Architecture Overview

STRATA has two parallel architectures that share the same design philosophy:

StrataNet (v2.5+) — Neural Network

Input: OHLCV window  (B, T, 5)
         │
         ▼
StrataEmbedding      learned OHLCV → semantic features
         │            Linear(5 → embed_dim) + LayerNorm + GELU
         ▼
StrataCoreCell  ×T   custom GRU-like recurrent cell
         │            update gate z, reset gate r
         │            candidate projected back to 4-dim
         │            bounds enforced per step
         ▼
StrataHead           hidden state → outputs
         │            Linear(4 → head_dim) + GELU
         ├──────────► action logits  (3: LONG/SHORT/HOLD)
         ├──────────► confidence     (1: sigmoid)
         └──────────► regime logits  (4: TRENDING/RANGING/TRANSITIONING/CHOPPY)

Classic State Machine (v1.0+) — Rule-Based

Raw OHLCV
    │
    ▼
┌─────────┐
│  SENSE  │  OHLCV → {trend, vol, liquidity_above, break_structure}
└────┬────┘
     │
     ▼
┌─────────┐     ┌──────────┐
│  CORE   │◄────│  MEMORY  │  pattern bank (Short/Mid/Long layers)
└────┬────┘     └──────────┘
     │
     ▼
┌─────────┐
│ DECIDE  │  state → {action, confidence, risk, regime}
└────┬────┘
     │
     ▼
┌─────────┐
│  GUARD  │  hard-rule override (NOT learned — explicit, auditable)
└────┬────┘
     │
     ▼
  Final Action
     │
     ▼
┌─────────┐
│  LOOP   │  P&L feedback → weight adaptation (optional)
└─────────┘

Shared Design Principles

Both architectures share the same hidden state design:

Dimension	Range	Meaning
bias	[-1, 1]	Directional conviction
momentum	[0, 1]	Structural energy from breakouts
trap_risk	[0, 1]	Adverse selection / liquidity trap
uncertainty	[0, 1]	Volatility-driven ambiguity

In the classic state machine, these are computed by explicit rules. In StrataNet, these are learned but still constrained to the same bounds.