forked from animatedread/Warrior_EA
| Filename | Latest commit message | Latest commit date |
|---|---|---|
70cdec2717
FeedForwardConv emits POSITION-MAJOR output, matrix_o[out + window_out * i],
so one bar's window_out filter responses are contiguous and consecutive bars
sit window_out apart. Both pooling implementations (FeedForwardProof and
CPU_FeedForwardProof) slide FLAT over that buffer - pos = i * step, reducing
`window` CONSECUTIVE elements. On a position-major layout those neighbours
are different FILTERS of the same bar, never one filter across time.
At the shipped 3/2 the pool computed max(bar0_f0, bar0_f1, bar0_f2), then
max(bar0_f2, bar0_f3, bar0_f4), with every 8th window straddling a bar
boundary. So it collapsed unrelated feature detectors into whichever fired
hardest, passed gradient to that winner only, and halved the feature map
while doing it - all below every learnable layer, where nothing above can
recover it. The removed inputs' own labels ("3 Bars") show time-axis pooling
was the intent throughout.
Measured cost: CONV sat pinned at ~40% balanced accuracy for 510 eras with
Sell recall 0%, while plain MLPs on the same data reached 57-61%. HYBRID,
which also carried this stage, came second-worst of the batch-norm group.
Not fixable in the topology: pooling one filter across time needs a stride
of window_out BETWEEN samples within a window, which a consecutive-window
kernel cannot express at any window/step. That needs a stride-aware kernel
in Network.cl + WarriorCPU.cpp + WarriorDML.cpp and a DLL rebuild, and is
only worth doing if a conv front-end earns its place without downsampling
first - with 20 sliding positions there is little to gain by halving them.
ConvPoolWindow/ConvPoolStep and their enums are removed with it, along with
the |CP: fingerprint term added earlier today.
Both builds compile 0 errors, 0 warnings.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
| .. | ||
| README.md | ||
| SignalCONV.mqh | ||
| SignalHYBRID.mqh | ||
| SignalIchimoku.mqh | ||
| SignalITF.mqh | ||
| SignalLSTM.mqh | ||
| SignalMA.mqh | ||
| SignalMACD.mqh | ||
| SignalMarketDepth.mqh | ||
| SignalNewsFilter.mqh | ||
| SignalPAI.mqh | ||
| SignalRiskGuard.mqh | ||
| SignalRSI.mqh | ||
| Signals.mqh | ||
| SignalSessionFilter.mqh | ||
Signals Subsystem (Signals/)
Overview
The Signals/ directory contains all trade signal generation logic for the Warrior EA. It includes both traditional indicator-based signals and advanced AI/ML-driven signals. Each signal is encapsulated in its own class, supporting modularity and extensibility.
Key Components
Signals.mqh
- Main orchestration file for signal modules.
- Includes both traditional and AI/ML signal classes.
- Facilitates integration of filters (news, session, etc.) and advanced signals.
AI/ML-Driven Signals
- SignalLSTM.mqh: Implements an LSTM-based neural network signal generator. Integrates with the AI subsystem, supports model training, loading, and inference. Designed for advanced, data-driven strategies.
- SignalPAI.mqh: Implements a Perceptron AI-based signal generator. Inherits from
CExpertSignalAIBase. Provides methods for initializing, training, and using a perceptron neural network for trade signal generation. Supports dynamic configuration, indicator integration, and modular AI/ML pipeline features. Designed for advanced, data-driven strategies and easy integration into the EA's AI subsystem.
Traditional Indicator-Based Signals
- SignalMA.mqh: Moving Average signal generator.
- SignalMACD.mqh: MACD oscillator signal generator.
- SignalRSI.mqh: Relative Strength Index signal generator.
- SignalStoch.mqh: Stochastic oscillator signal generator.
- SignalPB.mqh: Pin Bar pattern signal generator.
- SignalNewsFilter.mqh: News event filter for signals, configurable by impact and lookback period.
- SignalSessionFilter.mqh: Session-based filter (London, New York, Tokyo sessions).
Individual Signal Modules
Below is a comprehensive list of all signal modules in the Signals/ directory, with a brief description of each:
- SignalAC.mqh: (Removed)
- SignalAO.mqh: (Removed)
- SignalCCI.mqh: (Removed)
- SignalCONV.mqh: Convolutional AI signal. Uses a neural network for advanced pattern recognition.
- SignalDTDB.mqh: (Removed)
- SignalEB.mqh: (Removed)
- SignalIB.mqh: (Removed)
- SignalIchimoku.mqh: Ichimoku Kinko Hyo classic vote, written from scratch (no standard-library module exists). 12 patterns numbered weakest-to-strongest, covering the full repertoire: price/cloud bias, projected cloud colour and full Chikou Span confirmation (models 0-2, all at the standard-library floor weight of 10 since each is a standing state rather than a trigger); the TK cross graded weak/neutral/strong by cloud position (3/5/10, weights 10/40/90); Kumo twist (4); Kijun-sen cross and bounce (6/7); Kumo breakout and thin-cloud breakout (8/9); and Sanyaku Kōten/Gyakuten at 100 (11). Its class comment documents MT5's draw-shift-only buffer convention and the resulting lookahead hazard.
- SignalITF.mqh: Intraday Time Filter. Filters signals based on time-of-day and day-of-week.
- SignalLSTM.mqh: LSTM AI signal. Uses a recurrent neural network for sequence-based prediction.
- SignalMA.mqh: Moving Average classic vote (unified
ADMovingAveragecustom indicator; 4 patterns). - SignalMACD.mqh: MACD oscillator classic vote, ported from the MQL5 standard library. 6 patterns including single and double price/oscillator divergence — the only divergence model in the classic set.
- SignalNewsFilter.mqh: (Filters trading signals based on economic news events and impact levels. Configurable lookback window and impact threshold.)
- SignalPAI.mqh: Implements a Perceptron AI-based signal generator. Inherits from
CExpertSignalAIBase. Provides methods for initializing, training, and using a perceptron neural network for trade signal generation. Supports dynamic configuration, indicator integration, and modular AI/ML pipeline features. Designed for advanced, data-driven strategies and easy integration into the EA's AI subsystem. - SignalPB.mqh: (Removed)
- SignalRSI.mqh: RSI classic vote (4 patterns).
- SignalRVI.mqh: (Removed)
- Signals.mqh: Main orchestration file for all signals.
- SignalSAR.mqh: (Removed)
- SignalSessionFilter.mqh: (Removed)
- SignalStoch.mqh: (Removed)
- SignalWPR.mqh: (Removed)
Integration Notes
- All signals derive from a common base (typically
CExpertSignalCustomorCExpertSignalAIBase). - Modular design allows for easy addition/removal of signals and filters.
- Migration to a fully AI/ML-driven pipeline is recommended for future-proofing and improved performance.
- Some files (e.g., SignalPAI.mqh) may require conversion or external review due to non-text format.
Documented April 2026. For AI/ML migration and modernization, see AI_NETWORK.md and project roadmap.