1.[PAPER: DYNAMIC NETWORK MODEL FROM PARTIAL OBSERVATIONS]04:43
2.Can evolving network be inferred and modeled without directly observing their no...05:21
3.Dyference Framework06:39
4.Online Dynamic Network Inference08:22
5.Performance Evaluation08:50
6.Conclusion09:29
7.[PAPER: STOCHASTIC NONPARAMETRIC EVENT-TENSOR DECOMPOSITION]10:04
8.Tensor Data: 10:44
9.Tensor Decomposition11:05
10.Tensor Decomposition with Time11:50
11.Stochastic Nonparametric Event-Tensor Decomposition12:49
12.Welcome to14:23
13.[PAPER: ON GANS AND GMMS]14:33
14.GAN: Sharp and realistic generated samples, but...14:56
15.NDB - A Binning-based Two-Sample Test16:05
16.A Full-image GMM (Mixture of Factor Analyzers)16:48
17.But, Can GMMs Generate Sharp Images?17:52
18.Summary18:14
19.[PAPER: GILBO: ONE METRIC MEASURE THEM ALL]19:02
20.The Problem19:35
21.The Solution20:20
22.4 Datasets21:34
23.What makes a good GILBO?22:13
24.GILBO Validation:
Consistency24:13
25.Isn't the Mutual Information going to be infinite?24:25
26.Call to Action!24:35
27.[PAPER: ISOLATING SOURCES OF DISENTANGLEMENT IN VAES]24:47
28.Disentanglement25:33
29.Regularization in VAEs26:40
30.Different Forms of Regularization in the ELBO27:33
31.Isolating Different Forms of Regularization28:02
32.Stochastic Estimation of log q( .)29:50
33.Isolating Different Forms of Regularization & TCVAE30:59
34.Evaluating Disentanglement32:12
35.Datasets Used for Quantitative Experiments33:38
36.Penalizing Only Total Correlation Works Better34:01
37.How is Independence related to Disentanglement?34:51
38.Qualitative Results35:38
39.Future Directions36:32
40.Collaborators37:27
41.Q/A37:45
42.[PAPER: SPARSE COVARIANCE MODELING IN HIGH DIMENSIONS WITH GAUSSIAN PROCESSES]39:24
43.Example: Gene Regulatory Network39:52
44.Example: GRN Inference40:06
45.Key Challenges40:16
46.Solutions40:47
47.Covariance Modeling41:12
48.Gaussian Process and A Sparse Prior41:43
49.Performance Evaluation: GRN Inference41:57
50.Performance Evaluation: Crime Event Prediction42:55
51.[PAPER: EFFICIENT HIGH DIMENSIONAL BAYESIAN OPTIMIZATION WITH ADDITIVITY AND QUA...44:12
52.Bayesian Optimization44:33
53.Challenges of high dimensions45:41
54.Main tool: Quadrature Fourier Features (QFF)46:44
55.Example47:33
56.Algorithm48:13
57.[PAPER: REGRET BOUNDS FOR META BAYESIAN OPTIMIZATION WITH AN UNKNOWN GAUSSIAN PR...49:15
58.Bayesian optimization49:33
59.Bayesian optimization with an unknown GP prior51:06
60.Meta Bayesian optimization with an unknown GP prior51:32
61.Effect of N, the number of meta training functions52:24
62.Bounding the regret of meta BO with an unknown GP prior52:58
63.Empirical results on block picking and placing54:21
64.[PAPER: ADVERSARIALLY ROBUST OPTIMIZATION WITH GAUSSIAN PROCESSES]54:42
65.Gaussian Process Optimization55:01
66.Adversarial Robust GP Optimization55:38
67.Robust Algorithm: StableOpt56:51
68.Theoretical Result57:46
69.Variations58:23
70.[PAPER: APPROXIMATE KNOWLEDGE COMPILATION BY ONLINE COLLAPSED IMPORTANCE SAMPLIN...59:36
71.Motivation1:00:16
72.Motivation: Arithmetic Circuit1:01:20
73.But they don't scale!1:02:04
74.Collapsed Sampling (Rao-Blackwell)1:02:51
75.What to Sample?1:03:28
76.Online Collapsed Sampling1:04:54
77.How?1:05:24
78.Exact Inference1:06:27
79.Collapsed Compilation1:06:45
80.What/when do we sample?1:07:43
81.Conditional Exact Inference1:09:02
82.Online Collapsed Importance Sampling1:09:14
83.Experiments1:09:40
84.Knowledge Compilation Sampling1:10:32
85.Poster: Room 210 #51:11:14
86.Q/A1:11:22
87.[PAPER: DAGS WITH NO TEARS]1:13:49
88.Background1:14:12
89.Structure Learning: Where Are We?1:14:37
90.Results: Recovering Erdos-Renyi Graph1:18:06
91.Summary1:19:03
92.[PAPER: PROXIMAL GRAPHICAL EVENT MODEL]1:19:35
93.Parameter and Structure Learning1:22:57
94.Results: Synthetic Datasets1:24:17
95.[PAPER: HETEROGENEOUS MULTI-OUTPUT GAUSSIAN PROCESS PREDICTION]1:24:39
96.Motivation1:24:54
97.Heterogeneous Likelihoods1:26:02
98.Multi-parameter priors1:26:55
99.Scalable inference1:27:08
100.Results1:27:37
101.[PAPER: GPYTORCH-BLACKBOX MATRIX-MATRIX GAUSSIAN PROCESS INFERENCE WITH GPU ACCE...1:28:35
102.Deep Learning & Rapid Prototyping1:29:04
103.Gaussian Processes (GPS) & Rapid Prototyping1:29:54
104.Blackbox Matrix-Matrix (BBMM) Inference1:31:22
105.Gpytorch A BBMM Library1:33:08
106.Summary: Gps With BBMM Inference1:33:53