Online Instrumental Variable Regression with Applications to Online Linear System Identification
Authors: Arun Venkatraman, Wen Sun, Martial Hebert, J. Bagnell, Byron Boots
AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We experimentally demonstrate the efficacy of our algorithm in combination with popular no-regret online algorithms for the task of learning predictive dynamical system models and on a prototypical econometrics instrumental variable regression problem. Experiments We demonstrate the performance OIVR on a variety of dynamics benchmark and one illustrative econometrics problem. In Fig. 2, we show the convergence of the estimated At in OIVR to the A computed with IVR. As an additional performance metric, we report the observation prediction error with a constant covariance Kalman filter using At (Fig. 3) on a set of held out test trajectories. |
| Researcher Affiliation | Academia | Arun Venkatraman1, Wen Sun1, Martial Hebert1, J. Andrew Bagnell1, Byron Boots2 1Robotics Institute, Carnegie Mellon University, Pittsburgh, PA 15213 2School of Interactive Computing, Georgia Institute of Technology, Atlanta, GA 30332 {arunvenk,wensun,hebert,dbagnell}@cs.cmu.edu, bboots@cc.gatech.edu |
| Pseudocode | Yes | Algorithm 1 Batch Instrumental Variable Regression Input: Explanatory Variable Design Matrix X Rdx,n, Instrumental Variable Design Matrix Z Rdz,n, Prediction Targets Design Matrix Y Rdy,n Output: A Rdy,dx 1: M arg min M X MZ 2 F 2: X M Z 3: A arg min A Y A X 2 F 4: return A |
| Open Source Code | No | No explicit statement providing concrete access to the source code for the methodology described in this paper was found. The paper mentions an appendix for derivation, but not code. |
| Open Datasets | Yes | MG-10 The Mackey-Glass (MG) time-series is a standard dynamical modelling benchmark (Ralaivola and D Alche Buc 2004; Wingate and Singh 2006) generated from the nonlinear time-delay differential equation x(t) = bx(t) + ax(t τ) 1+x(t τ)10 . Helicopter The simulated helicopter from (Abbeel and Ng 2005) computes its dynamics in a 21-dimensional state space with a 4-dimensional control input. Airplane Flight Take Off We also consider the complex dynamics generated during a DA-42 airplane s take off in a flight simulator, X-plane (Research 2015), a well known program for training pilots. College Distance We finally consider an econometrics problem, a traditional application domain for instrumental variable regression, the College Distance vignette... (Kleiber and Zeileis 2008; Card 1993). |
| Dataset Splits | No | The paper mentions 'held out test trajectories' and 'training models', but does not provide specific details on training, validation, and test dataset splits needed for reproduction (e.g., percentages, sample counts, or predefined split citations). |
| Hardware Specification | No | No specific hardware details (e.g., exact GPU/CPU models, memory amounts, or detailed computer specifications) used for running its experiments were provided in the paper. |
| Software Dependencies | No | The paper mentions several algorithms and software like X-plane, OGD, ONS, i OGD, and FTRL, but does not provide specific version numbers for these software components or any other ancillary software dependencies required for replication. |
| Experiment Setup | No | The paper describes how inputs are constructed (e.g., 'maintain a k-step time window') and general experimental procedures, but does not provide specific experimental setup details such as concrete hyperparameter values, learning rates, batch sizes, or optimizer settings in the main text. |