Pete Warden, a member of the Google TensorFlow team and the head of TensorFlow Mobile, travels in the deep learning sea all year round.
In addition, these seemingly familiar books are also his works.
In addition, Pete has a new idea to share with everyone——
He firmly believes that the future of deep learning can run freely on tiny, low-power chips.
In other words, the single-chip microcomputer (MCU) will one day become the most fertile soil for deep learning.
The logic here is a bit detour, but it still seems to make sense.
Why is a single chip microcomputer
MCU everywhere
According to Peter's estimation, about 40 billion single-chip microcomputers (MCUs) will be sold worldwide this year.
There is a small CPU in the MCU, and the RAM is only a few kb, but it is used in medical equipment, automotive equipment, industrial equipment, and consumer electronic products.
Such a computer requires very little power and the price is very cheap, probably less than 50 cents.
The reason why it is not paid attention to is that under normal circumstances, MCUs are used to replace those old-fashioned electromechanical systems (such as in washing machines and remote controls)—the logic used to control the machine has not changed.
Energy consumption is the limiting factorAny equipment that requires Mains Electricity has great limitations. After all, you have to find a place to plug in wherever you go, even your mobile phones and PCs have to be charged frequently.
However, for smart products, it can be used anywhere without frequent maintenance.
So, let’s take a look at how fast the various parts of the smartphone use electricity-
· Display 400 mW
· Radio 800 mW
· Bluetooth 100mW
· Accelerometer 21 mW
· 130 mW gyroscope
· GPS 176 mW
In contrast, the MCU only needs 1 milliwatt, or less. However, a button battery has 2,000 joules of power, so even a 1 milliwatt device can only last for one month.
Of course, most of the current equipment uses Duty Cycling to prevent each component from being in working condition all the time. However, even so, the power distribution is still very tight.
CPU and sensors are less power-hungryThe power consumption of CPUs and sensors can basically be reduced to microwatts, such as Qualcomm's Glance vision chip.
In contrast, monitors and radios are particularly power-hungry. Even WiFi and Bluetooth are at least tens of milliwatts.
Because the energy required for data transmission seems to be proportional to the transmission distance. The CPU and sensor only transmit a few millimeters, while the transmission distance of the radio is much more expensive in meters.
This 10 -port USB 3.0 hub design uses 10 USB 3.0 port to extend to your computer, laptop, reader, network camera, moving HDD, Oculus, Oculus, printers and more USB devices.
This USB 3.0 fast data transmission driver The USB 3.0 center synchronization data transmission speed is as high as 5Gbps. Compatible with Windows XP / Vista / Win 7 / Win 8 / Win 10, Mac OS 10, Linux and above, compatible with USB 2.0 Connections.
10 Port Type-C HUB 3.0, USB HUB 3.0 10 Port, 10-Port USB Hub 3.0
shenzhen ns-idae technology co.,ltd , https://www.szbestchargers.com