Selvolution

Issac / Junhui/Jiang
The term ‘Selvolution’ means ‘Let the object and system can make progress or solve problems by itself’. For machines it is mainly to design and analyze some of the computer can automatically “learn” algorithm. Machine learning algorithm is a kind of automatic analysis from the data to obtain the law, and use the law of the unknown data to predict the algorithm. For systems it is design to have a process that can get in automatically and fix the problem or inprove itself and repeat again. By applying this Selvolution tactic in design, it will make my design more friendly, smart and useful to users. For example, one object designed with Selvolution can adjust itself day by day. The more user use it, the more inprovements it can make. The tactic selvolution is mainly based on highedge technology like cellphones and personal computers. But it can also be used at a lot of other aspects. It’s more like a add-on part to the whole design and make it better. But it need further development, In order to best perform in the generalization process environment, the complexity of the hypothesis should match the complexity of the function under the data. If the complexity of the hypothesis is less than the function, then the model will not fit the data. If the complexity of the model is enhanced as a response, the errors in training will be reduced. But if the hypothesis is too complicated, then the model will be over-fitting, and the generalization will be worse.

Apple Face ID

The main part of Face ID which is related to Selvolution is that it can also recongnise you with daily changes. Put on glasses. Wear a hat. Grow a beard. Your friends might not recognise you. But Face ID will. Users do not need to change their face data when they get scratched beard, Face ID will change together with users by itself.

Deepmind AlphaGo Zero

AlphaGo Zero is the latest version of AlphaGo, as old versions like Lee and Master, they all get human data as a reference to study playing Go. But Zero dose not. First day running AlphaGo Zero it has no prior knowledge of the game and only the basic rules as an input. Third day is surpasses the abilities of AlphaGo Lee by itself. At 21st day, Zero reaches the level of AlphaGo Master, the version that defeated 60 top professionals online and world champion Ke JIe in 3 out of 3 games. These achivements are all done by itself and there is a possibility that Zero can make itself more skillful day by day.