reorg

buds-lab · Aug 11, 2019 · 8e6ca55 · 8e6ca55
commit 8e6ca55
Show file tree

Hide file tree

Showing 183 changed files with 61,871 additions and 0 deletions.
diff --git a/.gitignore b/.gitignore
@@ -0,0 +1,2 @@
+*.csv
+*.zip
diff --git a/README.md b/README.md
@@ -0,0 +1,18 @@
+# Outline
+
+## Problem Statement
+- Gathering thermal comfort related data (or in general user-generated data)
+is costly and takes time.
+- Hard to generalize over a big population of people, as well as obtain a similar
+number of responses for each label/category (e.g. Thermal comfort labels, always
+a predominance in the 'comfort' class and not the rest)
+
+## Methods
+- Try different generative models for augment datasets
+    - GAN (and its variations: CGAN, WGAN, WCGAN, TGAN, TableGAN)
+    - Autoencoders (and the variations: Adversarial, variational)
+
+# Evaluation metrics
+- Baseline: Original train and test set
+- Train set and synthetic data as training set: Should increase the performance since classes would be more balanced
+- Synthetic data as training set: Performance should be comparable with the baseline showing the synthetic set captures the same characteristics as the real train set
diff --git a/__pycache__/RBM.cpython-37.pyc b/__pycache__/RBM.cpython-37.pyc
diff --git a/__pycache__/gan.cpython-37.pyc b/__pycache__/gan.cpython-37.pyc
diff --git a/__pycache__/utils.cpython-37.pyc b/__pycache__/utils.cpython-37.pyc
diff --git a/__pycache__/vae.cpython-37.pyc b/__pycache__/vae.cpython-37.pyc
diff --git a/data/.DS_Store b/data/.DS_Store