import altair as alt import pandas as pd import streamlit as st import numpy as np import matplotlib.pyplot as plt import seaborn as sns from sklearn.datasets import make_classification, make_moons, make_circles from sklearn.model_selection import train_test_split from sklearn.preprocessing import StandardScaler from mlxtend.plotting import plot_decision_regions from keras.models import Sequential from keras.layers import Dense from keras.optimizers import SGD # Streamlit page configuration st.set_page_config(page_title="DeepNet Playground", layout="wide", page_icon="🚀") # Apply seaborn styling sns.set_style("whitegrid") # Helper function def create_dataset(dataset_type, n_samples, noise, random_state=42, n_features=2, n_informative=2, n_redundant=0, n_clusters_per_class=1, class_sep=1.0): if dataset_type == 'Classification': X, y = make_classification( n_samples=n_samples, n_features=n_features, n_informative=n_informative, n_redundant=n_redundant, n_clusters_per_class=n_clusters_per_class, class_sep=class_sep, n_classes=2, random_state=random_state ) elif dataset_type == 'Moons': X, y = make_moons(n_samples=n_samples, noise=noise, random_state=random_state) elif dataset_type == 'Circles': X, y = make_circles(n_samples=n_samples, noise=noise, factor=0.5, random_state=random_state) else: X, y = None, None return X, y def build_model(input_dim, layer_configs): model = Sequential() model.add(Dense(layer_configs[0]['neurons'], activation=layer_configs[0]['activation'], input_dim=input_dim)) for config in layer_configs[1:]: model.add(Dense(config['neurons'], activation=config['activation'])) model.add(Dense(1, activation='sigmoid')) return model # Title Section st.markdown("

DeepNet Playground

", unsafe_allow_html=True) st.markdown("

Create your own dataset, build a neural network, train it live and visualize!

", unsafe_allow_html=True) st.markdown("---") # Sidebar with st.sidebar: st.markdown("## 🎛ī¸ Playground Control") if st.button("🔄 Reset Playground"): for key in list(st.session_state.keys()): del st.session_state[key] st.experimental_rerun() st.header("🔧 Configuration Panel") with st.expander("🛠ī¸ Dataset Settings"): dataset_type = st.selectbox("Dataset Type", ["Classification", "Moons", "Circles"]) n_samples = st.slider("Number of Samples", 100, 10000, 300, step=50) noise = st.slider("Noise Level", 0.0, 0.5, 0.2, step=0.01) if dataset_type == 'Classification': st.markdown("#### ⚙ī¸ Advanced Classification Settings") n_features = st.slider("Total Features", 2, 20, 5) n_informative = st.slider("Informative Features", 1, n_features, min(2, n_features)) max_redundant = n_features - n_informative if max_redundant > 0: n_redundant = st.slider("Redundant Features", 0, max_redundant, 0) else: n_redundant = 0 st.info("ℹī¸ No redundant features possible with current settings.") n_clusters_per_class = st.slider("Clusters per Class", 1, 3, 1) class_sep = st.slider("Class Separation", 0.5, 2.0, 1.0, step=0.1) with st.expander("🏗ī¸ Model Architecture"): n_layers = st.slider("Number of Hidden Layers", 1, 5, 2) layer_configs = [] for i in range(n_layers): neurons = st.number_input(f"Neurons in Layer {i+1}", min_value=1, max_value=512, value=8) activation = st.selectbox(f"Activation for Layer {i+1}", ["relu", "tanh", "sigmoid"], key=f"act_{i}") layer_configs.append({"neurons": neurons, "activation": activation}) with st.expander("🚀 Training Settings"): epochs = st.slider("Epochs", 10, 500, 100, step=10) batch_size = st.selectbox("Batch Size", [8, 16, 32, 64], index=2) learning_rate = st.selectbox("Learning Rate", [0.001, 0.01, 0.1], index=1) # Main Content col1, col2 = st.columns([2, 3]) with col1: st.subheader("📊 1. Generate Dataset") if st.button("🎲 Generate Dataset"): with st.spinner('✨ Creating Dataset...'): if dataset_type == "Classification": X, y = create_dataset(dataset_type, n_samples, noise, n_features=n_features, n_informative=n_informative, n_redundant=n_redundant, n_clusters_per_class=n_clusters_per_class, class_sep=class_sep) else: X, y = create_dataset(dataset_type, n_samples, noise) if X is not None: st.success(f"Generated {dataset_type} dataset with {n_samples} samples.") fig, ax = plt.subplots() sns.scatterplot(x=X[:, 0], y=X[:, 1], hue=y, palette="Set2", s=40, ax=ax) ax.set_title(f"{dataset_type} Dataset") st.pyplot(fig) st.session_state.X = X st.session_state.y = y else: st.error("Error generating dataset.") with col2: st.subheader("🧐 2. Build & Train Model") if 'X' in st.session_state and 'y' in st.session_state: if st.button("⚡ Train Model"): with st.spinner('Training the neural net...'): X, y = st.session_state.X, st.session_state.y scaler = StandardScaler() X_scaled = scaler.fit_transform(X) X_train, X_val, y_train, y_val = train_test_split(X_scaled, y, test_size=0.2, random_state=42) model = build_model(input_dim=X_scaled.shape[1], layer_configs=layer_configs) optimizer = SGD(learning_rate=learning_rate) model.compile(optimizer=optimizer, loss='binary_crossentropy', metrics=['accuracy']) history = model.fit(X_train, y_train, epochs=epochs, batch_size=batch_size, verbose=0, validation_data=(X_val, y_val)) st.session_state.model = model st.session_state.history = history st.session_state.scaler = scaler st.session_state.X_train, st.session_state.X_val = X_train, X_val st.session_state.y_train, st.session_state.y_val = y_train, y_val st.success("Model trained successfully! 🎉") st.markdown("---") if 'history' in st.session_state: st.subheader("📈 3. Training Performance") history = st.session_state.history perf_col1, perf_col2 = st.columns(2) with perf_col1: fig_loss, ax_loss = plt.subplots() ax_loss.plot(history.history['loss'], label='Train Loss', color='blue') ax_loss.plot(history.history['val_loss'], label='Validation Loss', color='red') ax_loss.set_title('Loss Curve', fontsize=13) ax_loss.set_xlabel('Epochs') ax_loss.set_ylabel('Loss') ax_loss.legend() st.pyplot(fig_loss) with perf_col2: fig_acc, ax_acc = plt.subplots() ax_acc.plot(history.history['accuracy'], label='Train Accuracy', color='green') ax_acc.plot(history.history['val_accuracy'], label='Validation Accuracy', color='orange') ax_acc.set_title('Accuracy Curve', fontsize=13) ax_acc.set_xlabel('Epochs') ax_acc.set_ylabel('Accuracy') ax_acc.legend() st.pyplot(fig_acc) st.markdown("---") st.subheader("🌈 4. Decision Boundary Visualization") class KerasClassifierWrapper: def __init__(self, model, scaler): self.model = model self.scaler = scaler def predict(self, X): X_scaled = self.scaler.transform(X) pred_probs = self.model.predict(X_scaled) return np.where(pred_probs >= 0.5, 1, 0).flatten() model_wrapper = KerasClassifierWrapper(st.session_state.model, st.session_state.scaler) fig_db, ax_db = plt.subplots(figsize=(8, 6)) plot_decision_regions(X=st.session_state.scaler.inverse_transform( np.vstack((st.session_state.X_train, st.session_state.X_val))), y=np.hstack((st.session_state.y_train, st.session_state.y_val)), clf=model_wrapper, legend=2, ax=ax_db) ax_db.set_title('Decision Boundary', fontsize=14) st.pyplot(fig_db) st.success("đŸŽ¯ You have completed the full cycle: Dataset ➔ Model ➔ Train ➔ Visualize!")