Machine Learning Nodes
Bring trained models into strategy logic: feature engineering, normalization, and inference as composable nodes. Machine Learning is coming soon; it is not available at launch.
Temporal Inference
Utilize LSTM networks for time-series forecasting. Optimized for capturing long-range dependencies in price history without vanishing gradients.
Feature Scaling
Deterministic nodes for Z-Score normalization and Min-Max scaling. Ensures input features are mapped to a unified space for stability.
Regime Classifiers
Unsupervised nodes utilizing Hidden Markov Models (HMM) to autonomously identify shifts in market volatility and trend direction.
Attention Mechanisms
Transformer-based nodes implementing multi-head attention. Identifies relevant historical bars for predicting price displacement.
GPU training
On-demand GPU compute
Training is planned to run on on-demand GPU compute provisioned per job, so you would only pay for what a training run uses. Trained models would then be attached to your strategy as inference nodes.
GPU is planned to be provisioned on demand for each training job.
Models will be trained on your chosen data and stay tied to your account.
Inspect a model’s behaviour in backtest before trusting it. Not investment advice.