recognition-rl

1.基于rl的性格辨识： python run_sa.py v6-4-1 python run_sa_eval.py v6-4-1 2.基于rl without attention python run_sa.py v6-4-3 python run_sa_eval.py v6-4-5 3.基于监督学习的性格辨识 python run_sa.py v6-5-6 4.监督学习 without attention python run_sa.py v6-5-9 5.离线监督学习 python generate_supervise_data.py v6-5-6 python run_supervise_offline.py v6-5-6

core/method_supervise_offline里面的datasize参数定义训练样本大小 查看当前文件夹下文件个数(不包括文件夹) ls -l | grep "^-" | wc -l

evaluate

multi-agent, social behavior

• 第一步：选择交互行为

### bottleneck
python run_ma_eval.py v-3-1 --evaluate -d evaluate_ray_isac_adaptive_character__social_behavior__bottleneck

### intersection
python run_ma_eval.py v-3-2 --evaluate -d evaluate_ray_isac_adaptive_character__social_behavior__intersection

### merge
python run_ma_eval.py v-3-3 --evaluate -d evaluate_ray_isac_adaptive_character__social_behavior__merge

### roundabout
python run_ma_eval.py v-3-4 --evaluate -d evaluate_ray_isac_adaptive_character__social_behavior__roundabout

注：上述命令，加入 --render --invert可查看可视化，例如

python run_ma_eval.py v-3-1 --evaluate -d evaluate_ray_isac_adaptive_character__social_behavior__bottleneck --render --invert

• 第二步：可视化交互行为

### bottleneck
python eval_social_behavior.py v1

### intersection (todo)
python eval_social_behavior.py v2

### merge (todo)
python eval_social_behavior.py v3

### roundabout (todo)
python eval_social_behavior.py v4