Construction of Continuous Score Model¶
Author: Dr Masahiro Ono¶
Date: 2025-02-16¶
Aim:¶
This notebook demonstrates how to construct and analyse a continuous score model by transferring learnt model weights of a pre-trained 2-class classifier to a new logit model.
Steps:
- Set up: import a pre-trained 2-class classifier model and load independent test data
- Continuous Score Model Construction
- Score Analysis Using Quadratic Regression
In [1]:
#Import Neccessary Libraries
import os
import numpy as np
import pandas as pd
import tensorflow as tf
import matplotlib.pyplot as plt
import seaborn as sns
import statsmodels.api as sm
from matplotlib.backends.backend_pdf import PdfPages
from tensorflow.keras import Model
from tensorflow.keras.layers import Dense
from tensorflow.keras.models import load_model
from tensorflow.keras.utils import to_categorical
from sklearn.metrics import accuracy_score, confusion_matrix, precision_score, recall_score
from sklearn.model_selection import train_test_split
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import LabelEncoder, PolynomialFeatures
from sklearn.linear_model import LinearRegression
import importlib.resources
import TockyConvNetPy
from TockyConvNetPy import InstanceNormalization, log_scale_age
In [2]:
# Load Trained 2-Class Classifier
custom_objects = {"InstanceNormalization": InstanceNormalization}
with importlib.resources.path("TockyConvNetPy.data.Foxp3DevAge.models", "Foxp3DevAge2ClassClassifier.keras") as model_path:
model = load_model(str(model_path),custom_objects=custom_objects)
model.summary()
Model: "model"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
image_input (InputLayer) [(None, 100, 100, 1 0 []
)]
conv1 (Conv2D) (None, 100, 100, 16 160 ['image_input[0][0]']
)
inst_norm1 (InstanceNormalizat (None, 100, 100, 16 32 ['conv1[0][0]']
ion) )
activation1 (Activation) (None, 100, 100, 16 0 ['inst_norm1[0][0]']
)
attention1_conv (Conv2D) (None, 100, 100, 1) 17 ['activation1[0][0]']
attention1_multiply (Multiply) (None, 100, 100, 16 0 ['activation1[0][0]',
) 'attention1_conv[0][0]']
max_pool1 (MaxPooling2D) (None, 50, 50, 16) 0 ['attention1_multiply[0][0]']
dropout1 (Dropout) (None, 50, 50, 16) 0 ['max_pool1[0][0]']
conv2 (Conv2D) (None, 50, 50, 32) 4640 ['dropout1[0][0]']
activation2 (Activation) (None, 50, 50, 32) 0 ['conv2[0][0]']
attention2_conv (Conv2D) (None, 50, 50, 1) 33 ['activation2[0][0]']
attention2_multiply (Multiply) (None, 50, 50, 32) 0 ['activation2[0][0]',
'attention2_conv[0][0]']
max_pool2 (MaxPooling2D) (None, 25, 25, 32) 0 ['attention2_multiply[0][0]']
dropout2 (Dropout) (None, 25, 25, 32) 0 ['max_pool2[0][0]']
flatten (Flatten) (None, 20000) 0 ['dropout2[0][0]']
age_input (InputLayer) [(None, 1)] 0 []
timer_pos_input (InputLayer) [(None, 1)] 0 []
concat (Concatenate) (None, 20002) 0 ['flatten[0][0]',
'age_input[0][0]',
'timer_pos_input[0][0]']
dense_fc (Dense) (None, 128) 2560384 ['concat[0][0]']
dropout_fc (Dropout) (None, 128) 0 ['dense_fc[0][0]']
output (Dense) (None, 2) 258 ['dropout_fc[0][0]']
==================================================================================================
Total params: 2,565,524
Trainable params: 258
Non-trainable params: 2,565,266
__________________________________________________________________________________________________
Construction of New Continuous Score Model¶
The following operations are performed here:
- Extracting the penultimate layer (
dense_fc) from the original classifier. - Adding a new dense layer (initially with 2 units and linear activation) to produce a combined output.
- Transferring weights from the corresponding layers of the original model.
- Replacing the final output with a logits layer that reuses the original output layer’s weights.
- Finally, printing the model summary to confirm the architecture
In [3]:
#Construct A Continuous Score Model
dense_fc_layer = model.get_layer('dense_fc').output
new_output = Dense(2, activation= None, name='combined_output')(dense_fc_layer)
continuous_score_model = Model(inputs=model.input, outputs=new_output)
for layer in continuous_score_model.layers:
if layer.name in [l.name for l in model.layers] and 'output' not in layer.name:
try:
original_layer = model.get_layer(layer.name)
original_weights = original_layer.get_weights()
continuous_score_model.get_layer(layer.name).set_weights(original_weights)
except Exception as e:
print(f"Could not transfer weights for layer: {layer.name}. Error: {e}")
penultimate_output = model.get_layer('dense_fc').output
original_output_layer = model.get_layer('output') #the output layer
logits_output = tf.keras.layers.Dense(
original_output_layer.units,
activation=None,
name='logits_output',
weights=original_output_layer.get_weights() # reuse weights and biases from the original output layer
)(penultimate_output)
continuous_score_model = Model(inputs=model.input, outputs=logits_output)
continuous_score_model.summary()
Model: "model_1"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
image_input (InputLayer) [(None, 100, 100, 1 0 []
)]
conv1 (Conv2D) (None, 100, 100, 16 160 ['image_input[0][0]']
)
inst_norm1 (InstanceNormalizat (None, 100, 100, 16 32 ['conv1[0][0]']
ion) )
activation1 (Activation) (None, 100, 100, 16 0 ['inst_norm1[0][0]']
)
attention1_conv (Conv2D) (None, 100, 100, 1) 17 ['activation1[0][0]']
attention1_multiply (Multiply) (None, 100, 100, 16 0 ['activation1[0][0]',
) 'attention1_conv[0][0]']
max_pool1 (MaxPooling2D) (None, 50, 50, 16) 0 ['attention1_multiply[0][0]']
dropout1 (Dropout) (None, 50, 50, 16) 0 ['max_pool1[0][0]']
conv2 (Conv2D) (None, 50, 50, 32) 4640 ['dropout1[0][0]']
activation2 (Activation) (None, 50, 50, 32) 0 ['conv2[0][0]']
attention2_conv (Conv2D) (None, 50, 50, 1) 33 ['activation2[0][0]']
attention2_multiply (Multiply) (None, 50, 50, 32) 0 ['activation2[0][0]',
'attention2_conv[0][0]']
max_pool2 (MaxPooling2D) (None, 25, 25, 32) 0 ['attention2_multiply[0][0]']
dropout2 (Dropout) (None, 25, 25, 32) 0 ['max_pool2[0][0]']
flatten (Flatten) (None, 20000) 0 ['dropout2[0][0]']
age_input (InputLayer) [(None, 1)] 0 []
timer_pos_input (InputLayer) [(None, 1)] 0 []
concat (Concatenate) (None, 20002) 0 ['flatten[0][0]',
'age_input[0][0]',
'timer_pos_input[0][0]']
dense_fc (Dense) (None, 128) 2560384 ['concat[0][0]']
logits_output (Dense) (None, 2) 258 ['dense_fc[0][0]']
==================================================================================================
Total params: 2,565,524
Trainable params: 258
Non-trainable params: 2,565,266
__________________________________________________________________________________________________
Analyse Test Data¶
Next, load independent test data (images, labels, age, and timer-positive values) from .npy and CSV files.
- Timer data is scaled, and age data is processed using
log_scale_age. - Labels are converted into organ names ("Spleen" or "Thymus"), then encoded into integer and one-hot formats.
- The number of samples per class is printed, and separate subsets for spleen and thymus are created, along with their corresponding ages.
- Visualisation
- Regression Analysis
In [15]:
#Import Independent Test Data
base_export_dir = 'Continious_Score_Model_Results'
os.makedirs(base_export_dir, exist_ok=True)
all_images = np.load('test_data/sample_images.npy')
all_labels = np.load('test_data/sample_labels.npy')
age_data = pd.read_csv('test_data/sampledef_age.csv')
timer_pos_data = pd.read_csv('test_data/timer_pos.csv')
timer_pos_data = timer_pos_data['timer_positive'].values
all_ages = age_data['age'].values
all_timers_scaled = timer_pos_data/100
age_scaled = log_scale_age(age_data['age'], 400)
new_labels = []
for i in range(len(all_labels)):
organ = "Spleen" if all_labels[i,0] == 1 else "Thymus"
new_labels.append(organ)
new_labels = np.array(new_labels)
encoder = LabelEncoder()
integer_labels = encoder.fit_transform(new_labels)
one_hot_labels = to_categorical(integer_labels, num_classes=2)
class_indices = {class_name: np.where(integer_labels == i)[0] for i, class_name in enumerate(encoder.classes_)}
age_scaled = age_scaled.values.reshape(-1, 1)
all_timers_scaled = all_timers_scaled.reshape(-1, 1)
spleen_indices = np.where(all_labels[:, 0] == 1)[0]
thymus_indices = np.where(all_labels[:, 0] == 0)[0]
spleen_images = all_images[spleen_indices]
thymus_images = all_images[thymus_indices]
spleen_labels = all_labels[spleen_indices]
thymus_labels = all_labels[thymus_indices]
age_vec = age_data['age']
age_data_spleen = age_vec[spleen_indices]
age_data_thymus = age_vec[thymus_indices]
continuous_scores = continuous_score_model.predict([all_images, age_scaled, all_timers_scaled])
continuous_scores = np.array(continuous_scores[:,0]).flatten() * (-1)
spleen_scores = continuous_scores[spleen_indices]
thymus_scores = continuous_scores[thymus_indices]
spleen_scores = np.array(spleen_scores).flatten()
thymus_scores = np.array(thymus_scores).flatten()
# --- Data Preparation ---
df_spleen = pd.DataFrame({
'Age': age_data_spleen,
'Score': spleen_scores,
'Class': 'Spleen'
})
df_thymus = pd.DataFrame({
'Age': age_data_thymus,
'Score': thymus_scores,
'Class': 'Thymus'
})
df_combined = pd.concat([df_spleen, df_thymus], ignore_index=True)
df_combined['Logged Age'] = np.log2(df_combined['Age'])
# --- Plotting Setup ---
fig, axes = plt.subplots(1, 2, figsize=(10, 5), sharey=False)
label_font = 22
title_font = 22
# Use all existing ages for vertical grid lines
unique_ages = np.array([1,2, 3,4, 7, 14, 21, 35,70,140 ])#np.unique(df_combined['Age'])
logged_unique_ages = np.log2(unique_ages)
# Panel 0: Original Age vs Score
sns.scatterplot(
data=df_combined, x='Age', y='Score', hue='Class',
palette={'Spleen': 'red', 'Thymus': 'blue'},
alpha=0.4, edgecolor=None, ax=axes[0], s=100
)
axes[0].set_xlabel("Age", fontsize=label_font)
axes[0].set_ylabel("Thymus-Spleen Model Score", fontsize=label_font)
axes[0].legend(title='Class', fontsize=18, title_fontsize=18, loc='upper right')
axes[0].set_xticks(unique_ages)
axes[0].grid(True, which='both', linestyle='--', linewidth=0.5, color='gray', alpha=0.7)
axes[0].tick_params(axis='y', labelsize=16)
axes[0].tick_params(axis='x', labelsize=16)
# Panel 2: Logged Age vs Score (Custom Ticks)
sns.scatterplot(
data=df_combined, x='Logged Age', y='Score', hue='Class',
palette={'Spleen': 'red', 'Thymus': 'blue'},
alpha=0.4, edgecolor=None, ax=axes[1], s=100,
legend=False
)
axes[1].set_xlabel("Age (log2)", fontsize=label_font)
axes[1].set_ylabel("Thymus-Spleen Model Score", fontsize=label_font)
axes[1].set_xticks(logged_unique_ages)
axes[1].set_xticklabels(unique_ages.astype(int), fontsize = 16)
axes[1].tick_params(axis='y', labelsize=16) # Set the font size for y-tick labels to 16
axes[1].grid(True, which='both', linestyle='--', linewidth=0.5, color='gray', alpha=0.7)
plt.tight_layout()
os.makedirs(base_export_dir, exist_ok=True)
pdf_path = os.path.join(base_export_dir, 'TestData_class_plot.pdf')
with PdfPages(pdf_path) as pdf:
pdf.savefig(fig)
print(f"Saved the combined plot to {pdf_path}")
csv_path = os.path.join(base_export_dir, 'TestData_class_plot.csv')
df_combined.to_csv(csv_path, index=False)
print(f"Saved the combined DataFrame to {csv_path}")
plt.show()
2/2 [==============================] - 0s 7ms/step Saved the combined plot to Continious_Score_Model_Results/TestData_class_plot.pdf Saved the combined DataFrame to Continious_Score_Model_Results/TestData_class_plot.csv
In [16]:
# Regression
# Compute continuous scores
continuous_scores = continuous_score_model.predict([all_images, age_scaled, all_timers_scaled])
continuous_scores = np.array(continuous_scores[:, 0]).flatten() * (-1)
spleen_scores = continuous_scores[spleen_indices]
thymus_scores = continuous_scores[thymus_indices]
spleen_scores = np.array(spleen_scores).flatten()
thymus_scores = np.array(thymus_scores).flatten()
# --- Data Preparation ---
df_spleen = pd.DataFrame({
'Age': age_data_spleen,
'Score': spleen_scores,
'Class': 'Spleen'
})
df_thymus = pd.DataFrame({
'Age': age_data_thymus,
'Score': thymus_scores,
'Class': 'Thymus'
})
df_combined = pd.concat([df_spleen, df_thymus], ignore_index=True)
df_combined['Logged Age'] = np.log2(df_combined['Age'])
# --- Quadratic Regression Setup ---
def fit_quadratic_model(X, y):
model = Pipeline([
('poly', PolynomialFeatures(degree=2)),
('linear', LinearRegression())
])
model.fit(X, y)
return model
min_age, max_age = df_combined['Age'].min(), df_combined['Age'].max()
min_logged, max_logged = df_combined['Logged Age'].min(), df_combined['Logged Age'].max()
age_grid = np.linspace(min_age, max_age, 100).reshape(-1, 1)
logged_age_grid = np.linspace(min_logged, max_logged, 100).reshape(-1, 1)
age_grid_df = pd.DataFrame(age_grid, columns=['Age'])
logged_age_grid_df = pd.DataFrame(logged_age_grid, columns=['Logged Age'])
models_age = {}
models_logged = {}
for cl in ['Spleen', 'Thymus']:
df_cl = df_combined[df_combined['Class'] == cl]
models_age[cl] = fit_quadratic_model(df_cl[['Age']], df_cl['Score'])
models_logged[cl] = fit_quadratic_model(df_cl[['Logged Age']], df_cl['Score'])
predictions_age = {cl: models_age[cl].predict(age_grid_df) for cl in ['Spleen', 'Thymus']}
predictions_logged = {cl: models_logged[cl].predict(logged_age_grid_df) for cl in ['Spleen', 'Thymus']}
fig, axes = plt.subplots(1, 2, figsize=(10, 5), sharey=False)
label_font = 22
unique_ages = np.array([1, 2, 3, 4, 7, 14, 21, 35, 70, 140])
logged_unique_ages = np.log2(unique_ages)
# Panel 0: Raw Age vs Score with Quadratic Regression Lines
sns.scatterplot(
data=df_combined, x='Age', y='Score', hue='Class',
palette={'Spleen': 'red', 'Thymus': 'blue'},
alpha=0.4, edgecolor=None, ax=axes[0], s=100, legend = None
)
for cl, color in zip(['Spleen', 'Thymus'], ['red', 'blue']):
axes[0].plot(age_grid, predictions_age[cl], color=color, lw=2, label=f'{cl} Quad Fit')
axes[0].set_xlabel("Age", fontsize=label_font)
axes[0].set_ylabel("Thymus-Spleen Model Score", fontsize=label_font)
axes[0].set_xticks(unique_ages)
axes[0].grid(True, which='both', linestyle='--', linewidth=0.5, color='gray', alpha=0.7)
axes[0].tick_params(axis='x', labelsize=16)
axes[0].tick_params(axis='y', labelsize=16)
# Panel 1: Logged Age vs Score with Quadratic Regression Lines
sns.scatterplot(
data=df_combined, x='Logged Age', y='Score', hue='Class',
palette={'Spleen': 'red', 'Thymus': 'blue'},
alpha=0.4, edgecolor=None, ax=axes[1], s=100,
legend=False
)
for cl, color in zip(['Spleen', 'Thymus'], ['red', 'blue']):
axes[1].plot(logged_age_grid, predictions_logged[cl], color=color, lw=2)
axes[1].set_ylabel("Thymus-Spleen Model Score", fontsize=label_font)
axes[1].set_xlabel("Age (log2)", fontsize=label_font)
axes[1].set_xticks(logged_unique_ages)
axes[1].set_xticklabels(unique_ages.astype(int), fontsize=16)
axes[1].grid(True, which='both', linestyle='--', linewidth=0.5, color='gray', alpha=0.7)
axes[1].tick_params(axis='y', labelsize=16)
plt.tight_layout()
os.makedirs(base_export_dir, exist_ok=True)
pdf_path = os.path.join(base_export_dir, 'TestData_class_plot_with_regression.pdf')
with PdfPages(pdf_path) as pdf:
pdf.savefig(fig)
print(f"Saved the combined plot to {pdf_path}")
csv_path = os.path.join(base_export_dir, 'TestData_class_plot_with_regression.csv')
df_combined.to_csv(csv_path, index=False)
print(f"Saved the combined DataFrame to {csv_path}")
# --- Model Parameters Export ---
parameters = []
for model_type, model_dict, col_name in [('Age', models_age, 'Age'), ('Logged Age', models_logged, 'Logged Age')]:
for cl, model in model_dict.items():
poly = model.named_steps['poly']
linear = model.named_steps['linear']
feature_names = poly.get_feature_names_out([col_name])
intercept = linear.intercept_
parameters.append({
'Class': cl,
'Model_Type': model_type,
'Feature': 'Intercept',
'Coefficient': intercept
})
for fname, coef in zip(feature_names, linear.coef_):
parameters.append({
'Class': cl,
'Model_Type': model_type,
'Feature': fname,
'Coefficient': coef
})
params_df = pd.DataFrame(parameters)
params_csv_path = os.path.join(base_export_dir, 'quadratic_model_parameters.csv')
params_df.to_csv(params_csv_path, index=False)
plt.show()
2/2 [==============================] - 0s 11ms/step Saved the combined plot to Continious_Score_Model_Results/TestData_class_plot_with_regression.pdf Saved the combined DataFrame to Continious_Score_Model_Results/TestData_class_plot_with_regression.csv
In [6]:
# Thymus Data vs Age
X = sm.add_constant(np.column_stack((df_thymus['Age'], df_thymus['Age']**2)))
y = df_thymus['Score']
q_model = sm.OLS(y, X).fit()
print(q_model.summary())
OLS Regression Results
==============================================================================
Dep. Variable: Score R-squared: 0.949
Model: OLS Adj. R-squared: 0.943
Method: Least Squares F-statistic: 167.4
Date: Sun, 16 Feb 2025 Prob (F-statistic): 2.34e-12
Time: 16:34:44 Log-Likelihood: -48.107
No. Observations: 21 AIC: 102.2
Df Residuals: 18 BIC: 105.3
Df Model: 2
Covariance Type: nonrobust
==============================================================================
coef std err t P>|t| [0.025 0.975]
------------------------------------------------------------------------------
const 19.3294 0.803 24.061 0.000 17.642 21.017
x1 -0.1965 0.055 -3.548 0.002 -0.313 -0.080
x2 -0.0002 0.000 -0.621 0.542 -0.001 0.001
==============================================================================
Omnibus: 3.509 Durbin-Watson: 1.581
Prob(Omnibus): 0.173 Jarque-Bera (JB): 2.678
Skew: -0.866 Prob(JB): 0.262
Kurtosis: 2.759 Cond. No. 1.07e+04
==============================================================================
Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.
[2] The condition number is large, 1.07e+04. This might indicate that there are
strong multicollinearity or other numerical problems.
In [7]:
# Thymus Data vs Log2(Age)
df_thymus['Logged Age'] = np.log2(df_thymus['Age'])
X = sm.add_constant(np.column_stack((df_thymus['Logged Age'], df_thymus['Logged Age']**2)))
y = df_thymus['Score']
q_model = sm.OLS(y, X).fit()
print(q_model.summary())
OLS Regression Results
==============================================================================
Dep. Variable: Score R-squared: 0.946
Model: OLS Adj. R-squared: 0.940
Method: Least Squares F-statistic: 158.8
Date: Sun, 16 Feb 2025 Prob (F-statistic): 3.68e-12
Time: 16:34:44 Log-Likelihood: -48.634
No. Observations: 21 AIC: 103.3
Df Residuals: 18 BIC: 106.4
Df Model: 2
Covariance Type: nonrobust
==============================================================================
coef std err t P>|t| [0.025 0.975]
------------------------------------------------------------------------------
const 12.8814 1.592 8.093 0.000 9.537 16.225
x1 6.4773 0.982 6.597 0.000 4.415 8.540
x2 -1.3974 0.126 -11.080 0.000 -1.662 -1.132
==============================================================================
Omnibus: 3.893 Durbin-Watson: 1.817
Prob(Omnibus): 0.143 Jarque-Bera (JB): 1.943
Skew: -0.540 Prob(JB): 0.378
Kurtosis: 4.026 Cond. No. 68.4
==============================================================================
Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.
In [12]:
# Spleen Data vs Age
X = sm.add_constant(np.column_stack((df_spleen['Age'], df_spleen['Age']**2)))
y = df_spleen['Score']
q_model = sm.OLS(y, X).fit()
print(q_model.summary())
OLS Regression Results
==============================================================================
Dep. Variable: Score R-squared: 0.741
Model: OLS Adj. R-squared: 0.706
Method: Least Squares F-statistic: 21.45
Date: Sun, 16 Feb 2025 Prob (F-statistic): 3.99e-05
Time: 16:37:51 Log-Likelihood: -50.595
No. Observations: 18 AIC: 107.2
Df Residuals: 15 BIC: 109.9
Df Model: 2
Covariance Type: nonrobust
==============================================================================
coef std err t P>|t| [0.025 0.975]
------------------------------------------------------------------------------
const 2.5364 1.575 1.611 0.128 -0.820 5.892
x1 -0.4839 0.099 -4.864 0.000 -0.696 -0.272
x2 0.0024 0.001 3.773 0.002 0.001 0.004
==============================================================================
Omnibus: 1.533 Durbin-Watson: 0.712
Prob(Omnibus): 0.465 Jarque-Bera (JB): 1.296
Skew: -0.570 Prob(JB): 0.523
Kurtosis: 2.346 Cond. No. 1.23e+04
==============================================================================
Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.
[2] The condition number is large, 1.23e+04. This might indicate that there are
strong multicollinearity or other numerical problems.
In [13]:
# Spleen Data vs Log2(Age)
df_spleen['Logged Age'] = np.log2(df_spleen['Age'])
X = sm.add_constant(np.column_stack((df_spleen['Logged Age'], df_spleen['Logged Age']**2)))
y = df_spleen['Score']
q_model = sm.OLS(y, X).fit()
print(q_model.summary())
OLS Regression Results
==============================================================================
Dep. Variable: Score R-squared: 0.920
Model: OLS Adj. R-squared: 0.909
Method: Least Squares F-statistic: 85.89
Date: Sun, 16 Feb 2025 Prob (F-statistic): 6.10e-09
Time: 16:38:00 Log-Likelihood: -40.053
No. Observations: 18 AIC: 86.11
Df Residuals: 15 BIC: 88.78
Df Model: 2
Covariance Type: nonrobust
==============================================================================
coef std err t P>|t| [0.025 0.975]
------------------------------------------------------------------------------
const 18.1497 2.847 6.376 0.000 12.082 24.217
x1 -9.3239 1.581 -5.898 0.000 -12.693 -5.954
x2 0.6413 0.183 3.506 0.003 0.251 1.031
==============================================================================
Omnibus: 2.639 Durbin-Watson: 1.908
Prob(Omnibus): 0.267 Jarque-Bera (JB): 1.270
Skew: -0.636 Prob(JB): 0.530
Kurtosis: 3.273 Cond. No. 131.
==============================================================================
Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.
In [14]:
continuous_score_model.save('Foxp3DevAge_Logit.keras')
WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.
In [ ]: