rhd008 commited on
Commit
9a6b09b
·
verified ·
1 Parent(s): e32e8eb

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +42 -0
README.md CHANGED
@@ -70,3 +70,45 @@ configs:
70
  - split: test
71
  path: p=40/test-*
72
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
70
  - split: test
71
  path: p=40/test-*
72
  ---
73
+
74
+ # Covariance Denoising Dataset
75
+
76
+ This dataset is designed for training and evaluating machine learning models for covariance denoising. It consists of synthetically generated data, including noisy and clean covariance matrices. You can access the dataset on Hugging Face: [https://huggingface.co/datasets/rhd008/covariance_denoising](https://huggingface.co/datasets/rhd008/covariance_denoising)
77
+
78
+ ## Data Description
79
+
80
+ The dataset contains the following features for each sample:
81
+
82
+ * **n_samples:** An array indicating the number of samples used to generate the noisy covariance matrices.
83
+ * **noisy_covariances:** A stack of noisy covariance matrices generated with different numbers of samples.
84
+ * **clean_covariance:** The ground truth (clean) covariance matrix.
85
+ * **clean_precision:** The ground truth (clean) precision matrix, which is the inverse of the covariance matrix.
86
+
87
+ ## Dataset Splits
88
+
89
+ The dataset is divided into three splits:
90
+
91
+ * **Train:** Used for training the model.
92
+ * **Validation:** Used for evaluating the model during training.
93
+ * **Test:** Used for final evaluation of the model.
94
+
95
+ ## Data Generation
96
+
97
+ The data is generated using the following steps:
98
+
99
+ 1. Generate sparse symmetric positive definite matrices using `make_sparse_spd_matrix` from `sklearn.datasets`.
100
+ 2. Calculate the corresponding covariance and precision matrices.
101
+ 3. Generate noisy covariance matrices by sampling from a multivariate normal distribution with the clean covariance matrix and varying numbers of samples.
102
+
103
+ ## Use Cases
104
+
105
+ This dataset can be used for various tasks, including:
106
+
107
+ * **Covariance Denoising:** Training models to estimate the clean covariance matrix from noisy observations.
108
+ * **Precision Matrix Estimation:** Training models to estimate the clean precision matrix.
109
+ * **Graphical Model Inference:** Inferring the structure of graphical models from the estimated precision matrix.
110
+
111
+ ## Citation
112
+
113
+ If you use this dataset in your research, please cite it as follows:
114
+