Add UCI experiments + refactor

cieg/experiments/__init__.py

@@ -0,0 +1,6 @@
+from .susy import Susy
+from .wine import Wine
+from .higgs import Higgs
+from .us_census import Census
+from .covertype import Covertype
+from .osteoarthritis import Osteoarthritis

cieg/experiments/covertype.py

@@ -0,0 +1,16 @@
+import pandas as pd
+import torch
+
+from .experiment import Experiment
+
+
+class Covertype(Experiment):
+
+    def __init__(self):
+        url = "https://archive.ics.uci.edu/ml/machine-learning-databases/covtype/covtype.data.gz"
+        self.load_dataset_if_not_exists('data', url=url)
+        # See covtype.info for description of the attributes
+        dataset = pd.read_csv("data/covtype.data.gz", compression='gzip', delimiter=',', header=None)
+        X = torch.transpose(torch.tensor(dataset.values).float(), 0, 1)
+
+        super().__init__(X[4:8, :], name="Covertype")

cieg/experiments/experiment.py

@@ -0,0 +1,64 @@
+import pandas as pd
+import requests
+
+from cieg.eigenvectors import *
+from cieg.experiments.methods import *
+from cieg.experiments.utils import *
+from cieg.utils.covariance import cov
+from cieg.utils.draw import *
+
+
+class Experiment:
+    def __init__(self, X, name):
+        self.X = X
+        self.name = name
+
+    @staticmethod
+    def load_dataset_if_not_exists(path, url):
+        if not os.path.exists(path):
+            os.mkdir(path)
+
+        filename = os.path.join(path, os.path.basename(url))
+        if not os.path.exists(filename):
+            print("Loading dataset")
+            data = requests.get(url).content
+            with open(filename, "wb") as file:
+                file.write(data)
+
+    def run(self, path):
+        X = preprocess(self.X)
+        X_pd = pd.DataFrame(torch.transpose(X, 0, 1).data.numpy(), dtype='float64')
+
+        sigma = cov(X)
+        emp_prec = torch.inverse(sigma)
+        eig = get_eig(sigma)
+
+        print("----------- RESAMPLING - bootstrap pmatrix -----------")
+        pmatrix_lower, pmatrix_upper = resample_pmatrix(X_pd, n_iterations=100, rng=check_random_state(0))
+        print_pmatrix_bounds(pmatrix_lower, pmatrix_upper, emp_prec)
+        check_pmatrix_bounds(pmatrix_lower, pmatrix_upper, emp_prec)
+        plot_and_save_bounds(pmatrix_lower,
+                             pmatrix_upper,
+                             emp_prec,
+                             f"Bounds on the precision matrix using resampling on {self.name}",
+                             path)
+
+        print("----------- OUR METHOD -----------")
+        # Bounds on eigendecomposition
+        eigvals_lower, eigvals_upper, eigvects_lower, eigvects_upper = cieg(X, sigma, eig)
+        print_eig_bounds(eigvals_lower, eigvals_upper, eigvects_lower, eigvects_upper, eig)
+        check_eig_bounds(eigvals_lower, eigvals_upper, eigvects_lower, eigvects_upper, eig)
+
+        # Bounds on precision matrix
+        pmatrix_lower, pmatrix_upper, _ = pmatrix_bounds(eigvals_lower,
+                                                         eigvals_upper,
+                                                         eigvects_lower,
+                                                         eigvects_upper,
+                                                         sigma, eig)
+        print_pmatrix_bounds(pmatrix_lower, pmatrix_upper, emp_prec)
+        check_pmatrix_bounds(pmatrix_lower, pmatrix_upper, emp_prec)
+        plot_and_save_bounds(pmatrix_lower,
+                             pmatrix_upper,
+                             emp_prec,
+                             f"Bounds on the precision matrix using our method on {self.name}",
+                             path)

cieg/experiments/higgs.py

@@ -0,0 +1,25 @@
+import pandas as pd
+import torch
+
+from .experiment import Experiment
+
+
+class Higgs(Experiment):
+
+    def __init__(self, **kwargs):
+        url = "https://archive.ics.uci.edu/ml/machine-learning-databases/00280/HIGGS.csv.gz"
+        columns = ['class', 'lepton pT', 'lepton eta', 'lepton phi', 'missing energy magnitude', 'missing energy phi',
+                   'jet 1 pt', 'jet 1 eta', 'jet 1 phi', 'jet 1 b-tag', 'jet 2 pt', 'jet 2 eta', 'jet 2 phi',
+                   'jet 2 b-tag', 'jet 3 pt', 'jet 3 eta', 'jet 3 phi', 'jet 3 b-tag', 'jet 4 pt', 'jet 4 eta',
+                   'jet 4 phi', 'jet 4 b-tag', 'm_jj', 'm_jjj', 'm_lv', 'm_jlv', 'm_bb', 'm_wbb', 'm_wwbb']
+
+        self.load_dataset_if_not_exists('data', url=url)
+        dataset = pd.read_csv("data/HIGGS.csv.gz",
+                              compression='gzip',
+                              delimiter=',',
+                              header=None,
+                              names=columns,
+                              usecols=kwargs.pop('usecols'))
+
+        X = torch.transpose(torch.tensor(dataset.values).float(), 0, 1)
+        super().__init__(X, kwargs.pop("name"))

cieg/experiments/osteoarthritis.py

@@ -1,64 +1,20 @@
-import warnings
-
 import pandas as pd
 
-from cieg.eigenvectors import *
-from cieg.experiments.methods import *
-from cieg.experiments.utils import *
 from cieg.utils import *
-from cieg.utils.covariance import cov
-
-warnings.filterwarnings(
-    "ignore", category=UserWarning
-)
-
-path = create_folders()
-
-columns_to_read = ['Side', 'WOMAC', 'OSTM']  # OSTM OSFM
-print(f"Features used: {columns_to_read}")
-
-# Read data
-X_pd = pd.read_csv('../data/oai_most_bl_aleksei_sep20_w_dataset_col.csv', usecols=columns_to_read)
-if 'Side' in columns_to_read:
-    X_pd['Side'] = X_pd['Side'].map({"R": 0, "L": 1})
-X_pd = X_pd.dropna()
-X = torch.transpose(torch.tensor(X_pd.values).float(), 0, 1)
-print(f"X shape: {X.shape}")
-
-X = preprocess(X)
-X_pd = pd.DataFrame(torch.transpose(X, 0, 1).data.numpy(), dtype='float64')
+from .experiment import Experiment
 
-sigma = cov(X)
-emp_prec = torch.inverse(sigma)
-eig = get_eig(sigma)
 
+class Osteoarthritis(Experiment):
 
-print("----------- RESAMPLING - bootstrap pmatrix -----------")
-pmatrix_lower, pmatrix_upper = resample_pmatrix(X_pd, n_iterations=100, rng=check_random_state(0))
-print_pmatrix_bounds(pmatrix_lower, pmatrix_upper, emp_prec)
-check_pmatrix_bounds(pmatrix_lower, pmatrix_upper, emp_prec)
-plot_and_save_bounds(pmatrix_lower,
-                     pmatrix_upper,
-                     emp_prec,
-                     "Osteoarthritis bounds on the precision matrix using resampling",
-                     path)
+    def __init__(self, **kwargs):
+        columns_to_read = kwargs.pop("usecols")
+        print(f"Features used: {columns_to_read}")
 
-print("----------- OUR METHOD -----------")
-# Bounds on eigendecomposition
-eigvals_lower, eigvals_upper, eigvects_lower, eigvects_upper = cieg(X, sigma, eig)
-print_eig_bounds(eigvals_lower, eigvals_upper, eigvects_lower, eigvects_upper, eig)
-check_eig_bounds(eigvals_lower, eigvals_upper, eigvects_lower, eigvects_upper, eig)
+        # Read data
+        X_pd = pd.read_csv('data/oai_most_bl_aleksei_sep20_w_dataset_col.csv', usecols=columns_to_read)
+        if 'Side' in columns_to_read:
+            X_pd['Side'] = X_pd['Side'].map({"R": 0, "L": 1})
+        X_pd = X_pd.dropna()
+        X = torch.transpose(torch.tensor(X_pd.values).float(), 0, 1)
 
-# Bounds on precision matrix
-pmatrix_lower, pmatrix_upper, _ = pmatrix_bounds(eigvals_lower,
-                                              eigvals_upper, 
-                                              eigvects_lower,
-                                              eigvects_upper, 
-                                              sigma, eig)
-print_pmatrix_bounds(pmatrix_lower, pmatrix_upper, emp_prec)
-check_pmatrix_bounds(pmatrix_lower, pmatrix_upper, emp_prec)
-plot_and_save_bounds(pmatrix_lower,
-                     pmatrix_upper,
-                     emp_prec,
-                     "Osteoarthritis bounds on the precision matrix using our method",
-                     path)
+        super().__init__(X, "Osteoarthritis")

cieg/experiments/run.py

@@ -0,0 +1,31 @@
+import warnings
+from cieg.experiments import *
+from cieg.utils import create_folders
+
+warnings.filterwarnings(
+    "ignore", category=UserWarning
+)
+
+census_columns = ["iClass", "dIncome1", "iEnglish", "dHours"]
+higgs_first = ['m_jj', 'm_jjj', 'm_lv', 'm_jlv']
+higgs_second = ['m_bb', 'm_wbb', 'm_wwbb']
+osteoarthritis = ['Side', 'WOMAC', 'OSTM']  # OSTM OSFM
+
+
+EXPERIMENTS = {
+    'higgs_first': Higgs(usecols=higgs_first, name='Higgs first'),
+    'higgs_second': Higgs(usecols=higgs_second, name='Higgs second'),
+    'osteoarthritis': Osteoarthritis(usecols=osteoarthritis)
+}
+
+
+def run():
+    for exp_name in EXPERIMENTS.keys():
+        print(f"Running {exp_name}")
+        path = create_folders()
+        experiment = EXPERIMENTS[exp_name]
+        experiment.run(path)
+
+
+if __name__ == '__main__':
+    run()

cieg/experiments/susy.py

@@ -0,0 +1,15 @@
+import pandas as pd
+import torch
+
+from .experiment import Experiment
+
+
+class Susy(Experiment):
+
+    def __init__(self):
+        url = "https://archive.ics.uci.edu/ml/machine-learning-databases/00279/SUSY.csv.gz"
+        self.load_dataset_if_not_exists('data', url=url)
+        dataset = pd.read_csv("data/SUSY.csv.gz", compression='gzip', delimiter=',', header=None)
+
+        X = torch.transpose(torch.tensor(dataset.values).float(), 0, 1)
+        super().__init__(X[:4, :], name="Susy")

cieg/experiments/us_census.py

@@ -0,0 +1,15 @@
+import pandas as pd
+import torch
+
+from .experiment import Experiment
+
+
+class Census(Experiment):
+
+    def __init__(self, **kwargs):
+        url = "https://archive.ics.uci.edu/ml/machine-learning-databases/census1990-mld/USCensus1990.data.txt"
+        self.load_dataset_if_not_exists('data', url=url)
+        X_input = pd.read_csv('data/USCensus1990.data.txt', usecols=kwargs.pop('usecols'))
+
+        X = torch.transpose(torch.tensor(X_input.values).float(), 0, 1)
+        super().__init__(X, name="Census")

cieg/experiments/wine.py

@@ -0,0 +1,12 @@
+from sklearn.datasets import load_wine
+
+from cieg.experiments.utils import *
+from .experiment import Experiment
+
+
+class Wine(Experiment):
+    def __init__(self):
+        dataset = load_wine()
+
+        X = torch.transpose(torch.tensor(dataset.data).float(), 0, 1)
+        super().__init__(X[4:7, :], name="Wine")