Abstract:
A semiconductor device that updates a weight coefficient used for arithmetic operation by an artificial neural network is provided. Each of the first to third memory cells draws a current corresponding to data of its retention node and changes the data in accordance with the potentials of first and second wirings. When a weight coefficient and first and second reference data are held in the retention nodes of the first to third memory cells, the first circuit supplies, to a third wiring, a constant currents drawn by the second and third memory cells. When input data is input to the first wiring, a difference current between the constant current and a current drawn by the first memory cell is changed, and the second circuit outputs arithmetic result data corresponding to the change. The third circuit inputs update data corresponding to the arithmetic result data to the second wiring.
Abstract:
An object is to reduce standby power in a data processing device, without loss of convenience, in a structure in which a power supply control device includes the data processing device. In a structure of a power supply control device which supplies power to an external device using a main switch, a data processing device is provided in the power supply control device to control the main switch; a sub-switch supplying power to the data processing device is provided; and a volatile memory unit and a nonvolatile memory unit are provided in the data processing device. Further, the sub-switch is off in a period in which data is stored in the nonvolatile memory unit of the data processing device, so that power supply to the data processing device is intermittently stopped.
Abstract:
An imaging device capable of image processing is provided. The imaging device can retain analog data (image data) obtained by an image-capturing operation in a pixel and perform a product-sum operation of the analog data and a predetermined weight coefficient in the pixel to convert the data into binary data. When the binary data is taken in a neural network or the like, processing such as image recognition can be performed. Since enormous volumes of image data can be retained in pixels in the state of analog data, processing can be performed efficiently.
Abstract:
A display device with less power consumption. The display device operates in a still image display period and a moving image display period, and includes a receiving circuit having a function of receiving an image signal, a driver circuit having a function of driving a display portion, and the display portion having a function of displaying an image. The display portion displays an image at a frame frequency of less than or equal to 1 Hz in the still image display period. The receiving circuit stops receiving an image signal in the still image display period. Switching from the still image display period to the moving image display period is accurately performed by an external signal.
Abstract:
A semiconductor device and the like with low power consumption can be provided. A shift register in which a plurality of register circuits are connected to each other in series. The plurality of register circuits each include a flip-flop circuit. An operation of the flip-flop circuit of the register circuit in one stage is determined by a clock signal, an output signal of the register circuit in the previous stage, an output signal of the register circuit in the one stage, and an output signal of the register circuit in the next stage. Data stored in the flip-flop circuits in the register circuits in stages that are two or more stages before the one stage and in the register circuits in stages that are two or more stages after the one stage are not rewritten.
Abstract:
An imaging device capable of image processing is provided. The imaging device can retain analog data (image data) obtained by an image-capturing operation in a pixel and perform a product-sum operation of the analog data and a predetermined weight coefficient in the pixel to convert the data into binary data. When the binary data is taken in a neural network or the like, processing such as image recognition can be performed. Since enormous volumes of image data can be retained in pixels in the state of analog data, processing can be performed efficiently.
Abstract:
An imaging device capable of image processing is provided. The imaging device can retain analog data (image data) obtained by an image-capturing operation in a pixel and perform a product-sum operation of the analog data and a predetermined weight coefficient in the pixel to convert the data into binary data. When the binary data is taken in a neural network or the like, processing such as image recognition can be performed. Since enormous volumes of image data can be retained in pixels in the state of analog data, processing can be performed efficiently.
Abstract:
A hysteresis comparator that has a small circuit area and low power consumption is provided. A differential pair in the comparator is formed using transistors each including a back gate. The comparator is configured to apply an inverted signal of a logic value of an output signal of the comparator to the back gate of the transistor. That is, the threshold voltage of the transistor is controlled by the inverted signal. By the change of the threshold voltage, hysteresis can be added to an input comparison voltage.
Abstract:
A semiconductor device including an integrator circuit, in which electric discharge from a capacitor can be reduced to shorten time required for charging the capacitor in the case where supply of power supply voltage is stopped and restarted, and a method for driving the semiconductor device are provided. One embodiment has a structure in which a transistor with small off-state current is electrically connected in series to a capacitor in an integrator circuit. Further, in one embodiment of the present invention, a transistor with small off-state current is electrically connected in series to a capacitor in an integrator circuit; the transistor is on in a period during which power supply voltage is supplied; and the transistor is off in a period during which supply of the power supply voltage is stopped.
Abstract:
The memory circuit has a first writing mode in which data can be retained for a long time and a second writing mode in which data can be written at high speed. The memory circuit in which data reading is performed on the basis of a determined conductive state of a transistor includes first and second capacitor parts that are connected through a switch and retain electric charge corresponding to the data. The first writing mode is a mode where the switch is on and electric charge corresponding to the data is accumulated in the first and second capacitor parts that are electrically connected. The second writing mode is a mode where the switch is off, electric charge corresponding to the data is accumulated in the first capacitor part, and electric charge corresponding to the data is not accumulated in the second capacitor part.