Picture two engineers on the same team. Engineer A gets assigned a feature. She looks at the problem, considers a few options, and picks the simplest. A straightforward implementation, maybe 50 lines of code. Easy to read, easy to test, easy for the next person to pick up. It works. She ships it in a couple of days and moves on.
更多精彩内容，关注钛媒体微信号（ID：taimeiti），或者下载钛媒体App
。业内人士推荐体育直播作为进阶阅读
Most visual neuroimaging experiments require participants to fixate on a point in the center of the screen during stimulus presentation. This is important because large parts of the visual system are retinotopically organized (a distorted retina-reflecting map). If participants freely move their eyes, the same image stimulates different parts of this cortical map. Eye movements can introduce strong confounds: models may end up decoding eye position on the cortical map and not stimulus-related cortical patterns. There is debate about the strict necessity of fixation though. It also reduces natural viewing behavior and may suppress activity in higher-level visual areas. In general, reconstruction on free-viewing datasets should be approached with caution.
That gives me the math for the title of this post. Each test user had a playfield with ~2,200 characters, and each character contains 2 pixels. The game runs at 10 FPS. 2500 * 2200 * 2 * 10 is a little over 100 million! Maybe that’s not a fair measurement, but it’s the one I chose.