-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathindex.html
executable file
·447 lines (423 loc) · 22.3 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
<!DOCTYPE HTML>
<html lang="en"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
<title>Michael Luo</title>
<meta name="author" content="Michael Luo">
<meta name="viewport" content="width=device-width, initial-scale=1">
<link rel="stylesheet" type="text/css" href="stylesheet.css">
<link rel="icon" type="image/png" href="images/seal_icon.png">
</head>
<body>
<table style="width:100%;max-width:800px;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;"><tbody>
<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>Michael Luo</name>
</p>
<p>I am a PhD student in the <a href="https://eecs.berkeley.edu/">UC Berkeley EECS</a> department advised by <a href="https://people.eecs.berkeley.edu/~istoica/">Prof. Ion Stoica</a>. My research interests are in Artificial Intelligence and Systems. I am associated with <a href="https://sky.cs.berkeley.edu//">SkyLab</a> and <a href="https://eecs.berkeley.edu/">Berkeley Artificial Intelligence Research</a> (BAIR).
</p>
<p>
Before that, I earned my M.S. in <a href="https://eecs.berkeley.edu/">EECS</a> under Ion Stoica and <a href="https://goldberg.berkeley.edu/">Ken Goldberg</a> from <a href="http://autolab.berkeley.edu/">AUTOLab</a> during 2021. I also earned a B.S. from UC Berkeley with a double major in <a href="https://eecs.berkeley.edu/">EECS</a> and <a href="https://haas.berkeley.edu/">Business Administration</a> in 2020.
</p>
<p style="text-align:center">
<a href="mailto:[email protected]">Email</a>  / 
<a href="data/Michael_Luo_CV.pdf">CV</a>  / 
<a href="https://scholar.google.com/citations?hl=en&user=XpO6-kEAAAAJ">Google Scholar</a>  / 
<a href="https://www.linkedin.com/in/michaelzhiluo/">LinkedIn</a>  / 
<a href="https://github.com/michaelzhiluo/">Github</a>
</p>
</td>
<td style="padding:2.5%;width:40%;max-width:40%">
<a href="images/michael_luo_1.jpg"><img style="width:100%;max-width:100%" alt="profile photo" src="images/michael_luo_1.jpg" class="hoverZoomLink"></a>
</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:100%;vertical-align:middle">
<heading>Research</heading>
<p>
Currently, my research involves building scalable systems for ML pracitioners that will fulfill the <a href="https://sigops.org/s/conferences/hotos/2021/papers/hotos21-s02-stoica.pdf">Sky Computing vision</a>. We are <a href="https://github.com/skypilot-org/skypilot">open source</a>. This research involves virtualizing GPUs to scale DL training to trillions of parameters and designing learnable scheduling policies for migrating jobs across different and regions clouds (incl. on-premise).
</p>
<p>
Previously,my masters's and undergraduate research primarily focused on practical problems and applications for reinforcement learning (RL), including NLP, query optimization for databases, and video streaming.
</p>
</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:25%;vertical-align:middle">
<img src="images/balsa.png" alt="project image" style="width:auto; height:auto; max-width:100%;" />
</td>
<td style="padding:20px;width:75%;vertical-align:middle">
<a href="https://arxiv.org/abs/1912.00167">
<papertitle>Balsa: Learning a Query Optimizer Without Expert Demonstrations</papertitle>
</a>
<br>
Zongheng Yang,
Wei-lin Chiang,
Frank Luan,
Gautam Mittal,
<strong>Michael Luo</strong>,
Ion Stoica
<br>
<em>Special Interest Group on Management of Data (<strong>SIGMOD</strong>), 2022</em>
<br>
<a href="https://arxiv.org/abs/2201.01441">Arxiv</a> |
<a href="https://speakerdeck.com/zongheng/balsa-learning-a-query-optimizer-without-expert-demonstrations">Video</a> |
<a href="https://github.com/balsa-project/balsa">Code</a>
<p></p>
<p> An end2end query optimizer trained via deep RL that exceeds the query-performance of expert solvers by up to 2.8x.
</p>
</td>
</tr>
<tr>
<td style="padding:20px;width:25%;vertical-align:middle">
<img src="images/mesa.png" alt="project image" style="width:auto; height:auto; max-width:100%;" />
</td>
<td style="padding:20px;width:75%;vertical-align:middle">
<a href="https://sites.google.com/view/safe-meta-rl/home">
<papertitle>MESA: Offline Meta-RL for Safe Adaptation and Fault Tolerance</papertitle>
</a>
<br>
<strong>Michael Luo</strong>,
Ashwin Balakrishna,
Brijen Thananjeyan,
Suraj Nair,
Julian Ibarz,
Jie Tan,
Chelsea Finn,
Ion Stoica,
Ken Goldberg
<br>
<em>Neural Information Processing Systems (<strong>NeurIPS</strong>) Safe Control Workshop, 2021 </em>
<br>
<a href="https://sites.google.com/view/safe-meta-rl/home">Website</a> |
<a href="https://arxiv.org/abs/2112.03575">Arxiv</a> |
<a href="https://www.youtube.com/watch?v=lK3w_SoGsyE">Video</a> |
<a href="https://github.com/michaelzhiluo/mesa-safe-rl">Code</a>
<p></p>
<p> Safe RL algorithm that meta-learns from offline datasets to safely adapt to unseen environments.
</p>
</td>
</tr>
<tr>
<td style="padding:20px;width:25%;vertical-align:middle">
<img src="images/impact.png" alt="project image" style="width:auto; height:auto; max-width:100%;" />
</td>
<td style="padding:20px;width:75%;vertical-align:middle">
<a href="https://arxiv.org/abs/1912.00167">
<papertitle>IMPACT: Importance Weighted Asynchronous Architectures with Clipped Target Networks</papertitle>
</a>
<br>
<strong>Michael Luo</strong>,
Jiahao Yao,
Richard Liaw,
Eric Liang,
Ion Stoica
<br>
<em>International Conference on Learning Representations (<strong>ICLR</strong>), 2020 </em>
<br>
<a href="https://docs.ray.io/en/latest/rllib/rllib-algorithms.html#appo">Website</a> |
<a href="https://arxiv.org/abs/1912.00167">Arxiv</a> |
<a href="https://iclr.cc/virtual_2020/poster_BJeGlJStPr.html">Video</a> |
<a href="https://github.com/ray-project/ray/blob/master/rllib/agents/ppo/appo.py">Code</a>
<p></p>
<p> An algorithm for distributed reinforcement learning that tunes the tradeoff between distributed data collection and learning sample efficiency to optimize for training speed by combining the sample efficiency of PPO and the data throughput from IMPALA.
</p>
</td>
</tr>
<tr>
<td style="padding:20px;width:25%;vertical-align:middle">
<img style="width:105%;max-width:105%" src="images/cognitive.png">
</td>
<td style="padding:20px;width:75%;vertical-align:middle">
<a href="https://arxiv.org/abs/2011.13782">
<papertitle>Connecting Context-specific Adaptation in Humans to Meta-learning</papertitle>
</a>
<br>
Rachit Dubey*,
Erin Grant*,
<strong>Michael Luo*</strong>,
Karthik Narasimhan,
Thomas L. Griffiths
<br>
<em>Preprint.</em>
<br>
<a href="https://arxiv.org/abs/2011.13782">Arxiv</a> |
<a href="https://github.com/michaelzhiluo/ray/tree/metaworld">Code</a>
<br>
<p></p>
<p>We introduce a framework for using contextual information about a task to guide the initialization of task-specific models before adaptation to online feedback, which leads to faster adaptation to online feedback than that of zero-shot multitask approaches.</p>
</td>
</tr>
<tr>
<td style="padding:20px;width:25%;vertical-align:middle">
<img style="width:105%;max-width:105%" src="images/recovery-rl.png">
</td>
<td style="padding:20px;width:75%;vertical-align:middle">
<a href="https://sites.google.com/berkeley.edu/recovery-rl/">
<papertitle>Recovery RL: Safe Reinforcement Learning with Learned Recovery Zones</papertitle>
</a>
<br>
Brijen Thananjeyan*,
Ashwin Balakrishna*,
Suraj Nair,
<strong>Michael Luo</strong>,
Krishnan Srinivasan,
Minho Hwang,
Joseph E. Gonzalez,
Julian Ibarz,
Chelsea Finn,
Ken Goldberg
<br>
<em>International Conference on Robotics and Automation (<strong>ICRA</strong>), 2021</em>
<br>
<a href="https://sites.google.com/berkeley.edu/recovery-rl/">Website</a> |
<a href="https://arxiv.org/abs/2010.15920">Arxiv</a> |
<a href="https://www.youtube.com/watch?v=XvO8fzRM0LA">Video</a> |
<a href="https://github.com/abalakrishna123/recovery-rl">Code</a>
<br>
<p></p>
<p>An algorithm for safe reinforcement learning which utilizes a set of offline data to learn about constraints before policy learning and a pair of policies which separate the often conflicting objectives of task directed exploration and constraint satisfaction to learn contact rich and visuomotor control tasks.</p>
</td>
</tr>
<tr>
<td style="padding:20px;width:25%;vertical-align:middle">
<img style="width:105%;max-width:105%" src="images/dataflow.png">
</td>
<td style="padding:20px;width:75%;vertical-align:middle">
<a href="https://proceedings.neurips.cc/paper/2021/hash/2bce32ed409f5ebcee2a7b417ad9beed-Abstract.html">
<papertitle>Distributed Reinforcement Learning is a Dataflow Problem</papertitle>
</a>
<br>
Eric Liang*,
Zhanghao Wu*,
<strong>Michael Luo</strong>,
Sven Mika,
Ion Stoica
<br>
<em>Neural Information Processing Systems (<strong>NeurIPS</strong>), 2021.</em>
<br>
<a href="https://arxiv.org/abs/2011.12719">Arxiv</a> |
<a href="https://slideslive.com/38967764/rllib-flow-distributed-reinforcement-learning-is-a-dataflow-problem?ref=recommended">Video</a> |
<a href="https://github.com/ray-project/ray/blob/master/rllib/">Code</a>
<br>
<p></p>
<p>We propose RLFlow, a hybrid actor-dataflow programming model for distributed RL, that leads to highly composable and performant implementations of RL algorithms, which results to faster training and significant code reductions.</p>
</td>
</tr>
<tr>
<td style="padding:20px;width:25%;vertical-align:middle">
<img src="images/impact.png" alt="project image" style="width:auto; height:auto; max-width:100%;" />
</td>
<td style="padding:20px;width:75%;vertical-align:middle">
<a href="https://berkeleyautomation.github.io/rlqp/">
<papertitle>Accelerating Quadratic Optimization with Reinforcement Learning</papertitle>
</a>
<br>
Jeffrey Ichnowski,
Paras Jain,
Bartolomeo Stellato,
Goran Banjac,
<strong>Michael Luo</strong>,
Francesco Borrelli,
Joseph E. Gonzalez,
Ion Stoica,
Ken Goldberg
<br>
<em>Neural Information Processing Systems (<strong>NeurIPS</strong>), 2021.</em>
<br>
<a href="https://berkeleyautomation.github.io/rlqp/">Website</a> |
<a href="https://arxiv.org/abs/2107.10847">Arxiv</a> |
<a href="https://slideslive.com/38967072/accelerating-quadratic-programming-with-reinforcement-learning?ref=speaker-18771">Video</a> |
<a href="https://github.com/BerkeleyAutomation/rlqp">Code</a>
<p></p>
<p> An intelligent application of RL that tunes the parameters of existing Quadratic Program (QP) solvers and improving solving times by up to 3x.
</p>
</td>
</tr>
<tr>
<td style="padding:20px;width:25%;vertical-align:middle">
<img style="width:105%;max-width:105%" src="images/voi.png">
</td>
<td style="padding:20px;width:75%;vertical-align:middle">
<a href="https://arxiv.org/abs/2110.15797">
<papertitle>Discovering Non-monotonic Autoregressive Orderings with Variational Inference</papertitle>
</a>
<br>
Xuanlin Li*,
Brandon Trabucco*,
Dong Huk Park,
Yang Gao,
<strong>Michael Luo</strong>,
Sheng Shen,
Trevor Darrell
<br>
<em>International Conference on Learning Representations (<strong>ICLR</strong>), 2021.</em>
<br>
<a href="https://iclr-blog-track.github.io/2022/03/25/non-monotonic-autoregressive-ordering/">Website</a> |
<a href="https://arxiv.org/abs/2110.15797">Arxiv</a> |
<a href="https://slideslive.com/38954040/discovering-nonmonotonic-autoregressive-orderings-with-variational-inference?ref=recommended">Video</a> |
<a href="https://github.com/xuanlinli17/autoregressive_inference">Code</a>
<br>
<p></p>
<p>We propose the first domain-independent unsupervised / self-supervised learner that discovers high-quality autoregressive orders through fully-parallelizable end-to-end training without domain-specific tuning.</p>
</td>
</tr>
<tr>
<td style="padding:20px;width:25%;vertical-align:middle">
<img style="width:105%;max-width:105%" src="images/lazydagger.png">
</td>
<td style="padding:20px;width:75%;vertical-align:middle">
<a href="https://sites.google.com/view/lazydagger/home">
<papertitle>LazyDAgger: Reducing Context Switching in Interactive Robot Imitation Learning</papertitle>
</a>
<br>
Ryan Hoque,
Ashwin Balakrishna,
Brijen Thanajeyan,
Carl Putterman,
<strong>Michael Luo</strong>,
Daniel Seita,
Daniel S. Brown,
Ken Goldberg
<br>
<em>Conference on Automation Science and Engineering (<strong>CASE</strong>), 2021.</em>
<br>
<a href="https://sites.google.com/view/lazydagger/home">Website</a> |
<a href="https://arxiv.org/abs/2104.00053">Arxiv</a> |
<a href="https://drive.google.com/file/d/1Z39PhREBuoy-I3jStkPZqkpEcVSNhxKx/view">Video</a>
<br>
<p></p>
<p>An algorithm for interactive imitation learning that learns to minimize human context switching through sustained interventions and maintains the same supervisor burden for prior algorithms.</p>
</td>
</tr>
<tr>
<td style="padding:20px;width:25%;vertical-align:middle">
<img style="width:105%;max-width:105%" src="images/garden.png">
</td>
<td style="padding:20px;width:75%;vertical-align:middle">
<a href="http://alphagarden.org/">
<papertitle>AlphaGarden: Learning Seed Placement and Automation Policies For Polyculture Farming with Companion Plants</papertitle>
</a>
<br>
Yahav Avigal, Anna Deza, William Wong, Sebastian Oehme, Mark Presten, Mark Theis, Jackson Chui, Paul Shao, Huang Huang, Atsunobu Kotani, Satvik Sharma,
<strong>Michael Luo</strong>,
Stefano Carpin, Joshua Viers, Stavros Vougioukas, Ken Goldberg
<br>
<em>International Conference on Robotics and Automation (<strong>ICRA</strong>), 2021</em>
<br>
<a href="http://alphagarden.org/">Website</a> |
<a href="https://ieeexplore.ieee.org/document/9561431">Paper</a> |
<a href="https://github.com/BerkeleyAutomation/AlphaGarden">Code</a>
<br>
<p></p>
<p>We investigate different seed placement and pruning algoritms in a polyculture garden simulator to jointly maximize diveristy and coverage of various plants types.</p>
</td>
</tr>
</tbody></table>
<table width="100%" align="center" border="0" cellspacing="0" cellpadding="20"><tbody>
<tr>
<td>
<heading>Open-source Projects</heading>
</td>
</tr>
</tbody></table>
<table width="100%" align="center" border="0" cellpadding="20"><tbody>
<tr>
<td style="padding:20px;width:25%;vertical-align:middle"><img style="width:105%;max-width:105%" src="images/skypilot.png"></td>
<td width="75%" valign="center">
<a href="https://github.com/skypilot-org/skypilot">
<papertitle>SkyPilot: A Broker for Sky Computing</papertitle>
</a>
<br>
Core contributor; Developed Sky Storage, Sky On-premise, and multi-node core features.
</td>
</tr>
<tr>
<td style="padding:20px;width:25%;vertical-align:middle"><img style="width:105%;max-width:105%" src="images/rllib.png"></td>
<td width="75%" valign="center">
<a href="https://github.com/ray-project/ray">
<papertitle>Ray/RLlib</papertitle>
</a>
<br>
Core contributor; Created distributed model-free, model-based, and meta-learning RL algorithms on RLlib, including APPO/IMPACT, MAML, MBMPO, and Google Dreamer.
</td>
</tr>
</tbody></table>
<table width="100%" align="center" border="0" cellspacing="0" cellpadding="20"><tbody>
<tr>
<td>
<heading>Teaching</heading>
</td>
</tr>
</tbody></table>
<table width="100%" align="center" border="0" cellpadding="20"><tbody>
<tr>
<td style="padding:20px;width:25%;vertical-align:middle">
<img style="width:105%;max-width:105%" src="images/berkeley_eecs.png">
</td>
<td width="75%" valign="center">
<p>
<b>CS 189: Introduction to Machine Learning</b> <br>
Teaching Assistant: <a href="https://inst.eecs.berkeley.edu/~cs189/fa19/">Fall 2019</a>
</p>
<p>
<b>CS 162: Operating Systems and System Programming</b> <br>
Reader: <a href="https://inst.eecs.berkeley.edu/~cs162/fa18/">Fall 2018</a>
</p>
</td>
</tr>
</tbody></table>
<table width="100%" align="center" border="0" cellspacing="0" cellpadding="20"><tbody>
<tr>
<td>
<heading>Work Experience</heading>
</td>
</tr>
</tbody></table>
<table width="100%" align="center" border="0" cellpadding="20"><tbody>
<tr>
<td style="padding:20px;width:25%;vertical-align:middle"><img style="width:105%;max-width:105%" src="images/anyscale.png"></td>
<td width="75%" valign="center">
<a href="https://www.anyscale.com/">Anyscale</a>
<br>
Software Development Intern
</td>
</tr>
<tr>
<td style="padding:20px;width:25%;vertical-align:middle"><img style="width:105%;max-width:105%" src="images/amazon.jpg"></td>
<td width="75%" valign="center">
<a href="https://www.amazon.com/">Amazon</a>
<br>
Software Development Intern
</td>
</tr>
<tr>
<td style="padding:20px;width:25%;vertical-align:middle"><img style="width:105%;max-width:105%" src="images/meraki.png"></td>
<td width="75%" valign="center">
<a href="https://meraki.cisco.com/">Cisco Meraki</a>
<br>
Computer Vision Intern
</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:0px">
<br>
<p style="text-align:right;font-size:small;">
<a href="https://github.com/jonbarron/jonbarron_website">Website template from Jon Barron</a>
</p>
</td>
</tr>
</tbody></table>
</td>
</tr>
</table>
</body>
</html>