forked from kmopuri/kmopuri.github.io
-
Notifications
You must be signed in to change notification settings - Fork 0
/
long-tail.html
39 lines (31 loc) · 2.54 KB
/
long-tail.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
<html lang="en"> <!-- <head> -->
<p style="padding:20px;width:100%;vertical-align:middle;text-align:center;font-size:20px;">Research related to Long-tailed Training Data</p>
<p style="width:100%;vertical-align:middle;text-align:center;font-size:15px"> <a href="index.html">Back to Home</a> </p>
<body>
<!-- UAI 2021 -->
<table style="width:100%;max-width:800px;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;"><tbody>
<td style="padding:20px;width:25%;vertical-align:middle">
<img src='images/class-balancing-GAN.jpg' width="160">
</td>
<td style="padding:20px;width:75%;vertical-align:middle">
<papertitle>Class Balancing GAN with a Classifier in the Loop</papertitle>
<br>
Harsh Rangwani,
<strong>Konda Reddy Mopuri</strong>,
<a href="http://cds.iisc.ac.in/faculty/venky/">R. Venkatesh Babu</a>
<br>
<em>UAI</em>, 2021
<br>
<a href="https://arxiv.org/pdf/2106.09402">PDF</a>
<p></p>
<p>GANs have swiftly evolved to imitate increasingly complex image distributions. However, majority of the developments focus on performance of GANs on
balanced datasets. We find that the existing GANs and their training regimes which work well on balanced datasets fail to be effective in case of imbalanced (i.e. long-tailed) datasets. In this work
we introduce a novel, theoretically motivated Class Balancing regularizer for training GANs. Our regularizer makes use of the knowledge from a pretrained classifier to ensure balanced learning of
all the classes in the dataset. This is achieved via modelling the effective class frequency based on the exponential forgetting observed in neural networks and encouraging the GAN to focus on underrepresented classes. We demonstrate the utility
of our regularizer in learning representations for long-tailed distributions via achieving better performance than existing approaches over multiple datasets. Specifically, when applied to an unconditional GAN, it improves the FID from 13.03 to
9.01 on the long-tailed iNaturalist-2019 dataset.</p>
</td>
</tr>
</tbody></table>
<p style="width:100%;vertical-align:middle;text-align:center;font-size:15px"> <a href="index.html">Back to Home</a> </p>
</body>