trade

examples/trade/reward/__init__.py

@@ -0,0 +1,4 @@
+from .base import *
+from .pa_penalty import *
+from .ppo_reward import *
+from .vp_penalty import *

examples/trade/reward/base.py

@@ -0,0 +1,38 @@
+import numpy as np
+
+
+class Abs_Reward(object):
+    """The abstract class for Reward."""
+
+    def __init__(self, config):
+        return
+
+    def get_reward(self):
+        """:return: reward"""
+        reward = 0
+        return reward
+
+    def __call__(self, *args, **kargs):
+        return self.get_reward(*args, **kargs)
+
+    def isinstant(self):
+        """:return: Whether the reward should be given at every timestep or only at the end of this episode."""
+        raise NotImplementedError
+
+
+class Instant_Reward(Abs_Reward):
+    def __init__(self, config):
+        self.ffr_ratio = config["ffr_ratio"]
+        self.vvr_ratio = config["vvr_ratio"]
+
+    def isinstant(self):
+        return True
+
+
+class EndEpisode_Reward(Abs_Reward):
+    def __init__(self, config):
+        self.ffr_ratio = config["ffr_ratio"]
+        self.vvr_ratio = config["vvr_ratio"]
+
+    def isinstant(self):
+        return False

examples/trade/reward/pa_penalty.py

@@ -0,0 +1,14 @@
+import numpy as np
+from .base import Instant_Reward
+
+
+class PA_Penalty(Instant_Reward):
+    """Reward: (Abs(tt_ratio_t - 1) * 10000 * v_t / target - v_t^2 * penalty) / 100"""
+
+    def __init__(self, config):
+        self.penalty = config["penalty"]
+
+    def get_reward(self, performance_raise, v_t, target, PA_t, *args):
+        reward = PA_t * v_t / target
+        reward -= self.penalty * (v_t / target) ** 2
+        return reward / 100

examples/trade/reward/ppo_reward.py

@@ -0,0 +1,22 @@
+import numpy as np
+from .base import Abs_Reward
+
+
+class PPO_Reward(Abs_Reward):
+    """The reward function defined in IJCAI 2020"""
+
+    def __init__(self, *args):
+        pass
+
+    def isinstant(self):
+        return False
+
+    def get_reward(self, performace_raise, ffr, this_tt_ratio, is_buy):
+        if is_buy:
+            this_tt_ratio = 1 / this_tt_ratio
+        if this_tt_ratio < 1:
+            return -1.0
+        elif this_tt_ratio < 1.1:
+            return 0.0
+        else:
+            return 1.0

examples/trade/reward/vp_penalty.py

@@ -0,0 +1,41 @@
+import numpy as np
+from .base import Instant_Reward
+
+
+class VP_Penalty_small(Instant_Reward):
+    """Reward: (Abs(vv_ratio_t - 1) * 10000 - v_t^2 * penalty) / 100"""
+
+    def __init__(self, config):
+        self.penalty = config["penalty"]
+
+    def get_reward(self, performance_raise, v_t, target, *args):
+        """
+
+        :param performance_raise: Abs(vv_ratio_t - 1) * 10000.
+        :param target: Target volume
+        :param v_t: The traded volume
+        """
+        assert target > 0
+        reward = performance_raise * v_t / target
+        reward -= self.penalty * (v_t / target) ** 2
+        assert not (
+            np.isnan(reward) or np.isinf(reward)
+        ), f"{performance_raise}, {v_t}, {target}"
+        return reward / 100
+
+
+class VP_Penalty_small_vec(VP_Penalty_small):
+    def get_reward(self, performance_raise, v_t, target, *args):
+        """
+
+        :param performance_raise: Abs(vv_ratio_t - 1) * 10000.
+        :param target: Target volume
+        :param v_t: The traded volume
+        """
+        assert target > 0
+        reward = performance_raise * v_t.sum() / target
+        reward -= self.penalty * ((v_t / target) ** 2).sum()
+        assert not (
+            np.isnan(reward) or np.isinf(reward)
+        ), f"{performance_raise}, {v_t}, {target}"
+        return reward / 100