Imitation Learning from Visual Data with Multiple Intentions

Authors: Aviv Tamar, Khashayar Rohanimanesh, Yinlam Chow, Chris Vigorito, Ben Goodrich, Michael Kahane, Derik Pridmore

ICLR 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate the effectiveness of our approach on learning a reaching task with the IRB-120 robot (Figure 1b)... Table 1 shows the overall success rate for every model across all 200 trials.
Researcher Affiliation Collaboration 1EECS Department, UC Berkeley 2Osaro Inc.
Pseudocode Yes Algorithm 1: IDS Input: A minibatch of K samples {ut, xt, . . . , ut+K, xt+K} from the same demonstration trajectory T i Output: An update direction for θ, and a sample from Qα 1 Sample z1, . . . , z N P(z) 2 Set z = arg maxzi P(ut:t+K|xt:t+K, zi; θ) 3 return θ log P(ut:t+K|xt:t+K, z ; θ), and z
Open Source Code No The paper does not explicitly state that the source code for the methodology is openly available or provide a link to a code repository. It only provides a link to video results.
Open Datasets No The paper describes a custom collected dataset: 'We collected demonstrations from 468 different task configurations, for a total of 1404 demonstration trajectories.' However, no access information (link, DOI, citation) is provided to make this dataset publicly available.
Dataset Splits Yes using 90% of the data set for training and 10% for validation.
Hardware Specification No The paper mentions the 'IRB-120 robot' used for the task, but does not specify the computing hardware (e.g., GPU/CPU models, memory, cloud instances) used for training or running the models.
Software Dependencies No The paper mentions using 'Adam (Kingma & Ba, 2014)' for optimization, but does not provide specific version numbers for Adam or any other software libraries or frameworks used in the implementation.
Experiment Setup Yes In all our experiments we used the following parameters for the SNN training... (1) Loss function: ...an L1 regression loss... (2) Monte Carlo samples: we chose N = 5... (3) Intention variable dimension: we chose z R5 for all experiments... (4) P(z): a 5-dimensional vector of independent uniform multinomials in {0 : 4}.