1.Welcome back to ICML 2019 presentations.
This session on Causality includes:00:00
2.[Paper: Causal Identification under Markov Equivalence: Completeness Results]03:34
3.Motivation04:15
4.Fundamental Problem of Solution: Causal Graph Which analysis should we believe?06:19
5.Anatomy of Causal Identification09:19
6.Challenges for Identification12:33
7.Can a causal diagram be learned from data?13:20
8.Data-Driven Causal Identification15:30
9.Result 1. Graphical
Properties & Factorization17:12
10.C-Component17:41
11.Possible C-Component19:10
12.Q-Factorization in DAGs-PAGs20:23
13.Q-Factorization in PAGs20:28
14.Result 2.
Identification Algorithm20:48
15.IDP - Example21:02
16.IDP - Completeness21:08
17.Conclusions21:57
18.Q&A22:23
19.[Paper: Counterfactual Off-Policy Evaluation with Linear Structural Causal Model...22:54
20.Motivation: Building trust in RL policies23:02
21.Using counterfactuals to "sanity check"23:59
22.Simulating counterfactual trajectories26:02
23.So, what should we use for the structural
causal model (SCM)?26:48
24.[Paper: Causal Discovery and Forecasting in Nonstationary Environmentswith State...27:51
25.Two tasks:28:24
26.Time-varying causal model:29:11
27.Causal Model, Identifiability, and Estimation30:02
28.Forecasting with time-varying causal model30:49
29.Causal discovery:31:11
30.Macroeconomics data31:50
31.Conclusion32:12
32.[Paper: Classifying Treatment Responders Under Causal Effect Monotonicity]32:54
33.Heterogeneous Treatment Effect Estimation33:05
34.Often Outcome is Binary33:37
35.Often We Want to Predict Response33:50
36.Classifying Responders: The Problem34:16
37.Monotonicity34:54
38.Classifying Responders35:52
39.Empirical Results: Synthetic36:39
40.Empirical Results: Census Data36:59
41.[Paper: Learning Models from Data with Measurement Error: Tackling Underreportin...37:43
42.Introduction37:51
43.Model38:36
44.Identifiability39:55
45.Maternal drug use and childhood obesity40:57
46.Thanks!41:23
47.[Paper: Adjustment Criteria for Generalizing Experimental Findings]42:00
48.Causal Effects and Experiments42:30
49.Motivating Example (1)
(Why is this problem non-trivial?)43:19
50.Controlled Experimentation - Randomization44:24
51.Randomization is not all there is!
Motivating Example (2)45:27
52.What are we missing?
Motivating Example (3)47:03
53.Two Challenges47:35
54.Formalizing the Problem49:24
55.Problem Statement50:47
56.Related Work51:10
57.Solution: Covariate st-Adjustment52:14
58.Challenge I. Covariate Admissibility52:51
59.Main Result I:
Complete Graphical Condition53:52
60.Understanding the criterion54:33
61.Getting the intuition behind the rules55:17
62.Challenge II.
Searching for Admissible Sets55:50
63.Main Result II: Listing Algorithm56:35
64.Conclusions57:35
65.Thank you!58:35
66.[Paper: Conditional Independence in Testing Bayesian Networks]1:01:09
67.Fuse Knowledge with Expressiveness1:01:18
68.Testing Bayesian Network1:03:37
69.Conditional Independence in TBN1:04:39
70.Fuse Network with Expressiveness1:05:13
71.[Paper: Sensitivity Analysis of Linear Structural Causal Models]1:05:55
72.Motivating example: smoking and cancer1:06:18
73.In summary: why sensitivity analysis?1:08:17
74.Current sensitivity analysis literature1:08:49
75.A systematic approach for sensitivity analysis1:09:21
76.For details, come to our poster session:1:10:05
77.[Paper: More efficient Off-Policy Evaluation through Regularized Targeted Learni...1:10:33
78.Problem statement1:10:49
79.Formalization1:11:31
80.Our base estimator1:11:36
81.Our ensemble estimator1:13:29
82.Empirical performance1:13:41
83.[Paper: Inferring Heterogeneous Casual Effects in Presence of Spatial Confoundin...1:15:21
84.Causal inference problem1:15:32
85.Approach1:17:04
86.Error-invariables model1:17:50
87.Proposed robust method1:18:39
88.Real data1:19:28
89.Conclusion1:20:04