driada 0.6.0__py3-none-any.whl

driada/__init__.py

@@ -0,0 +1,104 @@
+"""
+DRIADA - Dimensionality Reduction for Integrated Activity Data
+
+A library for single neuron and population-level analysis of inner workings
+of intelligent systems, from brain neural recordings in vivo to RNNs.
+"""
+
+__version__ = "0.6.0"
+
+# Core modules
+from . import intense
+from . import information
+from . import experiment
+from . import utils
+from . import rsa
+from . import network
+from . import dim_reduction
+from . import dimensionality
+from . import integration
+from . import gdrive
+
+# Key classes
+from .experiment import Experiment
+from .information import TimeSeries, MultiTimeSeries
+from .network import Network
+
+# Main INTENSE pipeline functions
+from .intense import (
+    compute_cell_feat_significance,
+    compute_feat_feat_significance,
+    compute_cell_cell_significance,
+)
+
+# Common information theory functions
+from .information import (
+    get_mi,
+    conditional_mi,
+    interaction_information,
+)
+
+# Experiment utilities
+from .experiment import (
+    load_experiment,
+    load_exp_from_aligned_data,
+    save_exp_to_pickle,
+    load_exp_from_pickle,
+    generate_synthetic_exp,
+    generate_circular_manifold_exp,
+    generate_2d_manifold_exp,
+    generate_3d_manifold_exp,
+    generate_mixed_population_exp,
+)
+
+# Dimensionality reduction classes
+from .dim_reduction import MVData, Embedding
+
+# Dimensionality estimation
+from .dimensionality import eff_dim, nn_dimension, pca_dimension
+
+__all__ = [
+    # Version
+    "__version__",
+    # Modules
+    "intense",
+    "information",
+    "experiment",
+    "utils",
+    "rsa",
+    "network",
+    "dim_reduction",
+    "dimensionality",
+    "integration",
+    "gdrive",
+    # Core classes
+    "Experiment",
+    "TimeSeries",
+    "MultiTimeSeries",
+    "Network",
+    # INTENSE pipelines
+    "compute_cell_feat_significance",
+    "compute_feat_feat_significance",
+    "compute_cell_cell_significance",
+    # Information theory
+    "get_mi",
+    "conditional_mi",
+    "interaction_information",
+    # Experiment utilities
+    "load_experiment",
+    "load_exp_from_aligned_data",
+    "save_exp_to_pickle",
+    "load_exp_from_pickle",
+    "generate_synthetic_exp",
+    "generate_circular_manifold_exp",
+    "generate_2d_manifold_exp",
+    "generate_3d_manifold_exp",
+    "generate_mixed_population_exp",
+    # Dimensionality reduction
+    "MVData",
+    "Embedding",
+    # Dimensionality estimation
+    "eff_dim",
+    "nn_dimension",
+    "pca_dimension",
+]

driada/dim_reduction/README.md

@@ -0,0 +1,74 @@
+# Dimensionality Reduction Module
+
+## Overview
+
+The dimensionality reduction module provides a comprehensive suite of algorithms for extracting low-dimensional representations from high-dimensional neural population data. It supports both linear and nonlinear methods with a unified interface.
+
+## Key Components
+
+### MVData Class
+The core data container that `MultiTimeSeries` inherits from. Provides:
+- Automatic downsampling
+- Preprocessing (normalization, standardization)
+- Direct access to dimensionality reduction methods via `get_embedding()`
+
+### Supported Methods
+
+**Linear Methods:**
+- PCA (Principal Component Analysis)
+- FA (Factor Analysis)
+- LLE (Locally Linear Embedding)
+- LEM (Laplacian Eigenmaps)
+- DM (Diffusion Maps)
+
+**Nonlinear Methods:**
+- Isomap (Isometric Mapping)
+- UMAP (Uniform Manifold Approximation and Projection)
+- t-SNE (t-Distributed Stochastic Neighbor Embedding)
+- Autoencoders (standard and variational)
+- MVU (Maximum Variance Unfolding)
+
+### Manifold Quality Metrics
+- `knn_preservation_rate()` - Proportion of neighbors preserved
+- `trustworthiness()` - Are close points in low-D truly close in high-D?
+- `continuity()` - Are close points in high-D still close in low-D?
+- `procrustes_analysis()` - Optimal alignment between embeddings
+- `circular_structure_preservation()` - For ring-like manifolds
+
+## Example Usage
+
+```python
+from driada.dim_reduction import MVData
+
+# Create MVData object (or use experiment.calcium directly)
+mvdata = MVData(neural_data, downsampling=5)
+
+# Apply dimensionality reduction
+embedding = mvdata.get_embedding(method='umap', n_components=2, n_neighbors=30)
+
+# Access coordinates
+coords = embedding.coords.T  # Shape: (n_samples, n_dims)
+
+# Evaluate quality
+from driada.dim_reduction import knn_preservation_rate
+quality = knn_preservation_rate(neural_data, coords, k=10)
+```
+
+## Advanced Features
+
+### Sequential Reduction
+Chain multiple methods for optimal results:
+```python
+from driada.dim_reduction import dr_sequence
+
+result = dr_sequence(
+    mvdata,
+    steps=[
+        ('pca', {'dim': 50}),    # Initial denoising
+        ('umap', {'dim': 2})     # Final embedding
+    ]
+)
+```
+
+### Graph-Based Methods
+The module includes `ProximityGraph` for methods that build on nearest-neighbor graphs, enabling shared graph construction across algorithms.

driada/dim_reduction/__init__.py

@@ -0,0 +1,71 @@
+"""
+Dimensionality Reduction Module for DRIADA
+
+This module provides various dimensionality reduction algorithms and
+utilities for analyzing high-dimensional neural data.
+"""
+
+from .data import MVData
+from .dr_base import (
+    METHODS_DICT,
+    DRMethod,
+    merge_params_with_defaults,
+    e_param_filter,
+    g_param_filter,
+    m_param_filter,
+)
+from .embedding import Embedding
+from .graph import ProximityGraph
+from .sequences import dr_sequence
+
+# Import manifold metrics
+from .manifold_metrics import (
+    compute_distance_matrix,
+    knn_preservation_rate,
+    trustworthiness,
+    continuity,
+    geodesic_distance_correlation,
+    stress,
+    circular_structure_preservation,
+    procrustes_analysis,
+    manifold_preservation_score,
+    # Reconstruction validation functions
+    circular_distance,
+    extract_angles_from_embedding,
+    compute_reconstruction_error,
+    compute_embedding_alignment_metrics,
+    compute_decoding_accuracy,
+    manifold_reconstruction_score,
+)
+
+__all__ = [
+    # Core classes
+    "MVData",
+    "METHODS_DICT",
+    "DRMethod",
+    "Embedding",
+    "ProximityGraph",
+    # Functions
+    "dr_sequence",
+    "merge_params_with_defaults",
+    "e_param_filter",
+    "g_param_filter",
+    "m_param_filter",
+    # Manifold metrics
+    "compute_distance_matrix",
+    "knn_preservation_rate",
+    "trustworthiness",
+    "continuity",
+    "geodesic_distance_correlation",
+    "stress",
+    "circular_structure_preservation",
+    "procrustes_analysis",
+    "manifold_preservation_score",
+    # Reconstruction validation functions
+    "circular_distance",
+    "extract_angles_from_embedding",
+    "compute_reconstruction_error",
+    "compute_embedding_alignment_metrics",
+    "compute_decoding_accuracy",
+    "manifold_reconstruction_score",
+]