deveshc23 commited on
Commit
24ae774
Β·
verified Β·
1 Parent(s): 80ebf40

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +139 -7
README.md CHANGED
@@ -1,13 +1,145 @@
1
  ---
2
- title: Quantagents
3
- emoji: 🏒
4
- colorFrom: red
5
- colorTo: indigo
6
  sdk: gradio
7
- sdk_version: 6.30.0
8
- python_version: '3.12'
9
  app_file: app.py
10
  pinned: false
 
11
  ---
12
 
13
- Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ title: QuantAgent OS - Autonomous Quant Trading & Self-Improving Alpha Discovery
3
+ emoji: πŸ“ˆ
4
+ colorFrom: indigo
5
+ colorTo: blue
6
  sdk: gradio
7
+ sdk_version: 4.42.0
 
8
  app_file: app.py
9
  pinned: false
10
+ license: mit
11
  ---
12
 
13
+ # πŸ›οΈ QuantAgent OS: Autonomous Quantitative Trading System
14
+ ### *Self-Improving Multi-Agent Quantitative Architecture based on arXiv:2402.03755*
15
+
16
+ ![Python](https://img.shields.io/badge/Python-3.10%2B-blue.svg)
17
+ ![Gradio](https://img.shields.io/badge/Gradio-4.42.0-orange.svg)
18
+ ![arXiv](https://img.shields.io/badge/arXiv-2402.03755-b31b1b.svg)
19
+ ![License](https://img.shields.io/badge/License-MIT-green.svg)
20
+
21
+ QuantAgent OS is an institutional-grade, self-improving quantitative investment system powered by agentic AI. It implements the two-layer loop architecture pioneered in the research paper:
22
+
23
+ > **QuantAgent: Seeking Holy Grail in Trading by Self-Improving Large Language Model**
24
+ > *Saizhuo Wang, Hang Yuan, Lionel M. Ni, Jian Guo (HKUST & IDEA Research)*
25
+ > [arXiv:2402.03755](https://arxiv.org/abs/2402.03755)
26
+
27
+ ---
28
+
29
+ ## πŸ—οΈ System Architecture
30
+
31
+ QuantAgent OS solves the core bottleneck of domain-specific financial AI: **how to autonomously discover, test, and accumulate quantitative trading signals (alphas) without manual human data curation.**
32
+
33
+ ```
34
+ +---------------------------------------+
35
+ | Outer Feedback Loop |
36
+ | (Slow / Real-World Backtest Oracle) |
37
+ +---------------------------------------+
38
+ | ^
39
+ Submit Alpha | | Ground-truth Metrics
40
+ for Backtest v | & Qualitative Review
41
+ +--------------------+ |
42
+ | Real Environment |-----------+
43
+ | (Backtesting Eng) |
44
+ +--------------------+
45
+ |
46
+ Updates KB v
47
+ +---------------------------------------------------------------------------------+
48
+ | Inner Reasoning Loop (Fast Simulation) |
49
+ | |
50
+ | +------------------------------------------------------------------+ |
51
+ | | Shared Context Buffer (ctx) | |
52
+ | +------------------------------------------------------------------+ |
53
+ | | ^ | ^ |
54
+ | 1. Query | 2. Writer Hint | 3. Generated | 4. Judge | 5. Review |
55
+ | v | Alpha Code v Hint | & Score |
56
+ | +---------------+ +---------------+ +---------------+ | |
57
+ | | KnowledgeBase | | Writer LLM | | Judge LLM |---+ |
58
+ | | (KB) | +---------------+ +---------------+ |
59
+ | +---------------+ |
60
+ +---------------------------------------------------------------------------------+
61
+ ```
62
+
63
+ ### 1. The Inner Reasoning Loop (Fast Heuristic Simulation)
64
+ - **Shared Context Buffer**: Holds task descriptions, retrieved knowledge hints, generated factor code, and mentor feedback.
65
+ - **Quant Alpha Writer**: Translates qualitative trading ideas into vectorized, lookahead-free Python factor code inheriting from the standardized `Factor` base class.
66
+ - **Quant Mentor Judge**: Evaluates candidate signals against 6 quantitative criteria:
67
+ 1. *Vectorization efficiency* (no single-asset Python loops)
68
+ 2. *Lookahead bias prevention* (no `shift(-1)` or future leakage)
69
+ 3. *Continuous nuance* (smooth tanh / z-score scaling into `[-1.0, 1.0]`)
70
+ 4. *Numerical stability* (`.clip()`, `.fillna()`, zero-division protection)
71
+ 5. *Code validity*
72
+ 6. *Economic intuition alignment*
73
+ - **Iterative Dialogue**: The Writer and Judge iterate for up to $T$ rounds until the candidate meets approval score $\beta$.
74
+
75
+ ### 2. The Outer Feedback Loop (Real Backtesting Oracle)
76
+ - Submits the approved alpha code to the vectorized backtesting engine against historical market cross-sections.
77
+ - Calculates empirical metrics:
78
+ - **Information Coefficient (IC)**: Cross-sectional Pearson correlation between signal and forward returns.
79
+ - **IC Information Ratio (IC_IR)**: $\text{Mean}(IC) / \text{Std}(IC)$.
80
+ - **Long-Short Sharpe Ratio, Max Drawdown, and Turnover Rate**.
81
+ - **Autonomous Knowledge Base (KB)**: Retains both successful and failed alphas (with failure diagnosis) to enable *pessimistic value updates*.
82
+
83
+ ### 3. Downstream XGBoost Alpha Ensemble (Paper Figure 5)
84
+ - Automatically pools all valid alphas from the KB as feature columns.
85
+ - Trains an XGBoost cross-validated non-linear regression model to predict cross-sectional asset returns, demonstrating how prediction Mean Squared Error (MSE) monotonically decreases as the KB accumulates more diverse alphas.
86
+
87
+ ---
88
+
89
+ ## 🌐 Free Public Data Sources Integrated
90
+
91
+ | Source | Description | Authentication |
92
+ | :--- | :--- | :--- |
93
+ | **Yahoo Finance** | Live & historical OHLCV, volume, dividends, valuation metrics, beta, market cap | 100% Free (Built-in browser session) |
94
+ | **SEC EDGAR API** | Official regulatory 10-K and 10-Q XBRL financial facts: Revenue, Net Income, Total Assets, Liabilities, Debt/Equity | 100% Free (Official SEC endpoint) |
95
+ | **FRED / Market Proxies** | 10Y Treasury, 3M T-bill, Yield Curve Slope, VIX Volatility Index, Crude Oil, USD Index | 100% Free (FRED key optional) |
96
+ | **Yahoo RSS & Finnhub** | Live news headlines, company announcements, financial sentiment polarity scoring | 100% Free (Finnhub key optional) |
97
+
98
+ ---
99
+
100
+ ## πŸš€ How to Run Locally
101
+
102
+ ```bash
103
+ # 1. Clone the repository
104
+ git clone https://github.com/your-username/quantagent-os.git
105
+ cd quantagent-os
106
+
107
+ # 2. Install dependencies
108
+ pip install -r requirements.txt
109
+
110
+ # 3. Launch the application
111
+ python app.py
112
+ ```
113
+ Open your browser at `http://localhost:7860` to access the interactive dashboard.
114
+
115
+ ---
116
+
117
+ ## ☁️ Deploying to Hugging Face Spaces
118
+
119
+ You can deploy this Space in one of two ways:
120
+
121
+ ### Method 1: Directly from the Web UI (Tab 6)
122
+ 1. Open the application locally or on your server.
123
+ 2. Navigate to **Tab 6: Hugging Face Deployment & Settings**.
124
+ 3. Enter your Hugging Face API Token (with `write` permission) and target Space ID (`username/quantagent-os`).
125
+ 4. Click **Deploy to Hugging Face Spaces**.
126
+
127
+ ### Method 2: Via Git
128
+ ```bash
129
+ git remote add space https://huggingface.co/spaces/your-username/quantagent-os
130
+ git push space main
131
+ ```
132
+
133
+ ---
134
+
135
+ ## πŸ“ Mathematical Formulation & Regret Guarantees
136
+
137
+ The framework formulates alpha mining as an Information-State Markov Decision Process $(S, A, \mathcal{T}, r, \gamma)$:
138
+ - **Inner Loop**: Under implicit Bayesian inference by the LLM (Assumption 4.1), Bayesian regret is sublinear in inner iterations $T$.
139
+ - **Outer Loop**: Grounded in Pessimistic Value Iteration (PEVI, Jin et al. 2021), epistemic uncertainty between simulated KB and real-world environment decays with outer iterations $K$.
140
+ - **Overall Regret**: Total Bayesian regret $R(TK)$ decomposes into four bounded terms and is proven **sublinear in $KT$** (Theorem 4.6).
141
+
142
+ ---
143
+
144
+ ## βš–οΈ License
145
+ Released under the MIT License. Educational and quantitative research use only.