updateRewardProb: Update Reward Probabilities
In jdtrat/dynaq: Tools to Simulate DynaQ Reinforcement Learning Algorithms

Description Usage Arguments Details Value

Update reward probabilities according to Solway et al. (2019) methods. It takes in an image's reward probability, and returns a final probability to be reassigned to the inputted parameter according to the above description.where probabilities shift according to random gaussian noise with mean 0 and standard deviation of 0.025. Reflecting boundaries are at 0.8 and 0.2.

1	updateRewardProb(imageProb)

imageProb

The current probability of receiving a reward for a given image.

According to personal communication with Solway: Each of the four second-stage images had independent probabilities.

A random probability was chosen for each for trial 1 in the valid range.
The probability for trial n+1 was: prob trial n + N(0, sd)
There were reflecting boundaries, meaning, for example, if prob on trial n was 0.7 and you drew +0.3, you would end up at 0.6 for trial n+1 (go up 0.1 to the 0.8 upper boundary, then down 0.2).

Updated probability

jdtrat/dynaq documentation built on July 24, 2020, 7:18 a.m.