From d8245421c9c698d64c392296c87c03b06908fbc7 Mon Sep 17 00:00:00 2001
From: sam chou <sam@thechous.com>
Date: Wed, 18 Dec 2024 14:47:13 -0800
Subject: [PATCH] removing unecessary file

---
 afqinsight/nn/tf_model_training.py | 124 -----------------------------
 1 file changed, 124 deletions(-)
 delete mode 100644 afqinsight/nn/tf_model_training.py

diff --git a/afqinsight/nn/tf_model_training.py b/afqinsight/nn/tf_model_training.py
deleted file mode 100644
index 8dbcf86..0000000
--- a/afqinsight/nn/tf_model_training.py
+++ /dev/null
@@ -1,124 +0,0 @@
-import tempfile
-
-import numpy as np
-import tensorflow as tf
-from neurocombat_sklearn import CombatModel
-from sklearn.impute import SimpleImputer
-from sklearn.model_selection import train_test_split
-
-from afqinsight import AFQDataset
-from afqinsight.nn.tf_models import mlp4
-
-dataset = AFQDataset.from_study("hbn")
-print(dataset)
-dataset.drop_target_na()
-# tensorflow_dataset = dataset.as_tensorflow_dataset(
-#     bundles_as_channels=True, channels_last=True
-# )
-# print(tensorflow_dataset)
-
-
-# converts the input array to a tensor, fit for training
-def array_to_tensor(input_array):
-    return np.transpose(
-        np.array([input_array[:, i * 100 : (i + 1) * 100] for i in range(48)]),
-        (1, 2, 0),
-    )
-
-
-# fills in missing values with the median of the column
-def prep_data(input_array, site):
-    return array_to_tensor(
-        CombatModel().fit_transform(
-            SimpleImputer(strategy="median").fit_transform(input_array), site
-        )
-    )
-
-
-X = dataset.X
-y = dataset.y[:, 0]
-site = dataset.y[:, 2, None]
-groups = dataset.groups
-feature_names = dataset.feature_names
-group_names = dataset.group_names
-subjects = dataset.subjects
-
-X_train, X_test, y_train, y_test, site_train, site_test = train_test_split(
-    X, y, site, test_size=0.2
-)
-
-X_train, X_val, y_train, y_val, site_train, site_val = train_test_split(
-    X_train, y_train, site_train, test_size=0.2
-)
-
-X_train = prep_data(X_train, site_train)
-X_test = prep_data(X_test, site_test)
-X_val = prep_data(X_val, site_val)
-
-train_dataset = tf.data.Dataset.from_tensor_slices(
-    (X_train.astype(np.float32), y_train.astype(np.float32))
-)
-print(train_dataset)
-val_dataset = tf.data.Dataset.from_tensor_slices(
-    (X_val.astype(np.float32), y_val.astype(np.float32))
-)
-
-for i in range(10):
-    print(f"Y_TRAIN {i}: {y_train[i]}")
-model = mlp4(
-    input_shape=X_train.shape[1:], n_classes=1, verbose=True, output_activation="linear"
-)
-# model = cnn_lenet(input_shape=(100, 48),
-#           n_classes=1, verbose=True, output_activation='linear')
-# model = cnn_vgg(input_shape=(100, 48, 1),
-#           n_classes=1, verbose=True, output_activation='linear')
-
-lr = 0.0001
-batch_size = 32
-n_epochs = 100
-
-model.compile(
-    loss="mean_squared_error",
-    optimizer=tf.keras.optimizers.Adam(learning_rate=lr),
-    metrics=[
-        "mean_squared_error",
-        tf.keras.metrics.RootMeanSquaredError(name="rmse"),
-        "mean_absolute_error",
-    ],
-)
-
-ckpt_filepath = tempfile.NamedTemporaryFile(suffix=".weights.h5").name
-ckpt = tf.keras.callbacks.ModelCheckpoint(
-    filepath=ckpt_filepath,
-    monitor="val_loss",
-    verbose=0,
-    save_best_only=True,
-    save_weights_only=True,
-    mode="auto",
-)
-
-early_stopping = tf.keras.callbacks.EarlyStopping(
-    monitor="val_loss", min_delta=0.001, mode="min", patience=100
-)
-
-reduce_lr = tf.keras.callbacks.ReduceLROnPlateau(
-    monitor="val_loss", factor=0.5, patience=20, verbose=1
-)
-
-callbacks = [early_stopping, ckpt, reduce_lr]
-train_dataset = train_dataset.batch(batch_size)
-val_dataset = val_dataset.batch(batch_size)
-
-history = model.fit(
-    train_dataset,
-    epochs=n_epochs,
-    validation_data=val_dataset,
-    callbacks=callbacks,
-    verbose=2,
-)
-
-results = model.evaluate(X_test.astype(np.float32), y_test.astype(np.float32))
-print(
-    f"Test Results - Loss: {results[0]}, RMSE: {
-    results[1]}, MAE: {results[2]}, MSE: {results[3]}"
-)