LOT use case (#145)

* ReadMe file for creating dataset from MIMICIII, and training LOS>3 classifier

* adding links to readMe file

* fixing the LR model

* Mimic_Hnadler, data_processeing, model and the main file

* all attacks are running

* add relevant files from MIMIC_Extract and mimic_code repos

* fix some bugs regarding nivduration

* revert conflicting tabular_mia files

* adding subset to the mimicDataset

* fixing loading the database and indices

* fixing physionet username, breaking the makefile in two commands.

* fixing data and output folder in gitingnore

* adding gitkeep

* adding gitkeep

* adding gitignore for data

* adding gitkeep to output

* adding gitkeep to output

* adding gitignore for output

* removing extra readme file

* update the readme file

* fixing copying output to the correct directory

* Add .gitignore to ignore contents of the data folder

* removing redundant files, fixind double sigmoid bug

* adding comments in the readmefile

examples/mia/LOS/.gitignore

@@ -0,0 +1,5 @@
+# Ignore everything inside the data folder
+data/*
+
+# But do not ignore the .gitkeep file
+!data/.gitkeep

examples/mia/LOS/audit.yaml

@@ -0,0 +1,43 @@
+audit:  # Configurations for auditing
+  random_seed: 1234  # Integer specifying the random seed
+  attack_list:
+    # rmia:
+    #   training_data_fraction: 0.5  # Fraction of the auxilary dataset to use for this attack (in each shadow model training)
+    #   attack_data_fraction: 0.5 # Fraction of auxiliary dataset to sample from during attack
+    #   num_shadow_models: 8 # Number of shadow models to train
+    #   online: True # perform online or offline attack
+    #   temperature: 2
+    #   gamma: 1.0
+    #   offline_a: 0.33 # parameter from which we compute p(x) from p_OUT(x) such that p_IN(x) = a p_OUT(x) + b.
+    #   offline_b: 0.66
+    # qmia:
+    #   training_data_fraction: 1.0  # Fraction of the auxilary dataset (data without train and test indices) to use for training the quantile regressor
+    #   epochs: 5  # Number of training epochs for quantile regression
+    # population:
+    #   attack_data_fraction: 1.0  # Fraction of the auxilary dataset to use for this attack
+    lira:
+      training_data_fraction: 0.5  # Fraction of the auxilary dataset to use for this attack (in each shadow model training)
+      num_shadow_models: 8 # Number of shadow models to train
+      online: True # perform online or offline attack
+      fixed_variance: True # Use a fixed variance for the whole audit
+      boosting: True
+    # loss_traj:
+    #   training_distill_data_fraction : 0.7 # Fraction of the auxilary dataset to use for training the distillation models D_s = (1-D_KD)/2
+    #   number_of_traj: 10 # Number of epochs (number of points in the loss trajectory)
+    #   label_only: False # True or False
+    #   mia_classifier_epochs: 100
+
+  output_dir: "./leakpro_output"
+  attack_type: "mia" #mia, gia
+
+target:
+  # Target model path
+  module_path: "utils/model.py"
+  model_class: "MimicLR" 
+  # Data paths
+  target_folder: "./target"
+  data_path: "./data/dataset.pkl"
+
+shadow_model:
+
+distillation_model:

examples/mia/LOS/mimic_handler.py

@@ -0,0 +1,65 @@
+
+import torch
+from torch import cuda, device, optim, sigmoid
+from torch.nn import BCELoss
+from torch.utils.data import DataLoader
+from tqdm import tqdm
+
+from leakpro import AbstractInputHandler
+
+class MimicInputHandler(AbstractInputHandler):
+    """Class to handle the user input for the CIFAR10 dataset."""
+
+    def __init__(self, configs: dict) -> None:
+        super().__init__(configs = configs)
+
+
+    def get_criterion(self)->None:
+        """Set the CrossEntropyLoss for the model."""
+        return BCELoss()
+
+    def get_optimizer(self, model:torch.nn.Module) -> None:
+        """Set the optimizer for the model."""
+        learning_rate = 0.1
+        momentum = 0.8
+        return optim.SGD(model.parameters(), lr=learning_rate, momentum=momentum)
+
+    def train(
+        self,
+        dataloader: DataLoader,
+        model: torch.nn.Module = None,
+        criterion: torch.nn.Module = None,
+        optimizer: optim.Optimizer = None,
+        epochs: int = None,
+    ) -> dict:
+        """Model training procedure."""
+
+        compute_device = device("cuda" if cuda.is_available() else "cpu")
+        model.to(compute_device)
+        model.train()
+
+        criterion = self.get_criterion()
+        optimizer = self.get_optimizer(model)
+
+        for e in tqdm(range(epochs), desc="Training Progress"):
+            model.train()
+            train_acc, train_loss = 0.0, 0.0
+
+            for data, target in dataloader:
+                target = target.float().unsqueeze(1)
+                data, target = data.to(compute_device, non_blocking=True), target.to(compute_device, non_blocking=True)
+                optimizer.zero_grad()
+                output = model(data)
+
+                loss = criterion(output, target)
+                pred = sigmoid(output) >= 0.5
+                train_acc += pred.eq(target).sum().item()
+
+                loss.backward()
+                optimizer.step()
+                train_loss += loss.item()
+
+        train_acc = train_acc/len(dataloader.dataset)
+        train_loss = train_loss/len(dataloader)
+
+        return {"model": model, "metrics": {"accuracy": train_acc, "loss": train_loss}}

examples/mia/LOS/mimiciii_prepration/MIMIC_Extract/.gitignore

@@ -0,0 +1,118 @@
+# Ignore all contents of output
+/output/*
+
+# Exclude the folder itself so it stay in the repository
+!/output/.gitkeep
+
+
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+.hypothesis/
+.pytest_cache/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# pyenv
+.python-version
+
+# celery beat schedule file
+celerybeat-schedule
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+
+
+*.ipynb_checkpoints*
+makejob
+
+# tags
+tags

examples/mia/LOS/mimiciii_prepration/MIMIC_Extract/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2019 MLforHealth
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

examples/mia/LOS/mimiciii_prepration/MIMIC_Extract/SQL_Queries/codes.sql

@@ -0,0 +1,11 @@
+SET SEARCH_PATH TO public,mimiciii;
+SELECT
+    i.icustay_id, d.subject_id, d.hadm_id,
+    array_agg(d.icd9_code ORDER BY seq_num ASC) AS icd9_codes
+FROM diagnoses_icd d 
+    LEFT OUTER JOIN (SELECT ccs_matched_id, icd9_code from ccs_dx) c
+    ON c.icd9_code = d.icd9_code
+    INNER JOIN icustays i
+    ON i.hadm_id = d.hadm_id AND i.subject_id = d.subject_id
+WHERE d.hadm_id IN ('{hadm_id}') AND seq_num IS NOT NULL
+GROUP BY i.icustay_id, d.subject_id, d.hadm_id

examples/mia/LOS/mimiciii_prepration/MIMIC_Extract/SQL_Queries/debug_codes.sql

@@ -0,0 +1,15 @@
+\echo "DEBUG ONLY"
+SET search_path TO mimiciii;
+SELECT
+    i.icustay_id, d.subject_id, d.hadm_id,
+    array_agg(d.icd9_code ORDER BY seq_num ASC) AS icd9_codes,
+    array_agg(c.ccs_matched_id ORDER BY seq_num ASC) AS ccs_codes
+FROM mimiciii.diagnoses_icd d 
+    LEFT OUTER JOIN (SELECT ccs_matched_id, icd9_code from mimiciii.ccs_dx) c
+    ON c.icd9_code = d.icd9_code
+    INNER JOIN icustays i
+    ON i.hadm_id = d.hadm_id AND i.subject_id = d.subject_id
+
+WHERE seq_num IS NOT NULL
+GROUP BY i.icustay_id, d.subject_id, d.hadm_id
+

examples/mia/LOS/mimiciii_prepration/MIMIC_Extract/SQL_Queries/debug_statics.sql

@@ -0,0 +1,81 @@
+\echo "This file is just for debugging"
+SET search_path TO public,mimiciii;
+select distinct
+    i.subject_id,
+    i.hadm_id,
+    i.icustay_id,
+    i.gender,
+    i.age as age,
+    i.ethnicity,
+    i.admission_type,
+    i.hospital_expire_flag,
+    i.hospstay_seq,
+    i.los_icu,
+    i.admittime,
+    i.dischtime,
+    i.intime,
+    i.outtime,
+    a.diagnosis AS diagnosis_at_admission,
+    a.insurance,
+    a.deathtime,
+    a.discharge_location,
+    CASE when a.deathtime between i.intime and i.outtime THEN 1 ELSE 0 END AS mort_icu,
+    CASE when a.deathtime between i.admittime and i.dischtime THEN 1 ELSE 0 END AS mort_hosp,
+    s.first_careunit,
+        c.fullcode_first,
+    c.dnr_first,
+    c.fullcode,
+    c.dnr,
+--     c.timednr_chart,
+    c.dnr_first_charttime,
+    c.cmo_first,
+    c.cmo_last,
+    c.cmo,
+    c.cmo_ds,
+--     c.timecmo_chart,
+    c.cmo_first_charttime,
+--     c.timecmo_nursingnote,
+    c.cmo_nursingnote_charttime,
+    sofa.sofa,
+    sofa.respiration as sofa_,
+    sofa.coagulation as sofa_,
+    sofa.liver as sofa_,
+    sofa.cardiovascular as sofa_,
+    sofa.cns as sofa_,
+    sofa.renal as sofa_,
+    sapsii.sapsii,
+    sapsii.sapsii_prob,
+    oasis.oasis,
+    oasis.oasis_prob,
+    COALESCE(f.readmission_30, 0) AS readmission_30
+FROM icustay_detail i
+    INNER JOIN admissions a ON i.hadm_id = a.hadm_id
+    INNER JOIN icustays s ON i.icustay_id = s.icustay_id
+    INNER JOIN code_status c ON i.icustay_id = c.icustay_id
+    LEFT OUTER JOIN (SELECT d.icustay_id, 1 as readmission_30
+              FROM icustays c, icustays d
+              WHERE c.subject_id=d.subject_id
+              AND c.icustay_id > d.icustay_id
+              AND c.intime - d.outtime <= interval '30 days'
+              AND c.outtime = (SELECT MIN(e.outtime) from icustays e 
+                                WHERE e.subject_id=c.subject_id
+                                AND e.intime>d.outtime)) f
+              ON i.icustay_id=f.icustay_id
+    LEFT OUTER JOIN (SELECT icustay_id, sofa,  respiration, coagulation, liver, cardiovascular, cns, renal 
+              FROM sofa) sofa
+              ON i.icustay_id=sofa.icustay_id
+    LEFT OUTER JOIN (SELECT icustay_id, sapsii,  sapsii_prob 
+                FROM sapsii) sapsii
+                ON sapsii.icustay_id=i.icustay_id
+    LEFT OUTER JOIN (SELECT icustay_id, oasis, oasis_prob
+                FROM oasis) oasis
+                ON oasis.icustay_id=i.icustay_id
+WHERE s.first_careunit NOT like 'NICU'
+    and i.hadm_id is not null and i.icustay_id is not null
+    and i.hospstay_seq = 1
+    and i.icustay_seq = 1
+    and i.age >= 16
+    and i.los_icu >= 1
+    and (i.outtime >= (i.intime + interval '12 hours'))
+    and (i.outtime <= (i.intime + interval '250 hours'))
+ORDER BY subject_id

examples/mia/LOS/mimiciii_prepration/MIMIC_Extract/SQL_Queries/notes.sql

@@ -0,0 +1,7 @@
+SELECT n.subject_id, n.hadm_id, i.icustay_id, n.chartdate, n.charttime, n.category, n.description, n.text
+FROM noteevents n INNER JOIN icustays i on i.hadm_id = n.hadm_id
+WHERE
+  iserror IS NULL
+  AND (n.chartdate <= i.outtime OR n.charttime <= i.outtime)
+  AND n.hadm_id IN ('{hadm_id}')
+  AND n.subject_id IN ('{subject_id}')