Berack/upo-ml
0
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Bugfixes

Browse Source 
- fix bug recursion on init
- fix bug linux path
- added regularization
This commit is contained in:
Giacomo Bertolazzi
2024-05-02 14:27:57 +02:00
parent 3a4e07afc8
commit 3e9bcb87e9
3 changed files with 31 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
.vscode/launch.json vendored
View File

@@ -8,7 +8,7 @@
"name": "Python Debugger: App", "name": "Python Debugger: App",
"type": "debugpy", "type": "debugpy",
"request": "launch", "request": "launch",
"program": "src\\app.py", "program": "src/app.py",
"console": "integratedTerminal" "console": "integratedTerminal"
} }
] ]

16
src/app.py
View File

@@ -3,8 +3,12 @@ from learning.supervised import LinearRegression, LogisticRegression, MultiLayer
from learning.ml import MLAlgorithm from learning.ml import MLAlgorithm
from typing import Callable from typing import Callable
DATASET = "datasets/"
REGRESSION = DATASET + "regression/"
CLASSIFICATION = DATASET + "classification/"
def auto_mpg() -> tuple[int, MLAlgorithm]: def auto_mpg() -> tuple[int, MLAlgorithm]:
ds = Dataset("datasets\\auto-mpg.csv", "MPG", TargetType.Regression) ds = Dataset(REGRESSION + "auto-mpg.csv", "MPG", TargetType.Regression)
ds.numbers(["HP"]) ds.numbers(["HP"])
ds.handle_na() ds.handle_na()
@@ -12,7 +16,7 @@ def auto_mpg() -> tuple[int, MLAlgorithm]:
return (1000, LinearRegression(ds, learning_rate=0.0001)) return (1000, LinearRegression(ds, learning_rate=0.0001))
def automobile() -> tuple[int, MLAlgorithm]: def automobile() -> tuple[int, MLAlgorithm]:
ds = Dataset("datasets\\regression\\automobile.csv", "symboling", TargetType.Regression) ds = Dataset(REGRESSION + "automobile.csv", "symboling", TargetType.Regression)
attributes_to_modify = ["fuel-system", "engine-type", "drive-wheels", "body-style", "make", "engine-location", "aspiration", "fuel-type", "num-of-cylinders", "num-of-doors"] attributes_to_modify = ["fuel-system", "engine-type", "drive-wheels", "body-style", "make", "engine-location", "aspiration", "fuel-type", "num-of-cylinders", "num-of-doors"]
ds.factorize(attributes_to_modify) ds.factorize(attributes_to_modify)
@@ -22,19 +26,19 @@ def automobile() -> tuple[int, MLAlgorithm]:
return (1000, LinearRegression(ds, learning_rate=0.004)) return (1000, LinearRegression(ds, learning_rate=0.004))
def power_plant() -> tuple[int, MLAlgorithm]: def power_plant() -> tuple[int, MLAlgorithm]:
ds = Dataset("datasets\\regression\\power-plant.csv", "energy-output", TargetType.Regression) ds = Dataset(REGRESSION + "power-plant.csv", "energy-output", TargetType.Regression)
ds.normalize() ds.normalize()
return (80, LinearRegression(ds, learning_rate=0.1)) return (80, LinearRegression(ds, learning_rate=0.1))
def electrical_grid() -> tuple[int, MLAlgorithm]: def electrical_grid() -> tuple[int, MLAlgorithm]:
ds = Dataset("datasets\\classification\\electrical_grid.csv", "stabf", TargetType.Classification) ds = Dataset(CLASSIFICATION + "electrical_grid.csv", "stabf", TargetType.Classification)
ds.factorize(["stabf"]) ds.factorize(["stabf"])
ds.normalize() ds.normalize()
return (1000, LogisticRegression(ds, learning_rate=0.08)) return (1000, LogisticRegression(ds, learning_rate=0.08))
def frogs() -> tuple[int, MLAlgorithm]: def frogs() -> tuple[int, MLAlgorithm]:
ds = Dataset("datasets\\classification\\frogs.csv", "Species", TargetType.MultiClassification) ds = Dataset(CLASSIFICATION + "frogs.csv", "Species", TargetType.MultiClassification)
ds.remove(["Family", "Genus", "RecordID"]) ds.remove(["Family", "Genus", "RecordID"])
ds.factorize(["Species"]) ds.factorize(["Species"])
return (1000, MultiLayerPerceptron(ds, learning_rate=0.08)) return (1000, MultiLayerPerceptron(ds, learning_rate=0.08))
@@ -55,5 +59,5 @@ def learn_dataset(function:Callable[..., tuple[int, MLAlgorithm]], epochs:int=10
return ml return ml
if __name__ == "__main__": if __name__ == "__main__":
ml = learn_dataset(automobile) ml = learn_dataset(electrical_grid)
print(ml.accuracy(ml.testset)) print(ml.accuracy(ml.testset))

24
src/learning/supervised.py
View File

@@ -8,16 +8,21 @@ from learning.data import Dataset, Data
class GradientDescent(MLAlgorithm): class GradientDescent(MLAlgorithm):
theta:np.ndarray theta:np.ndarray
alpha:float alpha:float
lambd:float
def __init__(self, dataset:Dataset, learning_rate:float=0.1) -> None: def __init__(self, dataset:Dataset, learning_rate:float=0.1, regularization:float=0.01) -> None:
self.__init__(dataset) super().__init__(dataset)
self.theta = np.random.rand(self.learnset.param) self.theta = np.random.rand(self.learnset.param)
self.alpha = max(0, learning_rate) self.alpha = max(0, learning_rate)
self.lambd = max(0, regularization)
def learning_step(self) -> float: def learning_step(self) -> float:
x, y, m, _ = self.learnset.as_tuple() x, y, m, _ = self.learnset.as_tuple()
self.theta -= self.alpha * (1/m) * np.sum((self._h0(x) - y) * x.T, axis=1) regularization = (self.lambd / m) * self.theta
regularization[0] = 0
derivative = self.alpha * (1/m) * np.sum((self._h0(x) - y) * x.T, axis=1)
self.theta -= derivative + regularization
return self._loss(x, y, m) return self._loss(x, y, m)
def predict_loss(self, dataset:Data) -> float: def predict_loss(self, dataset:Data) -> float:
@@ -54,5 +59,16 @@ class MultiLayerPerceptron(MLAlgorithm):
neurons: list[np.ndarray] neurons: list[np.ndarray]
def __init__(self, dataset:Dataset, layers:list[int]=[4,3]) -> None: def __init__(self, dataset:Dataset, layers:list[int]=[4,3]) -> None:
self.__init__(dataset) super().__init__(dataset)
def _h0(self, x:np.ndarray) -> np.ndarray:
pass
def learning_step(self) -> float:
pass
def predict_loss(self, dataset:np.ndarray) -> float:
pass
def get_parameters(self):
pass
def set_parameters(self, parameters):
pass