Question 1

What models can I train in the ML Studio?

Accepted Answer

Sixteen architectures and counting. Deep learning: LSTM, Transformer, CNN, GAN, Autoencoder. Gradient boosting: XGBoost, CatBoost, LightGBM. Classical ML: Random Forest, SVM. Time series: ARIMA, Prophet, GARCH, Kalman Filter, Wavelet Decomposition. You pick the model, the asset, the timeframe, and the hyperparameters — we handle the rest.

Question 2

Do I need my own GPU?

Accepted Answer

No. Everything runs on our dedicated compute cluster — four NVIDIA RTX GPUs with 16 GB VRAM each. Training is free, queued fairly, and typically completes in minutes. You get a full performance report: accuracy, precision, recall, feature importance rankings, and prediction visualisations.

Question 3

What data do the models train on?

Accepted Answer

Our institutional-grade database covering 173+ assets with years of historical candle data across multiple timeframes — from 1-minute bars up to daily. The Studio handles data extraction, cleaning, feature engineering, and train/test splitting automatically. You just select the asset and the model.