diff --git a/algorithms/a2c/agent.py b/algorithms/a2c/agent.py
index 46a93e6b..0af29f94 100644
--- a/algorithms/a2c/agent.py
+++ b/algorithms/a2c/agent.py
@@ -16,12 +16,12 @@
 import torch.nn.functional as F
 import wandb
 
-from algorithms.common.abstract.agent import Agent as AbstractAgent
+from algorithms.common.abstract.agent import Agent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(AbstractAgent):
+class A2CAgent(Agent):
     """1-Step Advantage Actor-Critic interacting with environment.
 
     Attributes:
@@ -55,7 +55,7 @@ def __init__(
             optims (tuple): optimizers for actor and critic
 
         """
-        AbstractAgent.__init__(self, env, args)
+        Agent.__init__(self, env, args)
 
         self.actor, self.critic = models
         self.actor_optimizer, self.critic_optimizer = optims
@@ -158,7 +158,7 @@ def save_params(self, n_episode: int):
             "critic_optim_state_dict": self.critic_optimizer.state_dict(),
         }
 
-        AbstractAgent.save_params(self, params, n_episode)
+        Agent.save_params(self, params, n_episode)
 
     def write_log(self, i: int, score: int, policy_loss: float, value_loss: float):
         total_loss = policy_loss + value_loss
diff --git a/algorithms/bc/ddpg_agent.py b/algorithms/bc/ddpg_agent.py
index 7c3e69e8..a51cf3d8 100644
--- a/algorithms/bc/ddpg_agent.py
+++ b/algorithms/bc/ddpg_agent.py
@@ -16,20 +16,20 @@
 import torch.nn.functional as F
 import wandb
 
-from algorithms.common.abstract.her import HER as AbstractHER
+from algorithms.common.abstract.her import HER
 from algorithms.common.buffer.replay_buffer import ReplayBuffer
 import algorithms.common.helper_functions as common_utils
 from algorithms.common.noise import OUNoise
-from algorithms.ddpg.agent import Agent as DDPGAgent
+from algorithms.ddpg.agent import DDPGAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(DDPGAgent):
+class BCDDPGAgent(DDPGAgent):
     """BC with DDPG agent interacting with environment.
 
     Attributes:
-        HER (AbstractHER): hinsight experience replay
+        her (HER): hinsight experience replay
         transitions_epi (list): transitions per episode (for HER)
         desired_state (np.ndarray): desired state of current episode
         memory (ReplayBuffer): replay memory
@@ -47,14 +47,14 @@ def __init__(
         models: tuple,
         optims: tuple,
         noise: OUNoise,
-        HER: AbstractHER,
+        her: HER,
     ):
         """Initialization.
         Args:
-            HER (AbstractHER): hinsight experience replay
+            her (HER): hinsight experience replay
 
         """
-        self.HER = HER
+        self.her = her
         DDPGAgent.__init__(self, env, args, hyper_params, models, optims, noise)
 
     # pylint: disable=attribute-defined-outside-init
@@ -66,7 +66,6 @@ def _initialize(self):
 
         # HER
         if self.hyper_params["USE_HER"]:
-            self.her = self.HER()
             if self.hyper_params["DESIRED_STATES_FROM_DEMO"]:
                 self.her.fetch_desired_states_from_demo(demo)
 
diff --git a/algorithms/bc/sac_agent.py b/algorithms/bc/sac_agent.py
index 1d1d4e42..d77b6e04 100644
--- a/algorithms/bc/sac_agent.py
+++ b/algorithms/bc/sac_agent.py
@@ -18,19 +18,19 @@
 import torch.nn.functional as F
 import wandb
 
-from algorithms.common.abstract.her import HER as AbstractHER
+from algorithms.common.abstract.her import HER
 from algorithms.common.buffer.replay_buffer import ReplayBuffer
 import algorithms.common.helper_functions as common_utils
-from algorithms.sac.agent import Agent as SACAgent
+from algorithms.sac.agent import SACAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(SACAgent):
+class BCSACAgent(SACAgent):
     """BC with SAC agent interacting with environment.
 
     Attrtibutes:
-        HER (AbstractHER): hinsight experience replay
+        her (HER): hinsight experience replay
         transitions_epi (list): transitions per episode (for HER)
         desired_state (np.ndarray): desired state of current episode
         memory (ReplayBuffer): replay memory
@@ -48,14 +48,14 @@ def __init__(
         models: tuple,
         optims: tuple,
         target_entropy: float,
-        HER: AbstractHER,
+        her: HER,
     ):
         """Initialization.
         Args:
-            HER (AbstractHER): hinsight experience replay
+            her (HER): hinsight experience replay
 
         """
-        self.HER = HER
+        self.her = her
         SACAgent.__init__(self, env, args, hyper_params, models, optims, target_entropy)
 
     # pylint: disable=attribute-defined-outside-init
@@ -67,7 +67,6 @@ def _initialize(self):
 
         # HER
         if self.hyper_params["USE_HER"]:
-            self.her = self.HER()
             if self.hyper_params["DESIRED_STATES_FROM_DEMO"]:
                 self.her.fetch_desired_states_from_demo(demo)
 
diff --git a/algorithms/ddpg/agent.py b/algorithms/ddpg/agent.py
index 8b774c6a..e00eed3f 100644
--- a/algorithms/ddpg/agent.py
+++ b/algorithms/ddpg/agent.py
@@ -16,7 +16,7 @@
 import torch.nn.functional as F
 import wandb
 
-from algorithms.common.abstract.agent import Agent as AbstractAgent
+from algorithms.common.abstract.agent import Agent
 from algorithms.common.buffer.replay_buffer import ReplayBuffer
 import algorithms.common.helper_functions as common_utils
 from algorithms.common.noise import OUNoise
@@ -24,7 +24,7 @@
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(AbstractAgent):
+class DDPGAgent(Agent):
     """ActorCritic interacting with environment.
 
     Attributes:
@@ -64,7 +64,7 @@ def __init__(
             noise (OUNoise): random noise for exploration
 
         """
-        AbstractAgent.__init__(self, env, args)
+        Agent.__init__(self, env, args)
 
         self.actor, self.actor_target, self.critic, self.critic_target = models
         self.actor_optimizer, self.critic_optimizer = optims
@@ -196,7 +196,7 @@ def save_params(self, n_episode: int):
             "critic_optim_state_dict": self.critic_optimizer.state_dict(),
         }
 
-        AbstractAgent.save_params(self, params, n_episode)
+        Agent.save_params(self, params, n_episode)
 
     def write_log(self, i: int, loss: np.ndarray, score: int):
         """Write log about loss and score"""
diff --git a/algorithms/dqn/agent.py b/algorithms/dqn/agent.py
index b6775322..dd81c76b 100644
--- a/algorithms/dqn/agent.py
+++ b/algorithms/dqn/agent.py
@@ -22,7 +22,7 @@
 from torch.nn.utils import clip_grad_norm_
 import wandb
 
-from algorithms.common.abstract.agent import Agent as AbstractAgent
+from algorithms.common.abstract.agent import Agent
 from algorithms.common.buffer.priortized_replay_buffer import PrioritizedReplayBuffer
 from algorithms.common.buffer.replay_buffer import NStepTransitionBuffer
 import algorithms.common.helper_functions as common_utils
@@ -31,7 +31,7 @@
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(AbstractAgent):
+class DQNAgent(Agent):
     """DQN interacting with environment.
 
     Attribute:
@@ -69,7 +69,7 @@ def __init__(
             optim (torch.optim.Adam): optimizers for dqn
 
         """
-        AbstractAgent.__init__(self, env, args)
+        Agent.__init__(self, env, args)
 
         self.use_n_step = hyper_params["N_STEP"] > 1
         self.epsilon = hyper_params["MAX_EPSILON"]
@@ -261,7 +261,7 @@ def save_params(self, n_episode: int):
             "dqn_optim_state_dict": self.dqn_optimizer.state_dict(),
         }
 
-        AbstractAgent.save_params(self, params, n_episode)
+        Agent.save_params(self, params, n_episode)
 
     def write_log(self, i: int, loss: np.ndarray, score: float):
         """Write log about loss and score"""
diff --git a/algorithms/fd/ddpg_agent.py b/algorithms/fd/ddpg_agent.py
index 39e058ce..cd690dba 100644
--- a/algorithms/fd/ddpg_agent.py
+++ b/algorithms/fd/ddpg_agent.py
@@ -17,12 +17,12 @@
 from algorithms.common.buffer.priortized_replay_buffer import PrioritizedReplayBufferfD
 from algorithms.common.buffer.replay_buffer import NStepTransitionBuffer
 import algorithms.common.helper_functions as common_utils
-from algorithms.ddpg.agent import Agent as DDPGAgent
+from algorithms.ddpg.agent import DDPGAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(DDPGAgent):
+class DDPGfDAgent(DDPGAgent):
     """ActorCritic interacting with environment.
 
     Attributes:
diff --git a/algorithms/fd/dqn_agent.py b/algorithms/fd/dqn_agent.py
index dde5170d..af305621 100644
--- a/algorithms/fd/dqn_agent.py
+++ b/algorithms/fd/dqn_agent.py
@@ -22,12 +22,12 @@
 from algorithms.common.buffer.priortized_replay_buffer import PrioritizedReplayBufferfD
 from algorithms.common.buffer.replay_buffer import NStepTransitionBuffer
 import algorithms.common.helper_functions as common_utils
-from algorithms.dqn.agent import Agent as DQNAgent
+from algorithms.dqn.agent import DQNAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(DQNAgent):
+class DQNfDAgent(DQNAgent):
     """DQN interacting with environment.
 
     Attribute:
diff --git a/algorithms/fd/sac_agent.py b/algorithms/fd/sac_agent.py
index 347b289b..30178c76 100644
--- a/algorithms/fd/sac_agent.py
+++ b/algorithms/fd/sac_agent.py
@@ -18,12 +18,12 @@
 from algorithms.common.buffer.priortized_replay_buffer import PrioritizedReplayBufferfD
 from algorithms.common.buffer.replay_buffer import NStepTransitionBuffer
 import algorithms.common.helper_functions as common_utils
-from algorithms.sac.agent import Agent as SACAgent
+from algorithms.sac.agent import SACAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(SACAgent):
+class SACfDAgent(SACAgent):
     """SAC agent interacting with environment.
 
     Attrtibutes:
diff --git a/algorithms/per/ddpg_agent.py b/algorithms/per/ddpg_agent.py
index edf21754..3821fd6d 100644
--- a/algorithms/per/ddpg_agent.py
+++ b/algorithms/per/ddpg_agent.py
@@ -13,12 +13,12 @@
 
 from algorithms.common.buffer.priortized_replay_buffer import PrioritizedReplayBuffer
 import algorithms.common.helper_functions as common_utils
-from algorithms.ddpg.agent import Agent as DDPGAgent
+from algorithms.ddpg.agent import DDPGAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(DDPGAgent):
+class DDPGPERAgent(DDPGAgent):
     """ActorCritic interacting with environment.
 
     Attributes:
diff --git a/algorithms/ppo/agent.py b/algorithms/ppo/agent.py
index dfd7baf6..ddbe6393 100644
--- a/algorithms/ppo/agent.py
+++ b/algorithms/ppo/agent.py
@@ -16,14 +16,14 @@
 import torch.nn as nn
 import wandb
 
-from algorithms.common.abstract.agent import Agent as AbstractAgent
+from algorithms.common.abstract.agent import Agent
 from algorithms.common.env.multiprocessing_env import SubprocVecEnv
 import algorithms.ppo.utils as ppo_utils
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(AbstractAgent):
+class PPOAgent(Agent):
     """PPO Agent.
 
     Attributes:
@@ -64,7 +64,7 @@ def __init__(
             optims (tuple): optimizers for actor and critic
 
         """
-        AbstractAgent.__init__(self, env_single, args)
+        Agent.__init__(self, env_single, args)
 
         if not self.args.test:
             self.env = env_multi
@@ -251,7 +251,7 @@ def save_params(self, n_episode: int):
             "actor_optim_state_dict": self.actor_optimizer.state_dict(),
             "critic_optim_state_dict": self.critic_optimizer.state_dict(),
         }
-        AbstractAgent.save_params(self, params, n_episode)
+        Agent.save_params(self, params, n_episode)
 
     def write_log(
         self,
diff --git a/algorithms/sac/agent.py b/algorithms/sac/agent.py
index 8eb56ed1..0791ca03 100644
--- a/algorithms/sac/agent.py
+++ b/algorithms/sac/agent.py
@@ -18,14 +18,14 @@
 import torch.optim as optim
 import wandb
 
-from algorithms.common.abstract.agent import Agent as AbstractAgent
+from algorithms.common.abstract.agent import Agent
 from algorithms.common.buffer.replay_buffer import ReplayBuffer
 import algorithms.common.helper_functions as common_utils
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(AbstractAgent):
+class SACAgent(Agent):
     """SAC agent interacting with environment.
 
     Attrtibutes:
@@ -71,7 +71,7 @@ def __init__(
             target_entropy (float): target entropy for the inequality constraint
 
         """
-        AbstractAgent.__init__(self, env, args)
+        Agent.__init__(self, env, args)
 
         self.actor, self.vf, self.vf_target, self.qf_1, self.qf_2 = models
         self.actor_optimizer, self.vf_optimizer = optims[0:2]
@@ -277,7 +277,7 @@ def save_params(self, n_episode: int):
         if self.hyper_params["AUTO_ENTROPY_TUNING"]:
             params["alpha_optim"] = self.alpha_optimizer.state_dict()
 
-        AbstractAgent.save_params(self, params, n_episode)
+        Agent.save_params(self, params, n_episode)
 
     def write_log(
         self, i: int, loss: np.ndarray, score: float = 0.0, delayed_update: int = 1
diff --git a/algorithms/td3/agent.py b/algorithms/td3/agent.py
index b9f0d41c..63ecefc8 100644
--- a/algorithms/td3/agent.py
+++ b/algorithms/td3/agent.py
@@ -16,7 +16,7 @@
 import torch.nn.functional as F
 import wandb
 
-from algorithms.common.abstract.agent import Agent as AbstractAgent
+from algorithms.common.abstract.agent import Agent
 from algorithms.common.buffer.replay_buffer import ReplayBuffer
 import algorithms.common.helper_functions as common_utils
 from algorithms.common.noise import GaussianNoise
@@ -24,7 +24,7 @@
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(AbstractAgent):
+class TD3Agent(Agent):
     """ActorCritic interacting with environment.
 
     Attributes:
@@ -66,7 +66,7 @@ def __init__(
             noise (GaussianNoise): random noise for exploration
 
         """
-        AbstractAgent.__init__(self, env, args)
+        Agent.__init__(self, env, args)
 
         self.actor, self.actor_target = models[0:2]
         self.critic_1, self.critic_2 = models[2:4]
@@ -225,7 +225,7 @@ def save_params(self, n_episode: int):
             "critic_optim": self.critic_optimizer.state_dict(),
         }
 
-        AbstractAgent.save_params(self, params, n_episode)
+        Agent.save_params(self, params, n_episode)
 
     def write_log(
         self, i: int, loss: np.ndarray, score: float = 0.0, delayed_update: int = 1
diff --git a/examples/lunarlander_continuous_v2/a2c.py b/examples/lunarlander_continuous_v2/a2c.py
index cb963f62..7173eb60 100644
--- a/examples/lunarlander_continuous_v2/a2c.py
+++ b/examples/lunarlander_continuous_v2/a2c.py
@@ -11,7 +11,7 @@
 import torch
 import torch.optim as optim
 
-from algorithms.a2c.agent import Agent
+from algorithms.a2c.agent import A2CAgent
 from algorithms.common.networks.mlp import MLP, GaussianDist
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
@@ -65,7 +65,7 @@ def run(env: gym.Env, args: argparse.Namespace, state_dim: int, action_dim: int)
     optims = (actor_optim, critic_optim)
 
     # create an agent
-    agent = Agent(env, args, hyper_params, models, optims)
+    agent = A2CAgent(env, args, hyper_params, models, optims)
 
     # run
     if args.test:
diff --git a/examples/lunarlander_continuous_v2/bc-ddpg.py b/examples/lunarlander_continuous_v2/bc-ddpg.py
index 0c266dc0..1a6e7b88 100644
--- a/examples/lunarlander_continuous_v2/bc-ddpg.py
+++ b/examples/lunarlander_continuous_v2/bc-ddpg.py
@@ -13,7 +13,7 @@
 import torch
 import torch.optim as optim
 
-from algorithms.bc.ddpg_agent import Agent
+from algorithms.bc.ddpg_agent import BCDDPGAgent
 from algorithms.common.networks.mlp import MLP
 from algorithms.common.noise import OUNoise
 from examples.lunarlander_continuous_v2.utils import LunarLanderContinuousHER
@@ -114,10 +114,10 @@ def run(env: gym.Env, args: argparse.Namespace, state_dim: int, action_dim: int)
     optims = (actor_optim, critic_optim)
 
     # HER
-    HER = LunarLanderContinuousHER if hyper_params["USE_HER"] else None
+    her = LunarLanderContinuousHER() if hyper_params["USE_HER"] else None
 
     # create an agent
-    agent = Agent(env, args, hyper_params, models, optims, noise, HER)
+    agent = BCDDPGAgent(env, args, hyper_params, models, optims, noise, her)
 
     # run
     if args.test:
diff --git a/examples/lunarlander_continuous_v2/bc-sac.py b/examples/lunarlander_continuous_v2/bc-sac.py
index a2fbac7b..e3258b58 100644
--- a/examples/lunarlander_continuous_v2/bc-sac.py
+++ b/examples/lunarlander_continuous_v2/bc-sac.py
@@ -12,7 +12,7 @@
 import torch
 import torch.optim as optim
 
-from algorithms.bc.sac_agent import Agent
+from algorithms.bc.sac_agent import BCSACAgent
 from algorithms.common.networks.mlp import MLP, FlattenMLP, TanhGaussianDistParams
 from examples.lunarlander_continuous_v2.utils import LunarLanderContinuousHER
 
@@ -117,10 +117,10 @@ def run(env: gym.Env, args: argparse.Namespace, state_dim: int, action_dim: int)
     optims = (actor_optim, vf_optim, qf_1_optim, qf_2_optim)
 
     # HER
-    HER = LunarLanderContinuousHER if hyper_params["USE_HER"] else None
+    her = LunarLanderContinuousHER() if hyper_params["USE_HER"] else None
 
     # create an agent
-    agent = Agent(env, args, hyper_params, models, optims, target_entropy, HER)
+    agent = BCSACAgent(env, args, hyper_params, models, optims, target_entropy, her)
 
     # run
     if args.test:
diff --git a/examples/lunarlander_continuous_v2/ddpg.py b/examples/lunarlander_continuous_v2/ddpg.py
index 5da373f4..0db10528 100644
--- a/examples/lunarlander_continuous_v2/ddpg.py
+++ b/examples/lunarlander_continuous_v2/ddpg.py
@@ -13,7 +13,7 @@
 
 from algorithms.common.networks.mlp import MLP
 from algorithms.common.noise import OUNoise
-from algorithms.ddpg.agent import Agent
+from algorithms.ddpg.agent import DDPGAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
@@ -101,7 +101,7 @@ def run(env: gym.Env, args: argparse.Namespace, state_dim: int, action_dim: int)
     optims = (actor_optim, critic_optim)
 
     # create an agent
-    agent = Agent(env, args, hyper_params, models, optims, noise)
+    agent = DDPGAgent(env, args, hyper_params, models, optims, noise)
 
     # run
     if args.test:
diff --git a/examples/lunarlander_continuous_v2/ddpgfd.py b/examples/lunarlander_continuous_v2/ddpgfd.py
index c4f6c4b3..d5a98f04 100644
--- a/examples/lunarlander_continuous_v2/ddpgfd.py
+++ b/examples/lunarlander_continuous_v2/ddpgfd.py
@@ -13,7 +13,7 @@
 
 from algorithms.common.networks.mlp import MLP
 from algorithms.common.noise import OUNoise
-from algorithms.fd.ddpg_agent import Agent
+from algorithms.fd.ddpg_agent import DDPGfDAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
@@ -109,7 +109,7 @@ def run(env: gym.Env, args: argparse.Namespace, state_dim: int, action_dim: int)
     optims = (actor_optim, critic_optim)
 
     # create an agent
-    agent = Agent(env, args, hyper_params, models, optims, noise)
+    agent = DDPGfDAgent(env, args, hyper_params, models, optims, noise)
 
     # run
     if args.test:
diff --git a/examples/lunarlander_continuous_v2/per-ddpg.py b/examples/lunarlander_continuous_v2/per-ddpg.py
index fc038c65..33769669 100644
--- a/examples/lunarlander_continuous_v2/per-ddpg.py
+++ b/examples/lunarlander_continuous_v2/per-ddpg.py
@@ -13,7 +13,7 @@
 
 from algorithms.common.networks.mlp import MLP
 from algorithms.common.noise import OUNoise
-from algorithms.per.ddpg_agent import Agent
+from algorithms.per.ddpg_agent import DDPGPERAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
@@ -104,7 +104,7 @@ def run(env: gym.Env, args: argparse.Namespace, state_dim: int, action_dim: int)
     optims = (actor_optim, critic_optim)
 
     # create an agent
-    agent = Agent(env, args, hyper_params, models, optims, noise)
+    agent = DDPGPERAgent(env, args, hyper_params, models, optims, noise)
 
     # run
     if args.test:
diff --git a/examples/lunarlander_continuous_v2/ppo.py b/examples/lunarlander_continuous_v2/ppo.py
index 26b4aac4..dafea19e 100644
--- a/examples/lunarlander_continuous_v2/ppo.py
+++ b/examples/lunarlander_continuous_v2/ppo.py
@@ -14,7 +14,7 @@
 
 from algorithms.common.env.utils import env_generator, make_envs
 from algorithms.common.networks.mlp import MLP, GaussianDist
-from algorithms.ppo.agent import Agent
+from algorithms.ppo.agent import PPOAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
@@ -92,7 +92,7 @@ def run(env: gym.Env, args: argparse.Namespace, state_dim: int, action_dim: int)
     optims = (actor_optim, critic_optim)
 
     # create an agent
-    agent = Agent(env_single, env_multi, args, hyper_params, models, optims)
+    agent = PPOAgent(env_single, env_multi, args, hyper_params, models, optims)
 
     # run
     if args.test:
diff --git a/examples/lunarlander_continuous_v2/sac.py b/examples/lunarlander_continuous_v2/sac.py
index ae5a4cfa..344d6c9a 100644
--- a/examples/lunarlander_continuous_v2/sac.py
+++ b/examples/lunarlander_continuous_v2/sac.py
@@ -13,7 +13,7 @@
 import torch.optim as optim
 
 from algorithms.common.networks.mlp import MLP, FlattenMLP, TanhGaussianDistParams
-from algorithms.sac.agent import Agent
+from algorithms.sac.agent import SACAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
@@ -106,7 +106,7 @@ def run(env: gym.Env, args: argparse.Namespace, state_dim: int, action_dim: int)
     optims = (actor_optim, vf_optim, qf_1_optim, qf_2_optim)
 
     # create an agent
-    agent = Agent(env, args, hyper_params, models, optims, target_entropy)
+    agent = SACAgent(env, args, hyper_params, models, optims, target_entropy)
 
     # run
     if args.test:
diff --git a/examples/lunarlander_continuous_v2/sacfd.py b/examples/lunarlander_continuous_v2/sacfd.py
index 82fdbe4a..665c9407 100644
--- a/examples/lunarlander_continuous_v2/sacfd.py
+++ b/examples/lunarlander_continuous_v2/sacfd.py
@@ -13,7 +13,7 @@
 import torch.optim as optim
 
 from algorithms.common.networks.mlp import MLP, FlattenMLP, TanhGaussianDistParams
-from algorithms.fd.sac_agent import Agent
+from algorithms.fd.sac_agent import SACfDAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
@@ -112,7 +112,7 @@ def run(env: gym.Env, args: argparse.Namespace, state_dim: int, action_dim: int)
     optims = (actor_optim, vf_optim, qf_1_optim, qf_2_optim)
 
     # create an agent
-    agent = Agent(env, args, hyper_params, models, optims, target_entropy)
+    agent = SACfDAgent(env, args, hyper_params, models, optims, target_entropy)
 
     # run
     if args.test:
diff --git a/examples/lunarlander_continuous_v2/td3.py b/examples/lunarlander_continuous_v2/td3.py
index 9015c6bf..b73a05c5 100644
--- a/examples/lunarlander_continuous_v2/td3.py
+++ b/examples/lunarlander_continuous_v2/td3.py
@@ -13,7 +13,7 @@
 
 from algorithms.common.networks.mlp import MLP
 from algorithms.common.noise import GaussianNoise
-from algorithms.td3.agent import Agent
+from algorithms.td3.agent import TD3Agent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
@@ -115,7 +115,7 @@ def run(env: gym.Env, args: argparse.Namespace, state_dim: int, action_dim: int)
     optims = (actor_optim, critic_optim)
 
     # create an agent
-    agent = Agent(env, args, hyper_params, models, optims, noise)
+    agent = TD3Agent(env, args, hyper_params, models, optims, noise)
 
     # run
     if args.test:
diff --git a/examples/lunarlander_continuous_v2/utils.py b/examples/lunarlander_continuous_v2/utils.py
index 0ab7fddd..7e8a5007 100644
--- a/examples/lunarlander_continuous_v2/utils.py
+++ b/examples/lunarlander_continuous_v2/utils.py
@@ -7,7 +7,7 @@
 
 import numpy as np
 
-from algorithms.common.abstract.her import HER as AbstractHER
+from algorithms.common.abstract.her import HER
 from algorithms.common.abstract.reward_fn import RewardFn
 
 
@@ -22,7 +22,7 @@ def __call__(self, transition: tuple, goal_state: np.ndarray) -> np.float64:
             return np.float64(-1.0)
 
 
-class LunarLanderContinuousHER(AbstractHER):
+class LunarLanderContinuousHER(HER):
     """HER for LunarLanderContinuous-v2 environment.
 
     Attributes:
@@ -33,7 +33,7 @@ class LunarLanderContinuousHER(AbstractHER):
 
     def __init__(self, reward_func: RewardFn = L1DistanceRewardFn):
         """Initialization."""
-        AbstractHER.__init__(self, reward_func=reward_func)
+        HER.__init__(self, reward_func=reward_func)
 
     # pylint: disable=attribute-defined-outside-init
     def fetch_desired_states_from_demo(self, demo: list):
diff --git a/examples/lunarlander_v2/dqfd.py b/examples/lunarlander_v2/dqfd.py
index 2c9155dc..3b373c96 100644
--- a/examples/lunarlander_v2/dqfd.py
+++ b/examples/lunarlander_v2/dqfd.py
@@ -12,7 +12,7 @@
 import torch.optim as optim
 
 from algorithms.dqn.networks import C51DuelingMLP
-from algorithms.fd.dqn_agent import Agent
+from algorithms.fd.dqn_agent import DQNfDAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
@@ -89,7 +89,7 @@ def get_fc_model():
     models = (dqn, dqn_target)
 
     # create an agent
-    agent = Agent(env, args, hyper_params, models, dqn_optim)
+    agent = DQNfDAgent(env, args, hyper_params, models, dqn_optim)
 
     # run
     if args.test:
diff --git a/examples/lunarlander_v2/dqn.py b/examples/lunarlander_v2/dqn.py
index 54a24de7..c824dfaf 100644
--- a/examples/lunarlander_v2/dqn.py
+++ b/examples/lunarlander_v2/dqn.py
@@ -11,7 +11,7 @@
 import torch
 import torch.optim as optim
 
-from algorithms.dqn.agent import Agent
+from algorithms.dqn.agent import DQNAgent
 from algorithms.dqn.networks import C51DuelingMLP
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
@@ -85,7 +85,7 @@ def get_fc_model():
     models = (dqn, dqn_target)
 
     # create an agent
-    agent = Agent(env, args, hyper_params, models, dqn_optim)
+    agent = DQNAgent(env, args, hyper_params, models, dqn_optim)
 
     # run
     if args.test:
diff --git a/examples/pong_no_frameskip_v4/dqn.py b/examples/pong_no_frameskip_v4/dqn.py
index 08d81da0..074b987e 100644
--- a/examples/pong_no_frameskip_v4/dqn.py
+++ b/examples/pong_no_frameskip_v4/dqn.py
@@ -12,7 +12,7 @@
 import torch.optim as optim
 
 from algorithms.common.networks.cnn import CNNLayer
-from algorithms.dqn.agent import Agent
+from algorithms.dqn.agent import DQNAgent
 from algorithms.dqn.networks import IQNCNN, IQNMLP
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
@@ -101,7 +101,7 @@ def get_cnn_model():
     models = (dqn, dqn_target)
 
     # create an agent
-    agent = Agent(env, args, hyper_params, models, dqn_optim)
+    agent = DQNAgent(env, args, hyper_params, models, dqn_optim)
     agent.env_name = env_name
 
     # run
diff --git a/examples/reacher_v2/bc-ddpg.py b/examples/reacher_v2/bc-ddpg.py
index 1a8a78aa..1e11e88a 100644
--- a/examples/reacher_v2/bc-ddpg.py
+++ b/examples/reacher_v2/bc-ddpg.py
@@ -12,7 +12,7 @@
 import torch
 import torch.optim as optim
 
-from algorithms.bc.ddpg_agent import Agent
+from algorithms.bc.ddpg_agent import BCDDPGAgent
 from algorithms.common.networks.mlp import MLP
 from algorithms.common.noise import OUNoise
 from examples.reacher_v2.utils import ReacherHER
@@ -36,7 +36,7 @@
     "INITIAL_RANDOM_ACTION": 10000,
     "MULTIPLE_LEARN": 1,
     # HER
-    "USE_HER": True,
+    "USE_HER": False,
     "SUCCESS_SCORE": -5.0,
     "DESIRED_STATES_FROM_DEMO": False,
 }
@@ -110,10 +110,10 @@ def run(env: gym.Env, args: argparse.Namespace, state_dim: int, action_dim: int)
     optims = (actor_optim, critic_optim)
 
     # HER
-    HER = ReacherHER if hyper_params["USE_HER"] else None
+    her = ReacherHER() if hyper_params["USE_HER"] else None
 
     # create an agent
-    agent = Agent(env, args, hyper_params, models, optims, noise, HER)
+    agent = BCDDPGAgent(env, args, hyper_params, models, optims, noise, her)
 
     # run
     if args.test:
diff --git a/examples/reacher_v2/bc-sac.py b/examples/reacher_v2/bc-sac.py
index 2d1558a6..5773eb95 100644
--- a/examples/reacher_v2/bc-sac.py
+++ b/examples/reacher_v2/bc-sac.py
@@ -12,7 +12,7 @@
 import torch
 import torch.optim as optim
 
-from algorithms.bc.sac_agent import Agent
+from algorithms.bc.sac_agent import BCSACAgent
 from algorithms.common.networks.mlp import MLP, FlattenMLP, TanhGaussianDistParams
 from examples.reacher_v2.utils import ReacherHER
 
@@ -114,10 +114,10 @@ def run(env: gym.Env, args: argparse.Namespace, state_dim: int, action_dim: int)
     optims = (actor_optim, vf_optim, qf_1_optim, qf_2_optim)
 
     # HER
-    HER = ReacherHER if hyper_params["USE_HER"] else None
+    her = ReacherHER() if hyper_params["USE_HER"] else None
 
     # create an agent
-    agent = Agent(env, args, hyper_params, models, optims, target_entropy, HER)
+    agent = BCSACAgent(env, args, hyper_params, models, optims, target_entropy, her)
 
     # run
     if args.test:
diff --git a/examples/reacher_v2/ddpg.py b/examples/reacher_v2/ddpg.py
index d625c8c5..2a066f80 100644
--- a/examples/reacher_v2/ddpg.py
+++ b/examples/reacher_v2/ddpg.py
@@ -13,7 +13,7 @@
 
 from algorithms.common.networks.mlp import MLP
 from algorithms.common.noise import OUNoise
-from algorithms.ddpg.agent import Agent
+from algorithms.ddpg.agent import DDPGAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
@@ -101,7 +101,7 @@ def run(env: gym.Env, args: argparse.Namespace, state_dim: int, action_dim: int)
     optims = (actor_optim, critic_optim)
 
     # create an agent
-    agent = Agent(env, args, hyper_params, models, optims, noise)
+    agent = DDPGAgent(env, args, hyper_params, models, optims, noise)
 
     # run
     if args.test:
diff --git a/examples/reacher_v2/sac.py b/examples/reacher_v2/sac.py
index 1f37124f..3f2607d2 100644
--- a/examples/reacher_v2/sac.py
+++ b/examples/reacher_v2/sac.py
@@ -13,7 +13,7 @@
 import torch.optim as optim
 
 from algorithms.common.networks.mlp import MLP, FlattenMLP, TanhGaussianDistParams
-from algorithms.sac.agent import Agent
+from algorithms.sac.agent import SACAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
@@ -106,7 +106,7 @@ def run(env: gym.Env, args: argparse.Namespace, state_dim: int, action_dim: int)
     optims = (actor_optim, vf_optim, qf_1_optim, qf_2_optim)
 
     # create an agent
-    agent = Agent(env, args, hyper_params, models, optims, target_entropy)
+    agent = SACAgent(env, args, hyper_params, models, optims, target_entropy)
 
     # run
     if args.test:
diff --git a/examples/reacher_v2/td3.py b/examples/reacher_v2/td3.py
index c553a649..c488afad 100644
--- a/examples/reacher_v2/td3.py
+++ b/examples/reacher_v2/td3.py
@@ -13,7 +13,7 @@
 
 from algorithms.common.networks.mlp import MLP
 from algorithms.common.noise import GaussianNoise
-from algorithms.td3.agent import Agent
+from algorithms.td3.agent import TD3Agent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
@@ -115,7 +115,7 @@ def run(env: gym.Env, args: argparse.Namespace, state_dim: int, action_dim: int)
     optims = (actor_optim, critic_optim)
 
     # create an agent
-    agent = Agent(env, args, hyper_params, models, optims, noise)
+    agent = TD3Agent(env, args, hyper_params, models, optims, noise)
 
     # run
     if args.test:
diff --git a/examples/reacher_v2/utils.py b/examples/reacher_v2/utils.py
index ec228599..d2c73ba7 100644
--- a/examples/reacher_v2/utils.py
+++ b/examples/reacher_v2/utils.py
@@ -7,7 +7,7 @@
 
 import numpy as np
 
-from algorithms.common.abstract.her import HER as AbstractHER
+from algorithms.common.abstract.her import HER
 from algorithms.common.abstract.reward_fn import RewardFn
 
 
@@ -22,12 +22,12 @@ def __call__(self, transition: tuple, _) -> np.float64:
         return reward_dist + reward_ctrl
 
 
-class ReacherHER(AbstractHER):
+class ReacherHER(HER):
     """HER for Reacher-v2 environment."""
 
     def __init__(self, reward_func: RewardFn = ReacherRewardFn):
         """Initialization."""
-        AbstractHER.__init__(self, reward_func=reward_func)
+        HER.__init__(self, reward_func=reward_func)
 
     def fetch_desired_states_from_demo(self, _: list):
         """Return desired goal states from demonstration data.