-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathindex.html
475 lines (431 loc) · 24.5 KB
/
index.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
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
---
layout: default
---
<!--<div class="home">-->
<!-- <h2>About</h2>-->
<!-- {% include image.html url="/images/IMG_9402.jpg" caption="Aishwarya Kamath" width=300 align="center" %}-->
<!-- <p> I am a PhD student at New York University's Center for Data Science, advised by Prof. Yann LeCun and Prof.-->
<!-- Kyunghyun Cho. My research focuses on using information from multiple sources such as text, images, video and-->
<!-- speech to improve commonsense reasoning capabilities of machines. Prior to this, I was advised during my Masters-->
<!-- by Prof. Andrew McCallum at University of Massachusetts Amherst in areas of natural language processing and-->
<!-- machine learning, with a special focus on structured prediction.</p>-->
<!-- <!– <p> I have also worked as a Machine Learning Engineer as part of the Machine Learning Research Group at Oracle Labs in Burlington, MA. </p>-->
<!-- <p> When I am not busy doing coursework or research, I love to paint and when time permits, keep my hobby of playing basketball alive. </p> –>-->
<!-- <br>-->
<!--
<!-- <hr>-->
<!-- <br>-->
<!--</div>-->
<table style="width:100%;max-width:800px;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;">
<tbody>
<table style="width:100%;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;">
<tbody>
<tr>
<td style="padding:20px;width:100%;vertical-align:middle">
<h2>About</h2>
</td>
</tr>
</tbody>
</table>
<tr style="padding:0px">
<td style="padding:0px">
<table style="width:100%;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;">
<tbody>
<tr style="padding:0px">
<td style="padding:2.5%;width:63%;vertical-align:middle">
<!-- <p style="text-align:center">-->
<!-- <name>Aishwarya Kamath</name>-->
<!-- </p>-->
<p>
I am currently joining as a Research Scientist at DeepMind (December 2023) in the Vision Team led by Andrew Zisserman.
</p>
<p>Previously, I earned my PhD on Fine-Grained Vision and Language Understanding at <a href="https://cds.nyu.edu/">New York University's Center for
Data Science</a> advised by <a
href="https://scholar.google.co.uk/citations?hl=en&user=WLN3QrAAAAAJ"> Prof. Yann
LeCun</a> and <a href="https://scholar.google.co.uk/citations?user=0RAmmIAAAAAJ&hl=en">Prof.
Kyunghyun Cho</a>. Prior to this I was advised during my
Masters by
<a href="https://scholar.google.com/citations?user=yILa1y0AAAAJ&hl=en">Prof. Andrew
McCallum</a> at <a href="https://www.cics.umass.edu/">University of Massachusetts
Amherst</a> in areas of natural language processing with a special focus on structured
prediction.
</p>
<p>
My current interests lie at the intersection of <strong>vision and language</strong>, and my
research focuses on using information from multiple sources such as text, images, and video
to improve reasoning capabilities of machines.
</p>
<p style="text-align:center">
<a href="mailto:aish@nyu.edu">Email</a>  / 
<a href="data/Resume_AishwaryaKamath_aug2021.pdf">CV</a>  / 
<a href="https://scholar.google.com/citations?user=WaW2C0UAAAAJ&hl=en&oi=ao">Google
Scholar</a>
 / 
<a href="https://twitter.com/ashkamath20">Twitter</a>  / 
<a href="https://github.com/ashkamath/">Github</a>
</p>
</td>
<td style="padding:2.5%;width:40%;max-width:40%">
<a href="/images/IMG_9402.jpg"><img style="width:100%;max-width:100%" alt="profile photo"
src="/images/IMG_9402.jpg" class="hoverZoomLink"></a>
</td>
</tr>
</tbody>
</table>
</td>
</tr>
<table style="width:100%;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;">
<tbody>
<tr>
<td style="padding:20px;width:100%;vertical-align:middle">
<h2>News</h2>
</td>
</tr>
</tbody>
</table>
<table style="width:100%;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;">
<tbody>
<tr>
<ul>
<li>
<em> September 2023 </em> Defended my PhD! A big thank you to my committee & all my friends and family for supporting me :) <a href="https://tinyurl.com/mr3tzxse">
<font color="black"><strong>Slides Here! (Thesis title: End-to-End Fine-Grained Vision and Language Understanding)</strong></font>
</a>
</li>
<li>
<em> September 2022 </em> New paper out on Vision and Language Navigation! <a href="https://t.co/XYtmIsC3MR">
<font color="black"><strong>A New Path: Scaling Vision-and-Language Navigation with Synthetic Instructions and Imitation Learning</strong></font>
</a>
</li>
<li>
<em> September 2022 </em> <a href="https://ashkamath.github.io/FIBER_page">
<font color="black"><strong>FIBER: Coarse-to-Fine Vision-Language Pre-training with Fusion in the Backbone</strong></font>
</a> accepted to NeurIPS 2022!
</li>
<li>
<em> December 2021 </em> Taught a guest lecture on multi-modal machine learning at Prof. Kyunghyun Cho's Natural Language Processing class at NYU.
</li>
<li>
<em> October 2021</em>: MDETR featured on the front page of the <a href="https://www.rsipvision.com/ICCV2021-Tuesday/">ICCV magazine</a> today.
</li>
<li>
<em>October 2021</em>: Presenting our paper : <a href="https://ashkamath.github.io/mdetr_page">
MDETR</a> at ICCV 2021, selected for Oral Presentation!
</li>
<li><em>September 2021</em>: Started Student Researcher position at Gooogle Research working on Vision
Language Navigation
(VLN)
</li>
<li><em>August 2021</em>: Invited to speak at Weights and Biases Reading Group on MDETR <a
href="https://www.youtube.com/watch?v=I68nn_0odto">video</a></li>
<li>
<em>July 2021</em>: Invited to speak at the Vision and Language Series at Microsoft Research.
Check out the <a
href="https://www.youtube.com/watch?v=QM07aZaSFak">video</a> and other talks here- <a
href="https://www.microsoft.com/en-us/research/videos/vision-language-summer-talk-series/">Series
link</a>
</li>
<li><em>June 2021</em>: Invited to speak at UKP Lab, TU Darmstadt & CIS Lab Ludwig Maximilian University
of Munich Joint Invited Talk Series
</li>
<li><em>June 2021</em>: Invited to speak at the Google AI Image Text group.</li>
<li><em>June 2021</em>: Started an internship at Google Research working with Peter Anderson, Jason
Baldridge and Zarana Parekh.
</li>
<li><em>May 2021</em>: Invited to speak at Intel's Deep Learning CoP Talk Series.</li>
</ul>
</tr>
</tbody>
</table>
<table style="width:100%;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;">
<tbody>
<tr>
<td style="padding:20px;width:100%;vertical-align:middle">
<h2>Research</h2>
</td>
</tr>
</tbody>
</table>
<table style="width:100%;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;">
<tbody>
<tr>
<td style="padding:20px;width:50%;vertical-align:middle">
<div class="one">
<img src='images/fib-pipeline.png' width="500">
</div>
</td>
<td style="padding:20px;width:50%;vertical-align:middle">
<a href="https://ashkamath.github.io/FIBER_page">
<font color="black"><strong>FIBER: Coarse-to-Fine Vision-Language Pre-training with Fusion in the Backbone</strong></font>
</a>
<br>
<a href="https://zdou0830.github.io/">
Zi-Yi Dou*‡
</a>
<strong><a href="https://www.semanticscholar.org/author/Aishwarya-Kamath/46174952">Aishwarya
Kamath*◊</a></strong>,
<a href="https://zhegan27.github.io/">
Zhe Gan*†
</a>
(*equal contribution)
<a href="https://pzzhang.github.io/pzzhang/">
Pengchuan Zhang†
</a>
<a href="https://scholar.google.com/citations?user=vJWEw_8AAAAJ&hl=en">
Jianfeng Wang†
</a>
<a href="https://scholar.google.com/citations?user=WR875gYAAAAJ&hl=en">
Linjie Li†
</a>
<a href="https://www.microsoft.com/en-us/research/people/zliu/?from=http%3A%2F%2Fresearch.microsoft.com%2F%7Ezliu">
Zicheng Liu†
</a>
<a href="https://people.csail.mit.edu/celiu/">
Ce Liu†
</a>
<a href="https://scholar.google.com/citations?hl=en&user=WLN3QrAAAAAJ">
Yann LeCun◊
</a>
<a href="https://vnpeng.net/">
Nanyun Peng‡
</a>
<a href="https://www.microsoft.com/en-us/research/people/jfgao/">
Jianfeng Gao†
</a>
<a href="https://www.microsoft.com/en-us/research/people/lijuanw/">
Lijuan Wang†
</a>
<br>
<em>NeurIPS 2022</em>
<br>
<a href="https://ashkamath.github.io/FIBER_page">Project page</a>
/
<a href="https://arxiv.org/abs/2206.07643">Paper</a>
/
<a href="https://github.com/microsoft/FIBER">Code & Model weights</a>
<p></p>
<p>We present FIBER (Fusion In-the-Backbone transformER) a novel Vision and Language architecture that performs deep multi-modal fusion. We also propose a new Vision-Language Pre-training (VLP) strategy, that first learns through coarse-grained image level objectives, and then obtains better fine-grained understanding capabilties by training on image-text-box data. While previous work required pseudo-annotating large amounts of image-text data to boost performance on fine-grained reasoning tasks, we show that we can equal and often surpass these results using our two-stage approach, using 25x less box annotated data. This opens the doors to scale up fine-grained models in an efficient manner without resorting to high resolution training using box annotated data. Our improved architecture also obtains state of the art performance on VQAv2, NLVR2, COCO captioning and Image-text Retrieval while being more efficient in terms of training time and memory than existing coarse and fine-grained models having similar performance.</p>
</td>
</tr>
<tr>
<td style="padding:20px;width:50%;vertical-align:middle">
<div class="one">
<img src='images/mdetr/pink.jpg' width="500">
</div>
</td>
<td style="padding:20px;width:50%;vertical-align:middle">
<a href="https://ashkamath.github.io/mdetr_page">
<font color="black"><strong>MDETR - Modulated Detection for End-to-End Multi-modal
Understanding</strong></font>
</a>
<br>
<strong><a href="https://www.semanticscholar.org/author/Aishwarya-Kamath/46174952">Aishwarya
Kamath</a></strong>,
<a href="https://scholar.google.com/citations?hl=en&user=QOO8OCcAAAAJ">Mannat Singh</a>,
<a href="https://scholar.google.com/citations?hl=en&user=WLN3QrAAAAAJ">Yann LeCun</a>,
<a href="https://scholar.google.com/citations?hl=en&user=wN9rBkcAAAAJ">Gabriel Synnaeve</a>,
<a href="https://scholar.google.com/citations?hl=en&user=WvufSLAAAAAJ">Ishan Misra</a>,
<a href="https://scholar.google.com/citations?hl=en&user=h8u3ll8AAAAJ">Nicolas Carion</a>
<br>
<em>ICCV 2021</em>,   <font color="red"><strong>(Oral Presentation, top 3% of
submissions)</strong></font>
<br>
<a href="https://ashkamath.github.io/mdetr_page">Project page</a>
/
<a href="https://arxiv.org/pdf/2104.12763">Paper</a>
/
<a href="https://github.com/ashkamath/mdetr">Code & Model weights</a>
/
<a href="https://colab.research.google.com/github/ashkamath/mdetr/blob/colab/notebooks/MDETR_demo.ipynb">Colab</a>
<p></p>
<p>We step away from existing approaches to multi-modal understanding that involve frozen
pre-trained object
detectors trained on a fixed label set, and instead achieve true end-to-end multi-modal
understanding by
<strong>detecting objects that are referred to in free form text</strong>. You can now detect and reason over
novel
combination of object classes and attributes like "a pink elephant"!</p>
</td>
</tr>
<tr>
<td style="padding:20px;width:50%;vertical-align:middle">
<div class="one">
<img src='images/adapter_fusion/adapterfusion.png' width="300">
</div>
</td>
<td style="padding:20px;width:50%;vertical-align:middle">
<font color="black"><strong>AdapterFusion: Non-Destructive Task Composition for Transfer
Learning</strong></font>
<br>
<a href="https://www.semanticscholar.org/author/Jonas-Pfeiffer/153733568">Jonas Pfeiffer</a>,
<strong><a href="https://www.semanticscholar.org/author/Aishwarya-Kamath/46174952">Aishwarya
Kamath</a></strong>,
<a href="https://www.semanticscholar.org/author/Andreas-R%C3%BCckl%C3%A9/22240011">Andreas
Rücklé</a>,
<a href="https://www.semanticscholar.org/author/Kyunghyun-Cho/1979489">Kyunghyun Cho</a>,
<a href="https://www.semanticscholar.org/author/Iryna-Gurevych/1730400">Iryna Gurevych</a>
<br>
<em>EACL 2021</em>,   <font color="red"><strong>(Oral Presentation)</strong></font>
<br>
<a href="https://adapterhub.ml/">Project page & Code</a>
/
<a href="https://aclanthology.org/2021.eacl-main.39/">Paper</a>
/
<a href="https://colab.research.google.com/github/Adapter-Hub/adapter-transformers/blob/master/notebooks/03_Adapter_Fusion.ipynb">Colab</a>
<p>We propose a <strong>new transfer learning algorithm </strong> that combines skills learned from multiple tasks in a
non-destructive manner.</p>
</td>
</tr>
<tr>
<td style="padding:20px;width:50%;vertical-align:middle">
<div class="one">
<img src='images/akbc/poster.png' width="500">
</div>
</td>
<td style="padding:20px;width:50%;vertical-align:middle">
<font color="black"><strong>A Survey on Semantic Parsing</strong></font>
<br>
<strong><a href="https://www.semanticscholar.org/author/Aishwarya-Kamath/46174952">Aishwarya
Kamath</a></strong>,
<a href="https://scholar.google.com/citations?user=FKoKAwIAAAAJ&hl=en">Rajarshi Das</a>
<br>
<em>AKBC 2019</em>
<br>
<a href="https://openreview.net/forum?id=HylaEWcTT7">Paper</a>
<p>A brief history of semantic parsing, with pointers to several seminal works. Check out the poster for
a TL;DR.</p>
</td>
</tr>
<tr>
<td style="padding:20px;width:50%;vertical-align:middle">
<div class="one">
<img src='images/postle/postle.png' width="500">
</div>
</td>
<td style="padding:20px;width:50%;vertical-align:middle">
<font color="black"><strong>Specializing Distributional Vectors of All Words for Lexical
Entailment</strong></font>
<br>
<strong><a href="https://www.semanticscholar.org/author/Aishwarya-Kamath/46174952">Aishwarya
Kamath*</a></strong>,
<a href="https://www.semanticscholar.org/author/Jonas-Pfeiffer/153733568">Jonas Pfeiffer*</a>,
<a href="https://www.semanticscholar.org/author/E.-Ponti/3381663">Edoardo M. Ponti</a>,
<a href="https://www.semanticscholar.org/author/Goran-Glavas/2472657">Goran Glavaš</a>,
<a href="https://www.semanticscholar.org/author/Ivan-Vulic/1747849">, Ivan Vulic´</a>
<br>
<em>Representation Learning for NLP Workshop, ACL 2019</em>,   <font color="red"><strong>(Best
Paper Award)</strong></font>
<br>
<a href="https://aclanthology.org/W19-4310/">Paper</a>
<p>We present the first word embedding <strong> post-processing method that specializes vectors of all vocabulary words</strong> –
including those unseen in the resources – for the asymmetric relation of lexical entailment (LE)
(i.e., hyponymy-hypernymy relation). We report consistent gains over state-of-the-art
LE-specialization methods, and successfully LE-specialize word vectors for <strong>languages without
any external lexical knowledge</strong>.</p>
</td>
</tr>
<tr>
<td style="padding:20px;width:50%;vertical-align:middle">
<div class="one">
<img src='images/spen/spen.png' width="500">
</div>
</td>
<td style="padding:20px;width:50%;vertical-align:middle">
<font color="black"><strong>Training Structured Prediction Energy Networks with Indirect
Supervision</strong></font>
<br>
<a href="https://www.semanticscholar.org/author/Pedram-Rooshenas/2844347">Amirmohammad
Rooshenas</a>,
<strong><a href="https://www.semanticscholar.org/author/Aishwarya-Kamath/46174952">Aishwarya
Kamath</a></strong>,
<a href="https://www.semanticscholar.org/author/A.-McCallum/143753639">Andrew McCallum</a>,
<br>
<em>NAACL 2018</em>,   <font color="red"><strong>(Oral Presentation)</strong></font>
<br>
<a href="https://aclanthology.org/N18-2021/">Paper</a>
<p>We train a structured prediction energy network (SPEN) <strong>without any labeled data instances</strong>, where the only source of supervision is a simple human-written scoring function. </p>
</td>
</tr>
<table style="width:100%;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;">
<tbody>
<tr>
<td style="padding:20px;width:100%;vertical-align:middle">
<h2>Awards</h2>
</td>
</tr>
</tbody>
</table>
<table style="width:100%;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;">
<tbody>
<tr>
<ul>
<li>
Deep Mind Fellowship 2019-2020
</li>
<li>
Center for Data Science Fellowship 2019-2024
</li>
<li>
NeurIPS travel award 2019
</li>
</ul>
</tr>
</tbody>
</table>
<table style="width:100%;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;">
<tbody>
<tr>
<td style="padding:20px;width:100%;vertical-align:middle">
<h2>Academic Service</h2>
</td>
</tr>
</tbody>
</table>
<table style="width:100%;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;">
<tbody>
<tr>
<ul>
<li>
Reviewer for NeurIPS 2019/20, AKBC 2019/21, EACL 2021, ICML 2021
</li>
<li>
Program Committee member for EurNLP 2019, DeeLIO 2020/21/22, RepL4NLP 2021
</li>
</ul>
</tr>
</tbody>
</table>
<table style="width:100%;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;">
<tbody>
<tr>
<td style="padding:20px;width:100%;vertical-align:middle">
<h2>Teaching</h2>
</td>
</tr>
</tbody>
</table>
<table style="width:100%;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;">
<tbody>
<tr>
<ul>
<li>
Teaching Assistant, <strong> Deep Learning at New York University</strong>, Fall 2020 - Course
by Kyunghyun Cho and Yann LeCun.
</li>
<li>
Teaching Assitant, <strong> Natural Language Processing at African Masters in Machine
Intelligence </strong> Spring 2021 - Course by Kyunghyun Cho.
</li>
<li>
Instructor, <strong> Computer Vision, NYU AI school</strong> Fall 2020.
</li>
</ul>
</tr>
</tbody>
</table>
</tbody>
</table>
<!-- <h2>Here's a list of my publications-</h2>
<strong><a href="https://arxiv.org/pdf/1909.04547.pdf">What do Deep Networks Like to Read?</a> </strong><br>
Jonas Pfeiffer* , Aishwarya Kamath* , Sebastian Ruder. Arxiv preprint, September 2019.
<br><br>