Open aforren1 opened 7 years ago
Maybe try a Gaussian centered on a 'typical' switch time, or do some mixture of an adaptive method + easy ones?
Gaussian centered on switch time would require some sort of calibration (tried shared Gaussian, very poor sampling of second person's speed/accuracy trade-off).
Maybe try a Gaussian centered on a 'typical' switch time, or do some mixture of an adaptive method + easy ones?