-
Notifications
You must be signed in to change notification settings - Fork 24
/
publication.html
260 lines (216 loc) · 11.6 KB
/
publication.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
---
layout: page
title: "Publications"
description: ""
header-img: "img/about-bg.jpg"
order: 2
---
<div id="aboutme" class="row">
<div class="col-md-12">
<div class="row">
<div class="col-md-2">
<img class="float-left" alt="Agustinus's Portrait" src="/img/portrait.jpg"
data-holder-rendered="true" style="width: 150px; border-radius: 50% !important;">
</div>
<div class="col-md-10" style="padding-top: 0.25em">
<h1 class="section-header" style="padding: 0;">Agustinus Kristiadi</h1>
<h4 class="section-header" style="padding: 0;">Postdoc at the Vector Institute, Toronto</h4>
</div>
</div>
<div class="row" style="padding-top: 3em;">
<div class="col-md-3"><h3>Conference</h3></div>
<div class="col-md-9">
<h4>Structured Inverse-Free Natural Gradient: Memory-Efficient & Numerically-Stable KFAC for Large Neural Nets</h4>
<div>
Wu Lin*, Felix Dangel*, Runa Eschenhagen, Kirill Neklyudov, <i>Agustinus Kristiadi</i>, Richard Turner, Alireza Makhzani
</div>
ICML 2024
[<a href="https://arxiv.org/abs/2312.05705">arxiv</a>]
[<a href="https://github.com/f-dangel/SINGD">code</a>]
<div class="row" style="padding-bottom: 1em"></div>
<h4>A Sober Look at LLMs for Material Discovery: Are They Actually Good for Bayesian Optimization Over Molecules?</h4>
<div>
<i>Agustinus Kristiadi</i>, Felix Strieth-Kalthoff, Marta Skreta, Pascal Poupart, Alán Aspuru-Guzik, Geoff Pleiss
</div>
ICML 2024
[<a href="https://arxiv.org/abs/2402.05015">arxiv</a>]
[<a href="https://github.com/wiseodd/lapeft-bayesopt">code</a>]
<div class="row" style="padding-bottom: 1em"></div>
<h4>Position Paper: Bayesian Deep Learning in the Age of Large-Scale AI</h4>
<div>
Theodore Papamarkou, Maria Skoularidou, Konstantina Palla, Laurence Aitchison, Julyan Arbel, David Dunson, Maurizio Filippone, Vincent Fortuin, Philipp Hennig, Jose Miguel Hernandez Lobato, Aliaksandr Hubin, Alexander Immer, Theofanis Karaletsos, Mohammad Emtiyaz Khan, <i>Agustinus Kristiadi</i>, Yingzhen Li, Stephan Mandt, Christopher Nemeth, Michael A. Osborne, Tim G. J. Rudner, David Rügamer, Yee Whye Teh, Max Welling, Andrew Gordon Wilson, Ruqi Zhang
</div>
ICML 2024
[<a href="https://arxiv.org/abs/2402.00809">arxiv</a>]
<div class="row" style="padding-bottom: 1em"></div>
<h4>Preventing Arbitrarily High Confidence on Far-Away Data in Point-Estimated Discriminative Neural Networks</h4>
<div>
Ahmad Rashid, Serena Hacker, Guojun Zhang, <i>Agustinus Kristiadi</i>, Pascal Poupart
</div>
AISTATS 2024
[<a href="https://arxiv.org/abs/2311.03683">arxiv</a>]
[<a href="https://github.com/serenahacker/preload">code</a>]
<div class="row" style="padding-bottom: 1em"></div>
<h4>The Geometry of Neural Nets' Parameter Spaces Under Reparametrization</h4>
<div>
<i>Agustinus Kristiadi</i>, Felix Dangel, and Philipp Hennig
</div>
NeurIPS 2023 <b style="color: crimson">[Spotlight]</b>
[<a href="https://arxiv.org/abs/2302.07384">arxiv</a>]
<div class="row" style="padding-bottom: 1em"></div>
<h4>Posterior Refinement Improves Sample Efficiency in Bayesian Neural Networks</h4>
<div>
<i>Agustinus Kristiadi</i>, Runa Eschenhagen, and Philipp Hennig
</div>
NeurIPS 2022
[<a href="https://arxiv.org/abs/2205.10041">arxiv</a>]
[<a href="https://github.com/runame/laplace-refinement">code</a>]
<div class="row" style="padding-bottom: 1em"></div>
<h4>Fast Predictive Uncertainty for Classification with Bayesian Deep Networks</h4>
<div>
Marius Hobbhahn, <i>Agustinus Kristiadi</i>, and Philipp Hennig
</div>
UAI 2022
[<a href="https://proceedings.mlr.press/v180/hobbhahn22a.html">paper</a>]
[<a href="https://github.com/mariushobbhahn/LB_for_BNNs_official">code</a>]
<div class="row" style="padding-bottom: 1em"></div>
<h4>Being a Bit Frequentist Improves Bayesian Neural Networks</h4>
<div>
<i>Agustinus Kristiadi</i>, Matthias Hein, and Philipp Hennig
</div>
AISTATS 2022
[<a href="https://proceedings.mlr.press/v151/kristiadi22a.html">paper</a>]
[<a href="https://github.com/wiseodd/bayesian_ood_training">code</a>]
<div class="row" style="padding-bottom: 1em"></div>
<h4>Discovering Inductive Bias with Gibbs Priors: A Diagnostic Tool for Approximate Bayesian Inference</h4>
<div>
Luca Rendsburg, <i>Agustinus Kristiadi</i>, Philipp Hennig, and Ulrike von Luxburg
</div>
AISTATS 2022
[<a href="https://proceedings.mlr.press/v151/rendsburg22a.html">paper</a>]
[<a href="https://github.com/tml-tuebingen/gibbs-prior-diagnostic">code</a>]
<div class="row" style="padding-bottom: 1em"></div>
<h4>An Infinite-Feature Extension for Bayesian ReLU Nets That Fixes Their Asymptotic Overconfidence</h4>
<div>
<i>Agustinus Kristiadi</i>, Matthias Hein, and Philipp Hennig
</div>
<div>
NeurIPS 2021 <b style="color: crimson">[Spotlight]</b>
[<a href="https://proceedings.neurips.cc/paper/2021/hash/9be40cee5b0eee1462c82c6964087ff9-Abstract.html">paper</a>]
[<a href="https://github.com/wiseodd/rgpr">code</a>]
</div>
<div class="row" style="padding-bottom: 1em"></div>
<h4>Laplace Redux -- Effortless Bayesian Deep Learning</h4>
<div>
Erik Daxberger*, <i>Agustinus Kristiadi</i>*, Alexander Immer*, Runa Eschenhagen*, Matthias Bauer, and Philipp Hennig
</div>
<div>
NeurIPS 2021
[<a href="https://proceedings.neurips.cc/paper/2021/hash/a7c9585703d275249f30a088cebba0ad-Abstract.html">paper</a>]
[<a href="https://github.com/AlexImmer/Laplace">code</a>]
</div>
<div class="row" style="padding-bottom: 1em"></div>
<h4>Learnable Uncertainty under Laplace Approximations</h4>
<div>
<i>Agustinus Kristiadi</i>, Matthias Hein, and Philipp Hennig
</div>
<div>
UAI 2021 <b style="color: crimson">[Long Talk]</b>
[<a href="https://proceedings.mlr.press/v161/kristiadi21a.html">paper</a>]
[<a href="https://github.com/wiseodd/lula">code</a>]
</div>
<div class="row" style="padding-bottom: 1em"></div>
<h4>Being Bayesian, Even Just a Bit, Fixes Overconfidence in ReLU Networks</h4>
<div>
<i>Agustinus Kristiadi</i>, Matthias Hein, and Philipp Hennig
</div>
<div>
ICML 2020
[<a href="http://proceedings.mlr.press/v119/kristiadi20a.html">paper</a>]
[<a href="https://github.com/wiseodd/last_layer_laplace">code</a>]
</div>
<div class="row" style="padding-bottom: 1em"></div>
<h4>Incorporating Literals into Knowledge Graph Embeddings</h4>
<div>
<i>Agustinus Kristiadi</i>*, Mohammad Asif Khan*, Denis Lukovnikov, Jens Lehmann, and Asja Fischer
</div>
<div>
ISWC 2019
[<a href="https://arxiv.org/abs/1802.00934">arxiv</a>]
[<a href="https://github.com/SmartDataAnalytics/LiteralE">code</a>]
</div>
<div class="row" style="padding-bottom: 1em"></div>
<h4>Improving Response Selection in Multi-turn Dialogue Systems by Incorporating Domain Knowledge</h4>
<div>
Debanjan Chauduri, <i>Agustinus Kristiadi</i>, Jens Lehmann, Asja Fischer
</div>
<div>
CoNLL 2018
[<a href="https://arxiv.org/abs/1809.03194">arxiv</a>]
[<a href="https://github.com/SmartDataAnalytics/AK-DE-biGRU">code</a>]
</div>
<div class="row" style="padding-bottom: 1em"></div>
<h4>Deep Convolutional Level Set Method for Image Segmentation</h4>
<div>
<i>Agustinus Kristiadi</i> and Pranowo
</div>
<div>
Journal of ICT Research and Applications 11.3 (2017)
[<a href="http://journals.itb.ac.id/index.php/jictra/article/download/3887/3046">pdf</a>]
[<a href="https://github.com/wiseodd/cnn-levelset">code</a>]
</div>
<div class="row" style="padding-bottom: 1em"></div>
<h4>Parallel Particle Swarm Optimization for Image Segmentation</h4>
<div>
<i>Agustinus Kristiadi</i>, Pranowo, and Paulus Mudjihartono
</div>
<div>
DEIS 2013
[<a href="http://sdiwc.net/digital-library/web-admin/upload-pdf/00000491.pdf">pdf</a>]
[<a href="https://github.com/wiseodd/swarm-image-segmenter">code</a>]
</div>
</div>
</div>
<div class="row" style="padding-top: 3em;">
<div class="col-md-3"><h3>Workshop</h3></div>
<div class="col-md-9">
<h4>Promises and Pitfalls of the Linearized Laplace in Bayesian Optimization
</h4>
<div>
<i>Agustinus Kristiadi</i>, Alexander Immer, Runa Eschenhagen, and Vincent Fortuin
</div>
AABI, ICML 2023
[<a href="https://arxiv.org/abs/2304.08309">arxiv</a>]
<div class="row" style="padding-bottom: 1em"></div>
<h4>Mixtures of Laplace Approximations for Improved <i>Post-Hoc</i> Uncertainty in Deep Learning</h4>
<div>
Runa Eschenhagen, Erik Daxberger, Philipp Hennig, <i>Agustinus Kristiadi</i>
</div>
Bayesian Deep Learning Workshop, NeurIPS 2021
[<a href="https://arxiv.org/abs/2111.03577">arxiv</a>]
<div class="row" style="padding-bottom: 1em"></div>
<h4>Predictive Uncertainty Quantification with Compound Density Networks</h4>
<div>
<i>Agustinus Kristiadi</i>, Sina Däubener, and Asja Fischer
</div>
<div>
Bayesian Deep Learning Workshop, NeurIPS 2019
[<a href="https://arxiv.org/abs/1902.01080">arxiv</a>]
[<a href="https://github.com/wiseodd/compound-density-networks">code</a>]
</div>
</div>
</div>
<div class="row" style="padding-top: 3em;">
<div class="col-md-3"><h3>Preprint</h3></div>
<div class="col-md-9">
<h4>On the Disconnect Between Theory and Practice of Overparametrized Neural Networks</h4>
<div>
Jonathan Wenger, Felix Dangel, <i>Agustinus Kristiadi</i>
</div>
ArXiv
[<a href="https://arxiv.org/abs/2310.00137">arxiv</a>]
<div class="row" style="padding-bottom: 1em"></div>
</div>
</div>
</div>
</div>