Table 1.
Parameter type | Parameter name | Value |
---|---|---|
SARSA-learning (see Appendix) | Learning rate α | 0.7 |
Discount factor γ | 0.7 | |
Environment/steps | Size | 150 cm × 150 cm |
Step size | 6 cm | |
Noise on the step size | ± 1.5 cm | |
Reward size | 15 cm × 15 cm | |
Place fieldsa | Number | 500 |
Width, through σ | 4.24 cm | |
Scaling factor A | 2.5 | |
Learning strategies | Exploration probability p e in E | 0.2 |
Probabilities for S | ||
p 1 | 0.5 | |
p 2 | 0.156 | |
p 3 | 0.063 | |
p 4 | 0.031 | |
p 5 | 0 | |
p 6 | 0.031 | |
p 7 | 0.063 | |
p 8 | 0.156 | |
Weighting factor w in S | 0.5 | |
Weight decay factor c f in F | 0.9995 | |
Zero weight threshold t f in F | 10 − 6 | |
Starting path length limitation in L | 200 | |
Path increase step in L, c l | 5 | |
Path limit in steps for any strategy | 300 |
aNote, additional justification for these default parameters is given in section ‘Place field size and density’