Abstract: A method of performing automatic tuning on a deep learning model includes: utilizing an instruction-based learned cost model to estimate a first type of operational performance metrics based on a tuned configuration of layer fusion and tensor tiling; utilizing statistical data gathered during a compilation process of the deep learning model to determine a second type of operational performance metrics based on the tuned configuration of layer fusion and tensor tiling; performing an auto-tuning process to obtain a plurality of optimal configurations based on the first type of operational performance metrics and the second type of operational performance metrics; and configure the deep learning model according to one of the plurality of optimal configurations.

Abstract: A wearable electrocardiogram measurement device and a method for compressing an electrocardiogram are disclosed. A heart detection unit detects a heart activity status of a human body to obtain electrocardiogram data. A processor of a portable host proceeds with bandwidth decomposition and quantization by using quantization scales to obtain quantized results. An integer part of each quantized result is set as a quantized datum. A decimal portion of each quantized result is set as a truncated error datum. Each quantization scale is multiplied by a corresponding truncated error datum to obtain a product. If a distortion value obtained by adding the products fulfills a distortion default value range, the quantized data are encoded to obtain compressed electrocardiogram data for storage in a data storage unit of the host. If the distortion value does not fulfill the distortion default value range, an adjustment step is repeated.

Abstract: A wearable electrocardiogram measurement device and a method for compressing an electrocardiogram are disclosed. A heart detection unit detects a heart activity status of a human body to obtain electrocardiogram data. A processor of a portable host proceeds with bandwidth decomposition and quantization by using quantization scales to obtain quantized results. An integer part of each quantized result is set as a quantized datum. A decimal portion of each quantized result is set as a truncated error datum. Each quantization scale is multiplied by a corresponding truncated error datum to obtain a product. If a distortion value obtained by adding the products fulfills a distortion default value range, the quantized data are encoded to obtain compressed electrocardiogram data for storage in a data storage unit of the host. If the distortion value does not fulfill the distortion default value range, an adjustment step is repeated.

Abstract: An electronic device integrally formed with an automatic switching-off structure for protecting the device from overheating. At least one conducting lead of the electronic device is mechanically and electrically connected to its corresponding electrode by means of a solder column having desired melting point, with an appropriate spacing being kept between the conducting lead and the electrode. The coating layer of the electronic device is formed by a material such as epoxy resin which will soften and release a gaseous substance when heated to a temperature near its ignition point. When the temperature of the device exceeds the melting point of the solder, the solder column melts to break the electrical circuit and prevent overheating of the device. If the electronic device per se has higher working temperature, the coating layer is formed by a material having a higher ignition point such as silicon resin or silicon rubber which will not soften and expand when heated to a temperature near its ignition point.