The next movement of the rat comes from the calculation of a weight matrix. Combination of motivation to explore and aversion for open and high spaces. Highly based on Montgomery’s research in the 50’s.
- Anxiolytic drug -> induces exploration.
- Anxiogenic drug -> reduces exploration.
- Earlier research about the elevated plus-maze suggests the rat tends to explore places where has never been or it has a long ago.
- Exploration is a combination of “motivation to explore and aversion for open and high spaces.”
- Divides the elevated plus-maze into 21 different squares. On top of that, has a line of squares around the maze, summing to a total of 65 squares. They determine the focus of the rat – the rat can’t leave the maze, but could stay on an edge looking outside.
- The weight matrix is given by the willingness to go from square j to i minus aversion to go from square j to i.
- Since the rat tends to focus on exploring new places, the weight j to i descreases after making this move. There’s a decay constant β in the formula. Also, another motivation constant α, decaying when the rat starts to explore the walls.
- Probabilistic model for the next movement. The distribution is the following: 40% of chances to go forward, 20% to go to each of the sides, 10% to go backward, and 10% to stay on the same square.
- Tested with 12 rats, four for each of the scenarios: standard elevated plus-maze, totally closed, and totally open.
- Used Analysis of variance (ANOVA) and Bonferroni test to compare time in each of the arms between real and virtual rats.
- The sample size of real rats population is only 12 animals. It would be interesting to confirm if this sample is really representative and test with more subjects.
- As stated by the author, the model emulates the behavior without seeking to reproduce the brain structures responsible for them. I assume that increasing the sample size or adding more actions to the rat would raise awareness of where the model can be improved.
- Largely based on Montgomery’s publications and assumptions.
- Explains a lot about the weight matrix – a combination of assumptions for rat behavior – but it wasn’t very clear on how the learning happens. The ending section mentions a neural network: “It is based on a neural network whose architecture reproduces the spatial structure of the elevated plus-maze.”