OPEN-SOURCE SCRIPT
AetherEdge NeuraNet Trend Predictor
🖊️ Overview
AetherEdge NeuraNet Trend Predictor is a next-generation AI-driven trend forecasting indicator that fully implements a true Multi-Layer Perceptron (MLP) neural network in Pine Script. Nine-dimensional features flow through input → hidden → output layers, with weights self-updating every bar via backpropagation + Nesterov momentum + L2 regularization. This is the pinnacle of machine learning engines on TradingView—a system that perpetually evolves with market structure changes.
🔶 Key Features
True MLP Implementation: 9-H-1 architecture (hidden 3–12 neurons)
Online Learning: Deterministic gradient descent updating weights every bar
Nesterov-Style Momentum: Accelerated convergence via gradient inertia
L2 Regularization: Overfitting suppression
Xavier Initialization: Theoretically optimal weight init
Binary Cross-Entropy Loss: Optimal for probabilistic prediction
9-D Feature Engine: Price/EMA/RSI/ATR/Volume/MACD/BB/DMI/Structure
Robust Normalization: z-score based statistical standardization
Live Accuracy Tracking: Cumulative prediction accuracy %
Loss EMA Monitoring: Visualized learning progress
Network Visualization: Real-time input → hidden activation → output
Probability Bar (right edge): 10-tier prediction visualization
Confidence Computation: Auto-calculated via |prob - 0.5| × 2
AI VERDICT: 5-tier decision support messaging
🧠 Technical Architecture
This indicator is an advanced neural network implementation grounded in genuine deep learning theory.
Network Structure:
Input layer: 9 neurons (feature vector)
Hidden layer: 3–12 neurons (tanh activation)
Output layer: 1 neuron (sigmoid, probability)
Total params: 9×H + H + H + 1
Feature Engineering (9-D):
f1: Price/EMA Fast deviation
f2: EMA Spread (Fast vs Slow)
f3: RSI normalized
f4: ATR% z-score
f5: Volume z-score (log-transformed)
f6: MACD Histogram normalized
f7: Bollinger Band Position
f8: DMI differential
f9: Structural position (high/low range)
Forward Pass:
h_j = tanh(Σ x_i × W1_ij + b1_j)
prob = sigmoid(Σ h_j × W2_j + b2)
Backpropagation:
Output error: err = ŷ - y (BCE gradient)
Hidden error: δ = err × W2 × (1 - h²) (tanh derivative)
L2-regularized gradient: g = δ × x + λ × W
Nesterov update: v = β×v - η×g; W += v
Target Signal: Next bar direction (close > close[1])
Warmup: Signals activate after 200 bars of training
Learning Statistics: Train Steps / Accuracy / Loss EMA / 100-bar Loss
⚙️ Recommended Settings & Tuning Guide
Crypto Defaults:
BTC (1H–4H): Standard (Hidden=6, η=0.05, β=0.9)
ETH (15M–1H): Hidden=8, η=0.03 for precision
SOL (high-vol): λ=0.005 to enhance overfitting suppression
XRP (range): Hidden=4, η=0.07 for rapid adaptation
Learning Rate (η):
0.01–0.03: Stability-focused, long-term trends
0.05: Balanced (recommended)
0.1–0.2: Fast adaptation for regime shifts
Hidden Neurons:
3–4: Simple, low overfit risk
6: Balanced (recommended)
8–12: High capacity, requires sufficient data
Warmup Period:
200 bars: Standard
500 bars: High-precision focus
100 bars: Quick activation
Thresholds:
Conservative: Long=0.75, Short=0.25, MinConf=0.7
Standard: Long=0.65, Short=0.35, MinConf=0.6
Aggressive: Long=0.6, Short=0.4, MinConf=0.5
💡 How to Use in Practice
VERDICT = STRONG LONG CONVICTION: Neural net at >80% conviction—strongest entry
Accuracy > 55%: Network in hot streak, high signal trust
Accuracy < 50%: Regime shift in progress, stand aside
Loss EMA declining: Learning healthy, precision improving
Confidence > 70%: High-quality signal, full position consideration
Feature panel watch: All features unidirectional = multi-faceted consensus
Hidden activations: ●●● (strong) aligned = internal network consensus
Probability bar extreme (>90%): Watch for reversal risk
AetherEdge Synergy:
All-in-One Dashboard HIGH-CONVICTION + NeuraNet STRONG = dual AI confirmation
Self-Evolving S/R Grid strong line + NeuraNet aligned = ML × RL fusion
Liquidity Sweep Bull Sweep + NeuraNet > 0.8 = institutional + AI sync
Multi-TF: 4H for direction → 1H for NeuraNet wait → 15M for precision entry
⚠️ Important Notes
Learning Reset: Weights initialize on every chart reload—warmup mandatory
Initial Warmup: Minimum 200 bars required before predictions
Accuracy Metric: Cumulative; treat <100 samples as preliminary
Overfit Risk: Excess Hidden + low λ = overfitting—balance carefully
Diverging η: η > 0.2 risks weight explosion—use cautiously
Regime Changes: Sudden market shifts temporarily reduce Accuracy, auto-recovers
Compute Load: Hidden=12 is heavy—use 6 on low-spec setups
🚨 Disclaimer
This indicator is a machine learning demonstration provided for educational and research purposes only and does not constitute financial advice. Neural network predictions are statistical estimates based on historical patterns and do not guarantee future profits. Do not blindly trust AI model outputs—combine with independent analysis and proper risk management.
AetherEdge NeuraNet Trend Predictor is a next-generation AI-driven trend forecasting indicator that fully implements a true Multi-Layer Perceptron (MLP) neural network in Pine Script. Nine-dimensional features flow through input → hidden → output layers, with weights self-updating every bar via backpropagation + Nesterov momentum + L2 regularization. This is the pinnacle of machine learning engines on TradingView—a system that perpetually evolves with market structure changes.
🔶 Key Features
True MLP Implementation: 9-H-1 architecture (hidden 3–12 neurons)
Online Learning: Deterministic gradient descent updating weights every bar
Nesterov-Style Momentum: Accelerated convergence via gradient inertia
L2 Regularization: Overfitting suppression
Xavier Initialization: Theoretically optimal weight init
Binary Cross-Entropy Loss: Optimal for probabilistic prediction
9-D Feature Engine: Price/EMA/RSI/ATR/Volume/MACD/BB/DMI/Structure
Robust Normalization: z-score based statistical standardization
Live Accuracy Tracking: Cumulative prediction accuracy %
Loss EMA Monitoring: Visualized learning progress
Network Visualization: Real-time input → hidden activation → output
Probability Bar (right edge): 10-tier prediction visualization
Confidence Computation: Auto-calculated via |prob - 0.5| × 2
AI VERDICT: 5-tier decision support messaging
🧠 Technical Architecture
This indicator is an advanced neural network implementation grounded in genuine deep learning theory.
Network Structure:
Input layer: 9 neurons (feature vector)
Hidden layer: 3–12 neurons (tanh activation)
Output layer: 1 neuron (sigmoid, probability)
Total params: 9×H + H + H + 1
Feature Engineering (9-D):
f1: Price/EMA Fast deviation
f2: EMA Spread (Fast vs Slow)
f3: RSI normalized
f4: ATR% z-score
f5: Volume z-score (log-transformed)
f6: MACD Histogram normalized
f7: Bollinger Band Position
f8: DMI differential
f9: Structural position (high/low range)
Forward Pass:
h_j = tanh(Σ x_i × W1_ij + b1_j)
prob = sigmoid(Σ h_j × W2_j + b2)
Backpropagation:
Output error: err = ŷ - y (BCE gradient)
Hidden error: δ = err × W2 × (1 - h²) (tanh derivative)
L2-regularized gradient: g = δ × x + λ × W
Nesterov update: v = β×v - η×g; W += v
Target Signal: Next bar direction (close > close[1])
Warmup: Signals activate after 200 bars of training
Learning Statistics: Train Steps / Accuracy / Loss EMA / 100-bar Loss
⚙️ Recommended Settings & Tuning Guide
Crypto Defaults:
BTC (1H–4H): Standard (Hidden=6, η=0.05, β=0.9)
ETH (15M–1H): Hidden=8, η=0.03 for precision
SOL (high-vol): λ=0.005 to enhance overfitting suppression
XRP (range): Hidden=4, η=0.07 for rapid adaptation
Learning Rate (η):
0.01–0.03: Stability-focused, long-term trends
0.05: Balanced (recommended)
0.1–0.2: Fast adaptation for regime shifts
Hidden Neurons:
3–4: Simple, low overfit risk
6: Balanced (recommended)
8–12: High capacity, requires sufficient data
Warmup Period:
200 bars: Standard
500 bars: High-precision focus
100 bars: Quick activation
Thresholds:
Conservative: Long=0.75, Short=0.25, MinConf=0.7
Standard: Long=0.65, Short=0.35, MinConf=0.6
Aggressive: Long=0.6, Short=0.4, MinConf=0.5
💡 How to Use in Practice
VERDICT = STRONG LONG CONVICTION: Neural net at >80% conviction—strongest entry
Accuracy > 55%: Network in hot streak, high signal trust
Accuracy < 50%: Regime shift in progress, stand aside
Loss EMA declining: Learning healthy, precision improving
Confidence > 70%: High-quality signal, full position consideration
Feature panel watch: All features unidirectional = multi-faceted consensus
Hidden activations: ●●● (strong) aligned = internal network consensus
Probability bar extreme (>90%): Watch for reversal risk
AetherEdge Synergy:
All-in-One Dashboard HIGH-CONVICTION + NeuraNet STRONG = dual AI confirmation
Self-Evolving S/R Grid strong line + NeuraNet aligned = ML × RL fusion
Liquidity Sweep Bull Sweep + NeuraNet > 0.8 = institutional + AI sync
Multi-TF: 4H for direction → 1H for NeuraNet wait → 15M for precision entry
⚠️ Important Notes
Learning Reset: Weights initialize on every chart reload—warmup mandatory
Initial Warmup: Minimum 200 bars required before predictions
Accuracy Metric: Cumulative; treat <100 samples as preliminary
Overfit Risk: Excess Hidden + low λ = overfitting—balance carefully
Diverging η: η > 0.2 risks weight explosion—use cautiously
Regime Changes: Sudden market shifts temporarily reduce Accuracy, auto-recovers
Compute Load: Hidden=12 is heavy—use 6 on low-spec setups
🚨 Disclaimer
This indicator is a machine learning demonstration provided for educational and research purposes only and does not constitute financial advice. Neural network predictions are statistical estimates based on historical patterns and do not guarantee future profits. Do not blindly trust AI model outputs—combine with independent analysis and proper risk management.
Mã nguồn mở
Theo đúng tinh thần TradingView, tác giả của tập lệnh này đã công bố nó dưới dạng mã nguồn mở, để các nhà giao dịch có thể xem xét và xác minh chức năng. Chúc mừng tác giả! Mặc dù bạn có thể sử dụng miễn phí, hãy nhớ rằng việc công bố lại mã phải tuân theo Nội quy.
Thông báo miễn trừ trách nhiệm
Thông tin và các ấn phẩm này không nhằm mục đích, và không cấu thành, lời khuyên hoặc khuyến nghị về tài chính, đầu tư, giao dịch hay các loại khác do TradingView cung cấp hoặc xác nhận. Đọc thêm tại Điều khoản Sử dụng.
Mã nguồn mở
Theo đúng tinh thần TradingView, tác giả của tập lệnh này đã công bố nó dưới dạng mã nguồn mở, để các nhà giao dịch có thể xem xét và xác minh chức năng. Chúc mừng tác giả! Mặc dù bạn có thể sử dụng miễn phí, hãy nhớ rằng việc công bố lại mã phải tuân theo Nội quy.
Thông báo miễn trừ trách nhiệm
Thông tin và các ấn phẩm này không nhằm mục đích, và không cấu thành, lời khuyên hoặc khuyến nghị về tài chính, đầu tư, giao dịch hay các loại khác do TradingView cung cấp hoặc xác nhận. Đọc thêm tại Điều khoản Sử dụng.