Time Compression

Thousands of years ago, humans made gunpowder (did you discover?). 60 years ago, he discovered the power of atom. 25 years ago, computers were mainly tools of universities, large companies, and government agencies. Personal computers are now commonplace in the home, children in the suburbs have more than 100 times computer skills than their company's ancestors. The Internet was in the same state as it was ten years ago. As it developed very quickly, the original framework is now explosive at the seams and can not meet the needs of the emerging "internet community" anymore.

An article means compressing time and space into one sharpness. This reminds me of some of the plots in the video game Final Fantasy VIII, a devil, a witch named Ultimecia, attempting to enter a state called time compression. She is at the center of that moment. When I write a paper I think that I am a witch Ultimecia. I use text to blend different levels of time and space, check the current moment and recreate past scenes. Even if you hide the information in the article slightly, this information only makes sense in my own future version. These versions reread the same text in a few years, at the same time amazed and happy, confusing in my opinion. In a nutshell, articles always apply.

The auto encoder performs seemingly simple tasks. It accepts the input, compresses it, and rebuilds it. In most cases compression will pass the input through multiple layer parameters (or weights) and then decode it to its original form. This compression is very irreversible. This means that the greatest automatic encoder tends to be inferior to common compression algorithms such as JPEG and MP3 because a lot of information is lost during processing. The interesting thing about automatic encoders is that they learn the compression mechanism itself without teachers. Assuming you know how to format your inputs, the same network can learn to encode and decode images, audio, and even written text. Through back propagation, this is a slow and subtle process of changing network parameters to learn the distribution of data.