dfm-python 0.5.55__tar.gz → 0.5.57__tar.gz

{dfm_python-0.5.55 → dfm_python-0.5.57}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dfm-python
-Version: 0.5.55
+Version: 0.5.57
 Summary: Dynamic Factor Model (DFM) estimation and forecasting in Python
 Project-URL: Homepage, https://github.com/yourusername/dfm-python
 Project-URL: Documentation, https://github.com/yourusername/dfm-python#readme
@@ -8,7 +8,6 @@ Project-URL: Repository, https://github.com/yourusername/dfm-python
 Project-URL: Issues, https://github.com/yourusername/dfm-python/issues
 Author: Minkey Chang
 License: MIT
-License-File: LICENSE
 Keywords: dfm,dynamic-factor-model,econometrics,forecasting,time-series
 Classifier: Development Status :: 4 - Beta
 Classifier: Intended Audience :: Science/Research
@@ -176,7 +175,7 @@ dfm-python/
 │   └── dfm_python/
 │       ├── models/          # Model implementations (DFM, DDFM, KDFM)
 │       ├── config/           # Configuration schemas
-│       ├── datamodule/       # PyTorch Lightning DataModules
+│       ├── datamodule/       # Data loading and preprocessing modules
 │       ├── trainer/          # Training procedures
 │       ├── ssm/              # State-space model components
 │       ├── dataset/          # Data loading and preprocessing
@@ -194,7 +193,7 @@ dfm-python/
 - **Structural identification**: Learnable structural shock identification (Cholesky, full, low-rank)
 - **Efficient computation**: Krylov FFT for O(T log T) forward pass
 - **Direct IRF computation**: Avoids numerical error accumulation
-- **Gradient descent training**: Uses PyTorch Lightning for training
+- **Gradient descent training**: Uses PyTorch for training with custom logging
 
 ### DFM (Dynamic Factor Model)
 
@@ -207,7 +206,7 @@ dfm-python/
 
 - **Nonlinear encoder**: Neural network encoder for nonlinear factor extraction
 - **MCMC training**: Denoising training procedure
-- **PyTorch-based**: Full PyTorch implementation with Lightning support
+- **PyTorch-based**: Full PyTorch implementation with custom logging system
 
 ## IRF Analysis: KDFM's PRIMARY CONTRIBUTION
 

{dfm_python-0.5.55 → dfm_python-0.5.57}/README.md

@@ -118,7 +118,7 @@ dfm-python/
 │   └── dfm_python/
 │       ├── models/          # Model implementations (DFM, DDFM, KDFM)
 │       ├── config/           # Configuration schemas
-│       ├── datamodule/       # PyTorch Lightning DataModules
+│       ├── datamodule/       # Data loading and preprocessing modules
 │       ├── trainer/          # Training procedures
 │       ├── ssm/              # State-space model components
 │       ├── dataset/          # Data loading and preprocessing
@@ -136,7 +136,7 @@ dfm-python/
 - **Structural identification**: Learnable structural shock identification (Cholesky, full, low-rank)
 - **Efficient computation**: Krylov FFT for O(T log T) forward pass
 - **Direct IRF computation**: Avoids numerical error accumulation
-- **Gradient descent training**: Uses PyTorch Lightning for training
+- **Gradient descent training**: Uses PyTorch for training with custom logging
 
 ### DFM (Dynamic Factor Model)
 
@@ -149,7 +149,7 @@ dfm-python/
 
 - **Nonlinear encoder**: Neural network encoder for nonlinear factor extraction
 - **MCMC training**: Denoising training procedure
-- **PyTorch-based**: Full PyTorch implementation with Lightning support
+- **PyTorch-based**: Full PyTorch implementation with custom logging system
 
 ## IRF Analysis: KDFM's PRIMARY CONTRIBUTION
 

dfm_python-0.5.57/config/ddfm_exchange.yaml

@@ -0,0 +1,38 @@
+# DDFM Configuration for Exchange Rate Dataset
+# Based on original TensorFlow DDFM implementation (DDFM/run_exchange_rate_original.py)
+# Data: 8 exchange rate series (AUD_USD, GBP_USD, CAD_USD, CHF_USD, CNY_USD, JPY_USD, SGD_USD, NZD_USD)
+# Clock: Daily (d) - all series are daily frequency
+# Date range: 1990-01-01 to 2010-10-10
+
+# Neural Network Architecture (matching original TensorFlow DDFM)
+encoder_layers: [16, 4]  # Hidden layers: [16] -> latent: 4 (matching original structure_encoder=(16, 4))
+activation: relu  # Matching original link='relu'
+use_batch_norm: true  # Matching original batch_norm=True
+
+# Training Parameters (matching original TensorFlow DDFM)
+learning_rate: 0.005  # Matching original learning_rate=0.005
+n_mc_samples: 10  # Number of MC samples per MCMC iteration (matching original epochs=10)
+window_size: 100  # Time-step batch size for training (matching original batch_size=100)
+
+# MCMC/Iterative Parameters (matching original TensorFlow DDFM)
+max_epoch: 200  # Maximum MCMC iterations (matching original max_iter=200)
+tolerance: 0.0005  # Convergence tolerance (matching original tolerance=0.0005)
+disp: 10  # Display frequency for training progress (matching original disp=10)
+seed: 3  # Random seed (matching original seed=3)
+
+# Data Handling (used by Dataset, not Model)
+clock: d  # Daily clock frequency (all series are daily)
+target_scaler: null  # No target scaler - data should be pre-scaled before passing to Dataset
+
+# Frequency mapping (all series are daily)
+# If not specified, all series use clock frequency (d)
+# Format: {"series_name": "frequency_code"}
+# frequency:
+#   AUD_USD: d
+#   GBP_USD: d
+#   CAD_USD: d
+#   CHF_USD: d
+#   CNY_USD: d
+#   JPY_USD: d
+#   SGD_USD: d
+#   NZD_USD: d

dfm_python-0.5.57/config/ddfm_finance.yaml

@@ -0,0 +1,31 @@
+# DDFM Configuration for Finance Dataset
+# Data: Finance data with market_forward_excess_returns as target variable
+# Clock: Monthly (m) - all series are monthly frequency
+
+# Neural Network Architecture
+encoder_layers: [64, 32, 2]  # Hidden layers: [64, 32] -> latent: 2 factors
+activation: relu  # Activation function: 'tanh', 'relu', 'sigmoid'
+use_batch_norm: true  # Use batch normalization
+
+# Training Parameters
+learning_rate: 0.001  # Learning rate for optimizer
+n_mc_samples: 1  # Number of MC samples per MCMC iteration (reduced for faster execution)
+window_size: 100  # Time-step batch size for training
+
+# MCMC/Iterative Parameters
+max_epoch: 200  # Maximum MCMC iterations
+tolerance: 0.0005  # Convergence tolerance
+disp: 10  # Display frequency for training progress
+seed: null  # Random seed (None uses default)
+
+# Data Handling (used by Dataset, not Model)
+clock: m  # Monthly clock frequency
+target_scaler: standard  # Scaler for target series: 'standard', 'robust', 'minmax', 'maxabs', 'quantile', or null
+
+# Frequency mapping (optional)
+# If not specified, all series use clock frequency
+# Format: {"series_name": "frequency_code"}
+# frequency:
+#   gdp: q
+#   unemployment: m
+#   interest_rate: m

dfm_python-0.5.57/config/ddfm_macro.yaml

@@ -0,0 +1,55 @@
+# DDFM Configuration for Macroeconomic Datasets
+# Based on Korean economic indicators (KO* series)
+# Data: 101 series, mixed weekly/monthly frequency (1985-04-05 to 2025-11-21)
+# Clock: Weekly (w) - base frequency for latent factors
+# Target: KOEQUIPTE (Investment, Equipment, Estimation, SA)
+
+# Neural Network Architecture
+encoder_layers: [64, 32, 2]  # Hidden layers: [64, 32] -> latent: 2 factors
+activation: relu  # Activation function: 'tanh', 'relu', 'sigmoid'
+use_batch_norm: true  # Use batch normalization
+
+# Training Parameters
+learning_rate: 0.001  # Learning rate for optimizer
+n_mc_samples: 1  # Number of MC samples per MCMC iteration (reduced for faster execution)
+window_size: 100  # Time-step batch size for training
+
+# MCMC/Iterative Parameters
+max_epoch: 200  # Maximum MCMC iterations
+tolerance: 0.0005  # Convergence tolerance
+disp: 10  # Display frequency for training progress
+seed: null  # Random seed (None uses default)
+
+# Data Handling (used by Dataset, not Model)
+clock: w  # Weekly clock frequency (base frequency for latent factors)
+target_scaler: standard  # Scaler for target series: 'standard', 'robust', 'minmax', 'maxabs', 'quantile', or null
+
+# Frequency mapping for mixed-frequency macro data
+# Clock is weekly (w), so monthly series are slower frequency
+# Grouped format: frequencies are specified as lists grouped by frequency code
+# Reduced set for faster testing (~50% reduction: 12 monthly series kept from original ~24)
+frequency:
+  # Monthly series (slower than weekly clock) - typically end-of-month observations
+  # Essential series for testing
+  m:
+    - KOEMPTOTO  # Total employment (monthly)
+    - KOHWRWEMP  # Hours worked (monthly)
+    - KOIMPCONA  # Investment, Construction (monthly)
+    - KOWRCCNSE  # Consumption (monthly)
+    - KOCONPRCF  # Production (monthly)
+    - KOCPCOREF  # Consumer prices (monthly)
+    - KOCPEXFEF  # Export prices (monthly)
+    - KOCREU..A  # Credit (monthly)
+    - KOLRLTHH  # Interest rates (monthly)
+    - KOBSIHAPR  # Business survey (monthly)
+    - KOCNFCONR  # Confidence (monthly)
+    - KOEQUIPTE  # Investment, Equipment (monthly, target series)
+  # Weekly series (same as clock frequency)
+  w:
+    - KORFBEV  # Retail, Food & Beverages (weekly)
+    - KORWHLS  # Retail, Wholesale (weekly)
+    - KORMCMP  # Retail, Miscellaneous & Comparison (weekly)
+    - KORELEC  # Retail, Electronics (weekly)
+    - KORCNST  # Retail, Construction (weekly)
+    - KORMACH  # Retail, Machinery (weekly)
+    - KOROCOINF  # Retail, Other Consumer & Information (weekly)

{dfm_python-0.5.55 → dfm_python-0.5.57}/config/dfm_macro.yaml

@@ -55,30 +55,32 @@ blocks:
 
 # Frequency mapping for mixed-frequency macro data
 # Clock is weekly (w), so monthly series are slower frequency
-# All frequencies must be explicitly specified
+# Grouped format: frequencies are specified as lists grouped by frequency code
 # Reduced set for faster testing (~50% reduction: 18 monthly series kept from original ~36)
 frequency:
   # Monthly series (slower than weekly clock) - typically end-of-month observations
   # Block_Real series
-  KOEMPTOTO: m  # Total employment (monthly)
-  KOHWRWEMP: m  # Hours worked (monthly)
-  KOIMPCONA: m  # Investment, Construction (monthly)
-  KOWRCCNSE: m  # Consumption (monthly)
-  KOCONPRCF: m  # Production (monthly)
-  # Block_Prices_Financial series
-  KOCPCOREF: m  # Consumer prices (monthly)
-  KOCPEXFEF: m  # Export prices (monthly)
-  KOCREU..A: m  # Credit (monthly)
-  KOLRLTHH: m  # Interest rates (monthly)
-  KOBSIHAPR: m  # Business survey (monthly)
-  KOCNFCONR: m  # Confidence (monthly)
-  # Target series
-  KOEQUIPTE: m  # Investment, Equipment (monthly, target series)
+  m:
+    - KOEMPTOTO  # Total employment (monthly)
+    - KOHWRWEMP  # Hours worked (monthly)
+    - KOIMPCONA  # Investment, Construction (monthly)
+    - KOWRCCNSE  # Consumption (monthly)
+    - KOCONPRCF  # Production (monthly)
+    # Block_Prices_Financial series
+    - KOCPCOREF  # Consumer prices (monthly)
+    - KOCPEXFEF  # Export prices (monthly)
+    - KOCREU..A  # Credit (monthly)
+    - KOLRLTHH  # Interest rates (monthly)
+    - KOBSIHAPR  # Business survey (monthly)
+    - KOCNFCONR  # Confidence (monthly)
+    # Target series
+    - KOEQUIPTE  # Investment, Equipment (monthly, target series)
   # Weekly series (same as clock frequency)
-  KORFBEV: w  # Retail, Food & Beverages (weekly)
-  KORWHLS: w  # Retail, Wholesale (weekly)
-  KORMCMP: w  # Retail, Miscellaneous & Comparison (weekly)
-  KORELEC: w  # Retail, Electronics (weekly)
-  KORCNST: w  # Retail, Construction (weekly)
-  KORMACH: w  # Retail, Machinery (weekly)
-  KOROCOINF: w  # Retail, Other Consumer & Information (weekly)
+  w:
+    - KORFBEV  # Retail, Food & Beverages (weekly)
+    - KORWHLS  # Retail, Wholesale (weekly)
+    - KORMCMP  # Retail, Miscellaneous & Comparison (weekly)
+    - KORELEC  # Retail, Electronics (weekly)
+    - KORCNST  # Retail, Construction (weekly)
+    - KORMACH  # Retail, Machinery (weekly)
+    - KOROCOINF  # Retail, Other Consumer & Information (weekly)