-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathresearch.html
More file actions
697 lines (611 loc) · 21 KB
/
research.html
File metadata and controls
697 lines (611 loc) · 21 KB
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
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<title>Hiring | Lee Optimization Group</title>
<style>
body {
font-family: system-ui, -apple-system, sans-serif;
line-height: 1.6;
margin: 0;
padding: 20px;
color: #333;
max-width: 800px;
margin-left: auto;
margin-right: auto;
}
header {
border-bottom: 1px solid #eee;
padding-bottom: 20px;
margin-bottom: 20px;
}
nav a {
margin-right: 15px;
text-decoration: none;
color: #007bff;
}
footer {
margin-top: 40px;
font-size: 0.9em;
color: #777;
border-top: 1px solid #eee;
padding-top: 20px;
display: flex;
align-items: center;
justify-content: space-between;
}
.footer-logos {
display: flex;
align-items: center;
gap: 12px;
}
.footer-logos img {
height: 36px;
width: auto;
object-fit: contain;
}
h3 {
margin-top: 30px;
margin-bottom: 5px; /* 제목 아래 간격 줄임 */
}
h3 + p {
margin-top: 0; /* 제목 바로 다음 문단 위 간격 제거 */
}
p {
margin-top: 0;
}
a {
text-decoration: none; /* 밑줄 제거 */
color: #007bff; /* 링크 색상 (원하는 색으로) */
}
a:hover {
text-decoration: underline; /* 마우스 올리면 밑줄 (선택사항) */
}
.research-item {
display: flex;
align-items: flex-start;
gap: 20px;
margin-bottom: 35px;
}
.research-item img {
width: 130px;
height: 130px;
object-fit: cover;
border: 1px solid #333;
padding: 5px;
}
.research-text h3 {
margin: 0 0 8px 0;
font-size: 20px;
}
.research-text p {
margin: 0;
color: #444;
}
/* Colored tags */
.tag.red { color: #d64541; font-weight: 600; }
.tag.orange { color: #f39c12; font-weight: 600; }
.tag.green { color: #27ae60; font-weight: 600; }
.tag.blue { color: #2e6bd9; font-weight: 600; }
.logo-title {
display: flex;
align-items: center;
}
/* ===== Publications Style ===== */
.paper {
margin-bottom: 22px;
}
.paper-title {
font-weight: 600;
color: #1a5fb4;
}
.paper-authors {
font-size: 0.95em;
color: #555;
}
.paper-venue {
font-size: 0.9em;
color: #777;
}
.spotlight {
color: red;
font-weight: 600;
}
.section-title {
font-weight: 700;
}
.red { color: #d64541; font-weight: 600; }
.orange { color: #f39c12; font-weight: 600; }
.green { color: #27ae60; font-weight: 600; }
.blue { color: #2e6bd9; font-weight: 600; }
.section-text p {
margin: 0;
line-height: 1.7;
}
.section img {
width: 140px;
height: auto;
object-fit: contain;
border: 1.5px solid #000;
flex-shrink: 0;
margin-top: 9px;
}
.section {
display: flex;
align-items: flex-start;
gap: 20px;
margin-bottom: 45px;
}
</style>
</head>
<body>
<header>
<h1>Lee Optimization Group</h1>
<!-- <div class="logo-title">
<img src="log_logo.png" alt="LOG Logo" class="logo">
<h1>Lee Optimization Group</h1>
</div> -->
<nav>
<a href="index.html">Home</a>
<a href="research.html">Research</a>
<a href="group.html">Members</a>
<a href="hiring.html">Hiring</a>
<a href="contact.html">Contact</a>
</nav>
</header>
<h2>Research Focus</h2>
<div class="section">
<img src="research_focus/opt-model.png" alt="Model Compression">
<div class="section-text">
<p>
<span class="section-title">Optimization</span> for
<span class="red">Model Compression</span>
As modern AI models, particularly deep neural networks, grow increasingly large and computationally demanding, model compression has become essential for enabling broader and more sustainable use of AI technologies.
We develop optimization methods to compress large models so that they can run more efficiently at training and inference without significantly compromising their performance.
</p>
</div>
</div>
<div class="section">
<img src="research_focus/opt-deep.jpeg" alt="Deep Learning">
<div class="section-text">
<p>
<span class="section-title">Optimization</span> for
<span class="orange">Deep Learning</span>
Optimization has been central to the success of deep learning, yet there remain many challenges to address such as non-convexity, high dimensionality, and scalability.
We study both theoretical foundations—such as convergence guarantees and generalization properties—and practical algorithms, including stochastic, adaptive, and second-order methods, so as to come up with efficient and principled optimization methods.
</p>
</div>
</div>
<div class="section">
<img src="research_focus/opt-collab.png" alt="Collaborative Learning">
<div class="section-text">
<p>
<span class="section-title">Optimization</span> for
<span class="green">Collaborative Learning</span>
Collaborative machine learning is a promising learning paradigm that allows participants to work together to train a model without necessarily sharing their raw data.
However, frequent exchange of updates can overwhelm networks, especially with large models or slow connections.
We address such a unique set of challenges due to its distributed and often decentralized nature by developing efficient and scalable solutions.
</p>
</div>
</div>
<div class="section">
<img src="research_focus/opt-applied.png" alt="Other Applications">
<div class="section-text">
<p>
<span class="section-title">Optimization</span> for
<span class="blue">Other Applications</span>
Optimization is a versatile tool that can be applied to a wide range of machine learning applications.
We are currently focused on addressing challenges surrounding large language models, particularly those involving learning in black-box environments, continual adaptation, and multi-modal contexts.
</p>
</div>
</div>
<h2>Publications</h2>
<!-- ===== Papers Start ===== -->
<div class="paper">
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/2602.18181v1" target="_blank">
SeedFlood: A Step Toward Scalable Decentralized Training of LLMs
</a>
</div>
<div class="paper-authors">
Jihun Kim, Namhoon Lee
</div>
<div class="paper-venue">arXiv 2026</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://www.arxiv.org/abs/2602.03515" target="_blank">
Mitigating Staleness in Asynchronous Pipeline Parallelism via Basis Rotation
</a>
</div>
<div class="paper-authors">
Hyunji Jung*, Sungbin Shin*, Namhoon Lee
</div>
<div class="paper-venue">arXiv 2026</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/2510.01650" target="_blank">
The Unseen Frontier: Pushing the Limits of LLM Sparsity with Surrogate-Free ADMM
</a>
</div>
<div class="paper-authors">
Kwanhee Lee, Hyeondo Jang, Dongyeop Lee, Dan Alistarh, Namhoon Lee
</div>
<div class="paper-venue">ICLR 2026</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/2510.25798" target="_blank">
MemIEC: A Step Toward Continual and Compositional Knowledge Editing
</a>
</div>
<div class="paper-authors">
Jin Seong*, Jiyun Park*, Wencke Liermann, Hongseok Choi, Yoonji Nam, Hyun Kim, Soojong Lim, Namhoon Lee
</div>
<div class="paper-venue">NeurIPS 2025</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/2505.16705" target="_blank">
An Analysis of Concept Bottleneck Models: Measuring, Understanding, and Mitigating the Impact of Noisy Annotations
</a>
</div>
<div class="paper-authors">
Seonghwan Park, Jueun Mun, Donghyun Oh, Namhoon Lee
</div>
<div class="paper-venue">NeurIPS 2025</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/2311.17539" target="_blank">
Critical Influence of Overparameterization on Sharpness-aware Minimization
</a>
</div>
<div class="paper-authors">
Sungbin Shin*, Dongyeop Lee*, Maksym Andriushchenko, Namhoon Lee
</div>
<div class="paper-venue">UAI 2025; ICML 2023 Workshop on High-dimensional Learning Dynamics</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/2506.06866" target="_blank">
SAFE: Finding Sparse and Flat Minima to Improve Pruning
</a>
</div>
<div class="paper-authors">
Dongyeop Lee, Kwanhee Lee, Jinseok Chung, Namhoon Lee
</div>
<div class="paper-venue">ICML 2025 <span class="spotlight">(Spotlight)</span></div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/2502.18153" target="_blank">
SASSHA: Sharpness-aware Adaptive Second-order Optimization with Stable Hessian Approximation
</a>
</div>
<div class="paper-authors">
Dahun Shin*, Dongyeop Lee*, Jinseok Chung, Namhoon Lee
</div>
<div class="paper-venue">ICML 2025</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/2504.06838" target="_blank">
ZIP: An Efficient Zeroth-order Prompt Tuning for Black-box Vision-Language Models
</a>
</div>
<div class="paper-authors">
Seonghwan Park, Jaehyeon Jeong, Yongjun Kim, Jaeho Lee, Namhoon Lee
</div>
<div class="paper-venue">ICLR 2025</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/2406.15524" target="_blank">
Rethinking Pruning Large Language Models: Benefits and Pitfalls of Reconstruction Error Minimization
</a>
</div>
<div class="paper-authors">
Sungbin Shin, Wonpyo Park, Jaeho Lee, Namhoon Lee
</div>
<div class="paper-venue">EMNLP 2024</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/2302.10494" target="_blank">
The Role of Masking for Efficient Supervised Knowledge Distillation of Vision Transformers
</a>
</div>
<div class="paper-authors">
Seungwoo Son, Jegwang Ryu, Namhoon Lee, Jaeho Lee
</div>
<div class="paper-venue">ECCV 2024; ICLR 2023 Workshop</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://opt-ml.org/papers.html" target="_blank">
Pruning Neural Networks with Velocity-Constrained Optimization
</a>
</div>
<div class="paper-authors">
Donghyun Oh*, Jinseok Chung*, Namhoon Lee
</div>
<div class="paper-venue">NeurIPS 2023 Workshop on Optimization for Machine Learning</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://sites.google.com/view/hidimlearning/papers" target="_blank">
On the Effectiveness of Sharpness-Aware Minimization with Large Mini-batches
</a>
</div>
<div class="paper-authors">
Jinseok Chung, Seonghwan Park, Jaeho Lee, Namhoon Lee
</div>
<div class="paper-venue">ICML 2023 Workshop on High-dimensional Learning Dynamics</div>
</div>
<div class="paper">
<div class="paper-title">
<a href=https://fl-icml2023.github.io/" target="_blank">
FedFwd: Federated Learning without Backpropagation
</a>
</div>
<div class="paper-authors">
Seonghwan Park, Dahun Shin, Jinseok Chung, Namhoon Lee
</div>
<div class="paper-venue">
ICML 2023 Workshop on Federated Learning and Analytics in Practice
</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://sites.google.com/view/aihci/papers" target="_blank">
Semi-Supervised Concept Bottleneck Models
</a>
</div>
<div class="paper-authors">
Jeeon Bae, Sungbin Shin, Namhoon Lee
</div>
<div class="paper-venue">
ICML 2023 Workshop on Artificial Intelligence and Human-Computer Interaction
</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/2302.14260" target="_blank">
A Closer Look at the Intervention Procedure of Concept Bottleneck Models
</a>
</div>
<div class="paper-authors">
Sungbin Shin, Yohan Jo, Sungsoo Ahn, Namhoon Lee
</div>
<div class="paper-venue">
ICML 2023; NeurIPS 2022 Workshop on Trustworthy and Socially Responsible ML
</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://openreview.net/forum?id=IcDTYTI0Nx" target="_blank">
Almost Sure Last Iterate Convergence of Sharpness-Aware Minimization
</a>
</div>
<div class="paper-authors">
Kyungchun Nam, Jinseok Chung, Namhoon Lee
</div>
<div class="paper-venue">
ICLR 2023 Tiny Papers
</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/2304.14082" target="_blank">
JaxPruner: A Concise Library for Sparsity Research
</a>
</div>
<div class="paper-authors">
Joo Hyung Lee, Wonpyo Park, Nicole Mitchell, Jonathan Pilault, Johan Obando-Ceron,
Han-Byul Kim, Namhoon Lee, Elias Frantar, Yun Long, Amir Yazdanbakhsh,
Shivani Agrawal, Suvinay Subramanian, Xin Wang, Sheng-Chun Kao,
Xingyao Zhang, Trevor Gale, Aart Bik, Woohyun Han, Milen Ferez,
Zhonglin Han, Hong-Seok Kim, Yann Dauphin,
Gintare Karolina Dziugaite, Pablo Samuel Castro, Utku Evci
</div>
<div class="paper-venue">
CPAL 2024 <span class="spotlight">(Oral)</span>; ICLR 2023 Workshop on Sparsity in Neural Networks
</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/2110.14678" target="_blank">
Meta-Learning Sparse Implicit Neural Representations
</a>
</div>
<div class="paper-authors">
Jaeho Lee*, Jihoon Tack*, Namhoon Lee, Jinwoo Shin
</div>
<div class="paper-venue">
NeurIPS 2021; SNN 2021
</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://ora.ox.ac.uk/objects/uuid:000e9d44-0229-48a3-84b0-dc17a8e96ccf" target="_blank">
Toward Efficient Deep Learning with Sparse Neural Networks
</a>
</div>
<div class="paper-authors">
Namhoon Lee
</div>
<div class="paper-venue">
Ph.D. Thesis, University of Oxford
</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://sites.google.com/view/sedl-workshop" target="_blank">
Optimal Mini-Batch Size for Stochastic Gradient Methods
</a>
</div>
<div class="paper-authors">
Namhoon Lee, Philip H. S. Torr, Richard Hartley
</div>
<div class="paper-venue">
ICLR 2021 Workshop on Science and Engineering of Deep Learning
</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/2003.11316" target="_blank">
Understanding the Effects of Data Parallelism and Sparsity on Neural Network Training
</a>
</div>
<div class="paper-authors">
Namhoon Lee, Thalaiyasingam Ajanthan, Philip H. S. Torr, Martin Jaggi
</div>
<div class="paper-venue">
ICLR 2021
</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/2003.11316" target="_blank">
Data Parallelism in Training Sparse Neural Networks
</a>
</div>
<div class="paper-authors">
Namhoon Lee, Philip H. S. Torr, Martin Jaggi
</div>
<div class="paper-venue">
ICLR 2020 Workshop on Practical ML
</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/1906.06307" target="_blank">
A Signal Propagation Perspective for Pruning Neural Networks at Initialization
</a>
</div>
<div class="paper-authors">
Namhoon Lee, Thalaiyasingam Ajanthan, Stephen Gould, Philip H. S. Torr
</div>
<div class="paper-venue">
ICLR 2020 <span class="spotlight">(Spotlight)</span>
</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/1810.02340" target="_blank">
SNIP: Single-shot Network Pruning based on Connection Sensitivity
</a>
</div>
<div class="paper-authors">
Namhoon Lee, Thalaiyasingam Ajanthan, Philip H. S. Torr
</div>
<div class="paper-venue">ICLR 2019</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/1804.02391" target="_blank">
Learn to Pay Attention
</a>
</div>
<div class="paper-authors">
Saumya Jetley, Nicholas Lord, Namhoon Lee, Philip H. S. Torr
</div>
<div class="paper-venue">
ICLR 2018
</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://link.springer.com/article/10.1007/s11263-018-1077-3" target="_blank">
Synthesizing a Scene-Specific Pedestrian Detector and Pose Estimator for Static Video Surveillance
</a>
</div>
<div class="paper-authors">
Hironori Hattori*, Namhoon Lee*, Vishnu N. Boddeti, Fares Beainy,
Kris M. Kitani, Takeo Kanade
</div>
<div class="paper-venue">
IJCV 2018
</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="#" target="_blank">
Forecasting Interactive Dynamics of Pedestrians with Fictitious Play
</a>
</div>
<div class="paper-authors">
Wei-Chiu Ma, De-An Huang, Namhoon Lee, Kris M. Kitani
</div>
<div class="paper-venue">
CVPR 2017
</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/1704.04394" target="_blank">
DESIRE: Distant Future Prediction in Dynamic Scenes with Interacting Agents
</a>
</div>
<div class="paper-authors">
Namhoon Lee, Wongun Choi, Paul Vernaza, Christopher B. Choy, Philip H. S. Torr, Manmohan Chandraker
</div>
<div class="paper-venue">
CVPR 2017 <span class="spotlight">(Spotlight)</span>
</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://namhoonlee.github.io/doc/namhoonlee_msthesis2015_final.pdf" target="_blank">
Modeling of Dynamic Environments for Visual Forecasting of American Football Plays
</a>
</div>
<div class="paper-authors">
Namhoon Lee (committee: Kris M. Kitani, Martial Hebert, Sebastian Scherer)
</div>
<div class="paper-venue">
M.S. Thesis, Carnegie Mellon University, Dec 2015
</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://arxiv.org/abs/1604.01431" target="_blank">
Forecasting Interactive Dynamics of Pedestrians with Fictitious Play
</a>
</div>
<div class="paper-authors">
Wei-Chiu Ma, De-An Huang, Namhoon Lee, Kris M. Kitani
</div>
<div class="paper-venue">
CVPR 2017
</div>
</div>
<div class="paper">
<div class="paper-title">
<a href="https://namhoonlee.github.io/doc/LK_WACV2016.pdf" target="_blank">
Predicting Wide Receiver Trajectories in American Football
</a>
</div>
<div class="paper-authors">
Namhoon Lee, Kris M. Kitani
</div>
<div class="paper-venue">
WACV 2016 <span class="spotlight">(Oral)</span>
</div>
</div>
</div>
</div>
<footer>
<p>© 2026 LOG. All rights reserved.</p>
<div class="footer-logos">
<a href="https://www.postech.ac.kr" target="_blank"><img src="logo/postech_logo.png" alt="POSTECH"></a>
<a href="https://ai.postech.ac.kr" target="_blank"><img src="logo/gsai_logo.png" alt="GSAI"></a>
<a href="https://cse.postech.ac.kr" target="_blank"><img src="logo/cs_logo.png" alt="CS"></a>
<img src="logo/log_logo.png" alt="LOG">
</div>
</footer>
</body>
</html>