Real world version #3

ouaelesi · 2024-07-22T02:52:26Z

No description provided.

mariemkallel16 · 2025-07-16T12:44:09Z

MEDfl/rw/client.py

+from sklearn.metrics import accuracy_score, roc_auc_score
+from model import Net  # Local model definition
+import socket
+import platform


socket et platform non utilisées dans le code

il faut aussi séparer les bibliothèques par groupe (bibliothèques standard (argparse), bibliothèques tierces (pandas, torch, flwr) et importations locales (from model import Net))

mariemkallel16 · 2025-07-16T12:53:25Z

MEDfl/rw/client.py

+        """
+        Get model parameters as a list of NumPy arrays.
+        """
+        return [val.cpu().numpy() for val in self.mo]()


self.model je pense? et j'ai pas compris pour les parenthèses à la fin

Olivier998 · 2025-07-15T21:00:05Z

MEDfl.egg-info/PKG-INFO

-The MEDfl package requires *python 3.9* or more to be run. If you don't have it installed  on your machine, check out the following link  [Python ](https://www.python.org/downloads/).
-It also requires MySQL database
+The MEDfl package requires *python 3.9* or more to be run. If you don't have it installed  on your machine, check out the following link [Python](https://www.python.org/downloads/).
+It also requires MySQL database.


python >=3.9
two spaces between "installed" and "on"

Should you also give the link to install MySQL?

Olivier998 · 2025-07-15T21:01:06Z

MEDfl.egg-info/PKG-INFO

+It also requires MySQL database.

 ### Package installation
 For now, you can  install the ``MEDfl``package as:


Olivier998 · 2025-07-15T21:03:09Z

MEDfl.egg-info/PKG-INFO

-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.
-```


Pourquoi as tu enlevé la license ici et la mention du consortium MEDOMICS?

Olivier998 · 2025-07-15T21:05:24Z

MEDfl/rw/client.py

+    Configuration for Differential Privacy (DP) settings.
+
+    Attributes:
+        noise_multiplier (float): Noise multiplier for DP.


C'est quoi? epsilon?

est-ce que le parametre delta est inclu?

Olivier998 · 2025-07-15T21:16:58Z

MEDfl/rw/client.py

+        dp_config: DPConfig = None,
+    ):
+        """
+        Initialize client by loading data, creating model, optimizer,


Initializes*

Olivier998 · 2025-07-16T12:35:28Z

MEDfl/rw/strategy.py

+    auc = sum(metrics.get("eval_auc", 0.0) * num_examples
+              for num_examples, metrics in results) / total_examples
+
+    return {"eval_loss": loss, "eval_accuracy": accuracy, "eval_auc": auc}


Les fonctions aggregate_fit_metrics et aggregate_eval_metrics devraient être regroupées en une seule fonction

Faudrait-il ajouter d'autres mesures comme std? L'aggrégation pondérée ici peut être biaisée par un centre plus populeux que les autres

Olivier998 · 2025-07-16T12:48:36Z

notebooks/banchmarking/network_architecture/net_bachmarking_conf_1.ipynb

  },
  {
   "cell_type": "code",
-   "execution_count": 1,


est-ce que ce fichier doit être dans le git?

Attention au nom du folder (banchmarking, bachmarking, ...)

Olivier998 · 2025-07-16T12:54:02Z

notebooks/banchmarking/network_architecture/net_bachmarking_conf_1.ipynb

   ]
  },
  {
   "cell_type": "code",


Commentaire en lien avec les folders de données: Il y a des données de MIMIC-Eicu dans le repo, ces données ne devraient pas être disponible ici puisque théoriquement, ce ne sont pas des données publiques.

Olivier998 · 2025-07-16T12:55:49Z

requirements.txt

 sphinx-rtd-dark-mode==1.2.4
 plotly==5.19.0
 optuna==3.5.0
-mysql


mysql n'était pas un package obligatoire ? Ou c'est seulement MySQL database? (selon README)

Olivier998 · 2025-07-16T12:56:30Z

setup.py

 setup(
    name="MEDfl",
-    version="0.1.37",
+    version="2.0.4.dev1",


mariemkallel16 · 2025-07-16T13:15:47Z

MEDfl/rw/strategy.py

+               for num_examples, metrics in results) / total_examples
+    accuracy = sum(metrics.get("train_accuracy", 0.0) * num_examples
+                   for num_examples, metrics in results) / total_examples
+    auc = sum(metrics.get("train_auc", 0.0) * num_examples


tu pourrais faire appel à une fonction générique pour appliquer la même formule

et prévoir aussi une condition pour vérifier que total_examples est différent de 0 ? (cas rare mais possible)

mariemkallel16 · 2025-07-16T13:35:50Z

MEDfl/rw/strategy.py

+        self.min_fit_clients = min_fit_clients
+        self.min_evaluate_clients = min_evaluate_clients
+        self.min_available_clients = min_available_clients
+        self.initial_parameters = initial_parameters or []


peut être mettre None au lieu de [] pour différencier les modèles non instanciés vs les modèles sans paramètres?

LaribiHakima · 2025-07-16T13:19:56Z

MEDfl/rw/client.py

+        max_grad_norm: float = 1.0,
+        batch_size: int = 32,
+        secure_rng: bool = False,
+    ):


Je sais que tu as décrit les attributs de la classe plus haut c'est pourquoi tu ne décris pas les arguments de la méthode ici, mais dans tout le reste du code tu décris les arguments et pas les attributs, peut être faire de même ici.

LaribiHakima · 2025-07-16T13:22:29Z

MEDfl/rw/client.py

+        # ---------- Load Data ----------
+        df = pd.read_csv(data_path)
+        X = df.iloc[:, :-1].values
+        y = df.iloc[:, -1].values


Plutôt ajouter deux arguments de "predictor columns" et "target column". Il arrive que certains datasets aient des colonnes d'identifiants, dates, etc qui ne sont pas utiles à la prédiction.

LaribiHakima · 2025-07-16T13:28:13Z

MEDfl/rw/client.py

+        self.X_tensor = torch.tensor(X, dtype=torch.float32)
+        self.y_tensor = torch.tensor(y, dtype=torch.float32)
+
+        batch_size = dp_config.batch_size if dp_config else 32


Ce paper parle du fait que la batch size n'affecte pas la performance du modèle, et qu'en fait, on devrait choisir le maximum que notre hardware est capable de supporter puis faire une optimisation d'hps dessus. i.e: pas besoin d'optimiser le batch size.

https://proceedings.neurips.cc/paper_files/paper/2017/file/a5e0ff62be0b08456fc7f1e88812af3d-Paper.pdf

LaribiHakima · 2025-07-16T13:29:02Z

MEDfl/rw/client.py

+        # ---------- Model and Optimizer ----------
+        input_dim = X.shape[1]
+        self.model = Net(input_dim)
+        self.criterion = nn.BCEWithLogitsLoss()


Clairement, tu fais uniquement de la classification binaire, peut être le préciser dans l'entête de ta classe.

LaribiHakima · 2025-07-16T13:37:31Z

MEDfl/rw/model.py

+        # Output layer: 32 → 1
+        self.fc3 = nn.Linear(32, 1)
+        # Dropout with 30% probability
+        self.dropout = nn.Dropout(0.3)


De mon expérience, c'est vraiment énorme 30% de dropout

LaribiHakima · 2025-07-16T13:39:16Z

MEDfl/rw/server.py

@@ -0,0 +1,134 @@
+import flwr as fl


Ajouter un entête

LaribiHakima · 2025-07-16T16:04:51Z

MEDfl/rw/strategy.py

+import flwr as fl
+from typing import Callable, Optional, Dict, Any, Tuple, List


LaribiHakima · 2025-07-16T17:50:40Z

MEDfl/rw/strategy.py

+        self.fit_metrics_aggregation_fn = fit_metrics_aggregation_fn or aggregate_fit_metrics
+        self.evaluate_metrics_aggregation_fn = evaluate_metrics_aggregation_fn or aggregate_eval_metrics
+
+        self.strategy_object: Optional[fl.server.strategy.Strategy] = None


Le typing n'est pas nécessaire à ce niveau

LaribiHakima · 2025-07-16T17:53:58Z

MEDfl/rw/strategy.py

+        StrategyClass = getattr(fl.server.strategy, self.name)
+
+        # Prepare parameters for strategy
+        params: Dict[str, Any] = {


En python, ce genre de typing n'est pas nécessaire

ouaelesi and others added 12 commits May 16, 2024 18:00

add: nodes config benchmarks

0fff63a

working on benchmarks

087fede

working on benchmarks

e0f3e5f

working on benchmarks

7add080

update authors

58b1e52

changing to sqllite

c7920f9

working on sqllite-benchmarks

c1985d5

publishing the verion 0.2.0

d82bd72

add: rw configuration

31af44b

add: rw server config messages

beabff3

add: comment classes

5d58e75

remove: build folder

94c57b3

ouaelesi changed the title ~~update authors~~ Real world version Jul 14, 2025

update: client code

b5e4076

mariemkallel16 reviewed Jul 16, 2025

View reviewed changes

Olivier998 reviewed Jul 16, 2025

View reviewed changes

mariemkallel16 reviewed Jul 16, 2025

View reviewed changes

LaribiHakima reviewed Jul 16, 2025

View reviewed changes

ouaelesi merged commit b5e4076 into main Oct 8, 2025
0 of 2 checks passed

		import flwr as fl
		from typing import Callable, Optional, Dict, Any, Tuple, List

Real world version #3

Real world version #3

Uh oh!

Conversation

ouaelesi commented Jul 22, 2024

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mariemkallel16 Jul 16, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

mariemkallel16 Jul 16, 2025 •

edited

Loading