Initial commit

4 files changed, 146 insertions, 0 deletions

diff --git a/code/sunlab/common/scaler/__init__.py b/code/sunlab/common/scaler/__init__.py
new file mode 100644
index 0000000..2a2281a
--- /dev/null
+++ b/code/sunlab/common/scaler/__init__.py
@@ -0,0 +1,2 @@
+from .max_abs_scaler import *
+from .quantile_scaler import *
diff --git a/code/sunlab/common/scaler/adversarial_scaler.py b/code/sunlab/common/scaler/adversarial_scaler.py
new file mode 100644
index 0000000..7f61725
--- /dev/null
+++ b/code/sunlab/common/scaler/adversarial_scaler.py
@@ -0,0 +1,44 @@
+class AdversarialScaler:
+    """# Scaler Class to use in Adversarial Autoencoder
+
+    For any scaler to be implemented, make sure to ensure each of the methods
+    are implemented:
+    - __init__ (optional)
+    - init
+    - load
+    - save
+    - __call__"""
+
+    def __init__(self, base_directory):
+        """# Scaler initialization
+
+        - Initialize the base directory of the model where it will live"""
+        self.base_directory = base_directory
+
+    def init(self, data):
+        """# Scaler initialization
+
+        Initialize the scaler transformation with the data
+        Should always return self in subclasses"""
+        raise NotImplementedError("Scaler initialization has not been implemented yet")
+
+    def load(self):
+        """# Scaler loading
+
+        Load the data scaler model from a file
+        Should always return self in subclasses"""
+        raise NotImplementedError("Scaler loading has not been implemented yet")
+
+    def save(self):
+        """# Scaler saving
+
+        Save the data scaler model"""
+        raise NotImplementedError("Scaler saving has not been implemented yet")
+
+    def transform(self, *args, **kwargs):
+        """# Scale the given data"""
+        return self.__call__(*args, **kwargs)
+
+    def __call__(self, *args, **kwargs):
+        """# Scale the given data"""
+        raise NotImplementedError("Scaler has not been implemented yet")
diff --git a/code/sunlab/common/scaler/max_abs_scaler.py b/code/sunlab/common/scaler/max_abs_scaler.py
new file mode 100644
index 0000000..56ea589
--- /dev/null
+++ b/code/sunlab/common/scaler/max_abs_scaler.py
@@ -0,0 +1,48 @@
+from .adversarial_scaler import AdversarialScaler
+
+
+class MaxAbsScaler(AdversarialScaler):
+    """# MaxAbsScaler
+
+    Scale the data based on the maximum-absolute value in each column"""
+
+    def __init__(self, base_directory):
+        """# MaxAbsScaler initialization
+
+        - Initialize the base directory of the model where it will live
+        - Initialize the scaler model"""
+        super().__init__(base_directory)
+        from sklearn.preprocessing import MaxAbsScaler as MAS
+
+        self.scaler_base = MAS()
+        self.scaler = None
+
+    def init(self, data):
+        """# Scaler initialization
+
+        Initialize the scaler transformation with the data"""
+        self.scaler = self.scaler_base.fit(data)
+        return self
+
+    def load(self):
+        """# Scaler loading
+
+        Load the data scaler model from a file"""
+        from pickle import load
+
+        with open(f"{self.base_directory}/portable/maxabs_scaler.pkl", "rb") as fhandle:
+            self.scaler = load(fhandle)
+        return self
+
+    def save(self):
+        """# Scaler saving
+
+        Save the data scaler model"""
+        from pickle import dump
+
+        with open(f"{self.base_directory}/portable/maxabs_scaler.pkl", "wb") as fhandle:
+            dump(self.scaler, fhandle)
+
+    def __call__(self, *args, **kwargs):
+        """# Scale the given data"""
+        return self.scaler.transform(*args, **kwargs)
diff --git a/code/sunlab/common/scaler/quantile_scaler.py b/code/sunlab/common/scaler/quantile_scaler.py
new file mode 100644
index 0000000..a0f53fd
--- /dev/null
+++ b/code/sunlab/common/scaler/quantile_scaler.py
@@ -0,0 +1,52 @@
+from .adversarial_scaler import AdversarialScaler
+
+
+class QuantileScaler(AdversarialScaler):
+    """# QuantileScaler
+
+    Scale the data based on the quantile distributions of each column"""
+
+    def __init__(self, base_directory):
+        """# QuantileScaler initialization
+
+        - Initialize the base directory of the model where it will live
+        - Initialize the scaler model"""
+        super().__init__(base_directory)
+        from sklearn.preprocessing import QuantileTransformer as QS
+
+        self.scaler_base = QS()
+        self.scaler = None
+
+    def init(self, data):
+        """# Scaler initialization
+
+        Initialize the scaler transformation with the data"""
+        self.scaler = self.scaler_base.fit(data)
+        return self
+
+    def load(self):
+        """# Scaler loading
+
+        Load the data scaler model from a file"""
+        from pickle import load
+
+        with open(
+            f"{self.base_directory}/portable/quantile_scaler.pkl", "rb"
+        ) as fhandle:
+            self.scaler = load(fhandle)
+        return self
+
+    def save(self):
+        """# Scaler saving
+
+        Save the data scaler model"""
+        from pickle import dump
+
+        with open(
+            f"{self.base_directory}/portable/quantile_scaler.pkl", "wb"
+        ) as fhandle:
+            dump(self.scaler, fhandle)
+
+    def __call__(self, *args, **kwargs):
+        """# Scale the given data"""
+        return self.scaler.transform(*args, **kwargs)