High Order Regularization for Semi-Supervised Learning of Structured Output Problems

Authors: Yujia Li, Rich Zemel

ICML 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments also demonstrate that this framework can utilize unlabeled data from a different source than the labeled data to significantly improve performance while saving labeling effort. We explore the efficacy of the proposed framework on two semi-supervised foreground-background segmentation tasks. Figures 3 and 4 show test accuracy and improvement over initial model, which are empirical results. The paper also includes sections on 'Datasets and Model Details' and 'Experimental Settings'.
Researcher Affiliation Academia Yujia Li1 YUJIALI@CS.TORONTO.EDU Richard Zemel1,2 ZEMEL@CS.TORONTO.EDU 1Department of Computer Science, University of Toronto, Toronto, ON, Canada 2Canadian Institute for Advanced Research, Toronto, ON, Canada
Pseudocode No The paper describes an iterative algorithm with 'Step 1' and 'Step 2' but does not present this as a formal pseudocode block or algorithm box.
Open Source Code No The paper does not contain any statement or link indicating that the source code for the described methodology is publicly available. The phrase 'these generated segmentations are available online' refers to a dataset generated by the authors, not the source code for their method.
Open Datasets Yes For the first task, we use the Weizmann Horse dataset (Borenstein & Ullman, 2002), a fully-labeled set of 328 images. For the unlabeled Horse dataset, we used images labeled horse in CIFAR-10 (Krizhevsky & Hinton, 2009), which are not segmented. For the second task, we constructed a labeled set of 214 bird images from the PASCAL VOC 2011 segmentation data (Everingham et al., 2010). The unlabeled Bird images come from the Caltech UCSD Bird (CUB) dataset (Welinder et al., 2010).
Dataset Splits Yes Note that while we vary the training and unlabeled sets in this way, the validation and test sets remain constant, in order to make comparisons fair. (Section 5.2) For the Horse task, there were up to 200 training images, 48 validation, and 80 test images... (Section 5.2) For the Bird task, there were up to 200 training images, 200 validation, and 200 test images... (Section 5.2) Table 1, under 'Experiment train validation test unlabeled' columns, explicitly lists sizes like 'W-48' and 'C-200' for validation.
Hardware Specification No The paper describes the computational models and experiments but does not provide specific details about the hardware used, such as GPU/CPU models, memory, or cloud computing instances.
Software Dependencies No The paper mentions the use of neural networks and graph cuts, but it does not specify any software libraries, frameworks, or their version numbers that would be required to reproduce the experiments.
Experiment Setup Yes Other hyper parameters, e.g. λ, µ, γ, are tuned using the validation set, see supplementary material for more details on parameter settings. (Section 5.1) During learning, the gradients are back-propagated through the neural network to update parameters. Since neural networks are highly nonlinear models, it is hard to find the optimal w in Eq. 7 in every Step 2 of our algorithm. Instead, we only take a few gradient steps in Step 2 of each iteration. (Section 5.1)