MachineLearning

Library "MachineLearning"

Quantum-TA • Machine – Adaptive ML Toolkit for Pine v6

Bring modern data-science techniques to any TradingView script without external servers or heavy tensors.
This library blends low-lag filtering, regime detection, information-theory gauges …and two tiny inference-only models – a KAN (Kolmogorov-Arnold Network) and a lite Temporal-Fusion Transformer (TFT) – then lets a self-training ensemble decide which one to trust bar-by-bar.

clamp(value, minVal, maxVal)
  Parameters:
    value (float)
    minVal (float)
    maxVal (float)

q_log(x)
  Parameters:
    x (float)

tanh(x)
  Parameters:
    x (float)

fisher_volatility(src, len)
  Parameters:
    src (float)
    len (simple int)

ema(src, len)
  Parameters:
    src (float)
    len (int)

normalizeArray(arr)
  Parameters:
    arr (array<float>)

hmm_volatility_regime(atr_current)
  Parameters:
    atr_current (float)

tft_model(inputVector, len, learningRate, regime_probs)
  Parameters:
    inputVector (array<float>)
    len (int)
    learningRate (float)
    regime_probs (array<float>)

normalizeWeights(w1, w2)
  Parameters:
    w1 (float)
    w2 (float)

final_prediction(kan_pred, attn_pred, w_kan, w_attn)
  Parameters:
    kan_pred (float)
    attn_pred (float)
    w_kan (float)
    w_attn (float)

ensemble_weight_predictor(target_weight, kan_err, tft_err, atr_norm, regime_probs)
  Parameters:
    target_weight (float)
    kan_err (float)
    tft_err (float)
    atr_norm (float)
    regime_probs (array<float>)

ensemble_weights(kan_err, tft_err, atr, regime_probs)
  Parameters:
    kan_err (float)
    tft_err (float)
    atr (float)
    regime_probs (array<float>)

render(source)
  Parameters:
    source (float)