Spaces:
Sleeping
Sleeping
replaced markdown with html
Browse files
app.py
CHANGED
|
@@ -178,13 +178,13 @@ with gr.Blocks(css=custom_css) as interface:
|
|
| 178 |
outputs=sample_display,
|
| 179 |
)
|
| 180 |
|
| 181 |
-
gr.
|
| 182 |
-
|
| 183 |
|
| 184 |
-
|
| 185 |
|
| 186 |
-
|
| 187 |
-
|
| 188 |
with gr.Row(elem_classes="custom-row"):
|
| 189 |
dff_gallery = gr.Gallery(
|
| 190 |
label="Deep Feature Factorization",
|
|
|
|
| 178 |
outputs=sample_display,
|
| 179 |
)
|
| 180 |
|
| 181 |
+
gr.HTML("""
|
| 182 |
+
<span style="color: purple;">Concept Discovery</span> involves identifying interpretable high-level features or concepts within a deep learning model's representation. It aims to understand what a model has learned and how these learned features relate to meaningful attributes in the data.
|
| 183 |
|
| 184 |
+
<span style="color: purple;">Deep Feature Factorization (DFF)</span> is a technique that decomposes the deep features learned by a model into disentangled and interpretable components. It typically involves matrix factorization methods applied to activation maps, enabling the identification of semantically meaningful concepts captured by the model.
|
| 185 |
|
| 186 |
+
Together, these methods enhance model interpretability and provide insights into the decision-making process of neural networks.
|
| 187 |
+
""")
|
| 188 |
with gr.Row(elem_classes="custom-row"):
|
| 189 |
dff_gallery = gr.Gallery(
|
| 190 |
label="Deep Feature Factorization",
|