Inverse Active Sensing: Modeling and Understanding Timely Decision-Making

Authors: Daniel Jarrett, Mihaela Van Der Schaar

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We show archetypical examples that exercise our framework through numerical simulation. Examples 1 2 give intuition for optimal active sensing, and 3 5 exemplify potential applications of IAS. Due to space limitation, commentary is necessarily brief; Appendix A gives more context and detail. Figure 4(a) depicts the output of our MAP and MCMC solutions, showing (relevant dimensions of) recovered estimates for optimal softmax strategies, along with the true weights.
Researcher Affiliation Academia 1Department of Mathematics, University of Cambridge, UK. 2Department of Electrical Engineering, UCLA, USA. Correspondence to: Daniel Jarrett <daniel.jarrett@maths.cam.ac.uk>.
Pseudocode Yes Algorithm 1 Posterior Sampler for IAS
Open Source Code No The paper does not provide any explicit statement about releasing code or a link to a code repository.
Open Datasets No The paper describes using 'numerical simulation' and generating 'episodes' for its examples (e.g., 'N = 300 episodes are simulated'), but does not use or provide concrete access information for a publicly available or open dataset.
Dataset Splits No The paper uses numerical simulations for illustrative examples but does not mention specific training, validation, or test splits for any dataset in the context of model evaluation.
Hardware Specification No The paper mentions 'numerical simulation' but provides no details about the specific hardware (e.g., CPU, GPU models) used to run these simulations.
Software Dependencies No The paper describes the mathematical framework and algorithms but does not mention any specific software libraries or their version numbers (e.g., Python, PyTorch, TensorFlow versions).
Experiment Setup Yes Example 3 (Differential Importance)... we simulate a random collection D of decision episodes for a Bayes-optimal softmax decision agent driven by ηa = (0.25, 0.75). N = 300 episodes are simulated for optimal softmax agents in (a) and the (biased) individual agent in (b), and N = 1000 for the (unbiased) population agent in (b); a greedy lookahead softmax agent (N = 300) is used in (c). Uniform priors P{η| } and P{ρ} are employed in all instances.