutional neural network; for each data layer, and according to the storage parameter, combining the plurality of characteristic graphs of the data layer into the corresponding large texture to store; for each processing layer, obtaining the large texture corresponding the previous data layer connected with the processing layer, calling the graphic program interface to carry out the rendering processing corresponding to the processing layer to generate the plurality of characteristic graphs of the next data layer.

The invention discloses a method to control automobile to drive in a cluster type formation. Automobile driving is controlled by using a basic driving unit-cluster, and a speed control model is provided on the basis of horizontal and vertical formation models. The method can be used to control the speed, acceleration and displacement via horizontal and vertical control, delay of time signals can be reduced, and is significant.

The invention discloses an accelerating method and system of an online game. The accelerating method includes intercepting a time function in online game application programming interface functions, and modifying a returned value of the time function according to the accelerating multiple input by a user. According to the accelerating method and system of the online game, the returned value of the time function is modified to cheat an application program so that speeds of an online game application program seem changed, accelerating of online games is achieved, and user experience is improved.

The invention discloses an environment-friendly unmanned aerial vehicle for constructional engineering. The environment-friendly unmanned aerial vehicle for constructional engineering comprises a vehicle body, a tail wing and vehicle wings and is characterized in that the vehicle wings are connected with the vehicle body through a folding mechanism; an undercarriage is arranged below the vehicle body and comprises an installation seat connected to the position below the vehicle body, a third rotary shaft arranged on the installation seat, a supporting column, a buffer spring, a cross beam, a supporting beam, two supporting legs and two connecting rods; the upper end of the supporting column penetrates the cross beam and is fixedly connected with the third rotary shaft; the lower end of the supporting column is inserted on the supporting beam; the supporting beam can move linearly along the supporting column; the supporting column is sleeved with the buffer spring; and the buffer spring is located between the cross beam and the supporting beam. By the adoption of the environment-friendly unmanned aerial vehicle, due to the fact that the undercarriage is arranged below the vehicle body, a buffering effect can be achieved in the taking-off and landing processes, and the vehicle body is protected against damage.

The invention belongs to the technical field of neural networks, and relates to a convolution kernel cutting method based on an entropy importance criterion model. In order to enable a convolutional neural network model with large parameter quantity, large calculated quantity and excellent performance to meet a real-time requirement in practical application. The invention provides a convolutionalkernel cutting method based on an entropy importance criterion model. The convolution kernel with less information amount is cut by taking a mode of solving image entropy of an activation channel of each convolution layer as a criterion for evaluating the importance of the corresponding convolution kernel. Therefore,, a small model which is excellent in performance and small in parameter amount and calculation amount is obtained, and the small model has the performance advantage and can meet the real-time performance and precision requirements of practical scene application.

The invention discloses auxiliary driving equipment based on voice control. The auxiliary driving equipment comprises an environment sensing module, a driver monitoring module, a voice processing module, a driving decision module, an execution module and a man-machine interaction module. The invention also discloses an auxiliary driving method based on voice control. The method is executed by theauxiliary driving equipment based on voice control, and comprises the following steps: waking up: waking up the auxiliary driving equipment according to a voice waking-up instruction of a driver; sensing an environment: sensing the driving state of the vehicle and the surrounding driving environment, and generating an environment sensing signal; monitoring a driver: monitoring the state of the driver and generating a driver monitoring signal; acquiring a voice instruction: acquiring and processing a sound signal in the vehicle and identifying the voice instruction of the driver from the soundsignal; making a driving decision: generating a driving instruction according to the environment sensing signal, the driver monitoring signal and the voice instruction; and executing: adjusting the driving state of the vehicle according to the driving instruction.

The invention relates to an electric automobile and a control method thereof. The control method of the electric automobile comprises the step of detecting the current position of a pedal, the step of calculating the position change state and movement speed of the pedal according to detection signals, the step of connecting a motor to an output loop of a first battery when the pedal position is changed into the speed regulating area from the braking area and a switch of the whole automobile is in a turn-on state, the step of cutting off the connection between the motor and the output loop of the first battery and connecting the motor to a storage loop of a second battery and starting a brake device when the pedal position is changed into the braking area from the speed regulating area, the step of cutting off the connection between the motor and the output loop of the first battery and connecting the motor to the storage loop of the second battery and starting the brake device when the movement speed of the pedal is larger than the preset value, and the step of adjusting the revolving direction and the revolving speed of the motor according to the gear information and the pedal position value when the pedal changes in the speed regulating area and the switch of the whole automobile is in a turn-on state.

The present invention relates to a dual-mass flywheel for a drive train of a motor vehicle, having one primary flywheel mass and one secondary flywheel mass (13), which are coupled to one another in a rotationally elastic manner via at least one spring device (19, 39, 53). At least one gear device (21, 31, 51, 51'') is disposed between the spring device and at least one of the two flywheel masses, said gear device causing a deflection movement of the spring device relative (19, 39, 53) to a rotational movement of the two flywheel masses in relation to one another, wherein a characteristic curve of the spring device may be modified by the gear device (21, 31, 51, 51'').

A working process of a content delivery network (CND) comprises that when a user visits a website which has joined in a CDN service, an optimum CDN nodal point which is closest to the user is determined by using a CDN re-direction technique and request of the user is enabled to point to the nodal joint. When a nodal point of the user arrives at the designated nodal point, a server (cache of the nodal point) of the CDN is responsible for providing content which the user requests for the user, namely that the user inputs the realm name of a website which the user visits on a browser of the user, the browser requests a local DNS to analyze the realm name, the local DSN sends the request to a main DNS of the website, the main DNS determines the optimum CDN nodal point at that moment by adopting a series of strategies and sends an analyzing result (internet protocol (IP) address) to the user and the user asks the designated CDN nodal point for the content of the corresponding website. The CDN system is capable of sending the content to the user in a high-efficiency way by judging user accessibility and server load. A good service can be provided for the user.

The invention relates to a pruning method for a deep convolutional neural network, computer equipment and an application method, and the method comprises the steps: (1), taking a layer as a unit, andexecuting a convolution kernel spectral clustering algorithm; (2) carrying out intra-layer convolution kernel pruning, dynamically calculating a spectral clustering pruning rate of this round, cuttingoff convolution kernels of non-clustering centers, and finely adjusting a certain number of rounds to recover precision; (3) calculating the importance of the convolution kernel by taking the globalsituation as a unit; (4) carrying out global convolution kernel pruning, calculating the global pruning rate of the current round dynamically, cutting off convolution kernels with low importance, andadjusting a certain round number finely to recover precision; (5) circularly executing the steps (1) to (4) until the target pruning rate is reached; and (6) keeping the weight of the descending convolution kernel as initialization, and continuing to finely adjust a certain number of rounds to obtain a final pruning network. According to the method, the problem that a large number of redundant parameters exist in an existing deep convolutional neural network is solved, and compression and acceleration of the network are achieved on a universal hardware platform.

An engine spray flow field near-field region optical diagnosis system comprises a light splitting system, a multiplexing system, an optical Kerr gate strobing imaging system, and a comparison imaging system, which are disposed along the incident direction of femtosecond pulse laser. The light splitting system split the incident femtosecond pulse laser into three beams of light pulses which are different in wavelength and polarization characteristic and adjustable in time interval. The three beams of light pulses respectively record information of a spray flow field near-field region at different time points through the multiplexing system and then go through the optical Kerr gate strobing imaging system for imaging respectively. The optical Kerr gate strobing trajectory light imaging system can be used to effectively remove the influence of diffusion light and increase the imaging signal-to-noise ratio. The particle speed and acceleration information can be calculated through a single particle track in three flow field graphs recorded by the comparison imaging system.

The invention discloses a three-dimensional object detection method based on Hash description and an iteration nearest point. Posture data of a target object is detected in three-dimensional information; According to the method, more abundant operation tasks and more replicated operation operations can be completed by operation tools such as a robot, structured production is converted into unstructured production, three-bit data is uniquely explained by adopting Hash description, matching of invalid data is reduced, and the feature matching speed is increased. And a point-to-tangent plane iteration nearest point method is adopted, so that the matching is more reliable, and the precision is higher.

A steering manipulator mechanism for car is composed of steering wheel, steering axle, steering sleeve pipe, steering mechanism and braking loop. Said steering wheel has a handle connected to braking bar and throttle wire. Turning the steering wheel left or right can steer the car. Pushing the steering wheel forward can brake the car. Pulling the steering wheel backward can accelerate the car.

The invention provides a resistance strain gauge, a sensing component, a force sensor and a skateboard. The resistance strain gauge comprises a strain gauge body and resistance wires, wherein the strain gauge body has a symmetric axis, the resistance wires are multiple in number and are divided into two groups, the two groups of resistance wires are electrically connected and are symmetrically arranged at two sides of the symmetric axis, and each of the resistance wires and the symmetric axis form an acute angle. The resistance strain gauge is advantaged in that a problem of influence on use experience of a user for an electric skateboard because of poor control performance of the electric skateboard in the prior is solved.

The invention relates to an SOPC (system on programmable chip) software and hardware cooperative system based on Linux, which comprises an SOPC module, a Linux system module and a network server module, wherein the Linux system module comprises a Linux kernel module and a Linux driver module, the network server module comprises a DHCP (dynamic host configuration protocol) server module and an NFS (network file system) server module, and the NFS server module comprises a file system mounting module and a file system interactive module. In the system of the invention, the SOPC and the NFS file system Linux are combined, and the software and hardware cooperative design system based on the combination is realized. The system of the invention has the advantages of fast development flow and simple implementation and is convenient in operation and easy in transplantation and expandability.

A method for actuating an automated transmission, in particular an automated transmission of a commercial vehicle, which is coupled to a drive motor, where data and/or signals are supplied to a controller of the automated transmission via an interface. When a defined event occurs, a temporally predictive non-steady profile of the drive motor torque for this event is transmitted to the controller of the automated transmission. The transmission controller ascertains and/or specifies a transmission shifting strategy on the basis of the transmitted data. There is also provided an apparatus for actuating an automated transmission, and also a test bench.