Combining Logical Abduction and Statistical Induction: Discovering Written Primitives with Human Knowledge

Authors: Wang-Zhou Dai, Zhi-Hua Zhou

AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section we report two experimental results of LASIN on 3 real handwritten characters datasets.
Researcher Affiliation Academia Wang-Zhou Dai and Zhi-Hua Zhou National Key Laboratory for Novel Software Technology, Nanjing University, Nanjing 210023, China Collaborative Innovation Center of Novel Software Technology and Industrialization, Nanjing 210023, China {daiwz, zhouzh}@lamda.nju.edu.cn
Pseudocode Yes Algorithm 1: LASIN
Open Source Code No The paper does not provide explicit statements about releasing source code for the described methodology or a link to a code repository.
Open Datasets Yes MNIST (Le Cun et al. 2001): This dataset consists of 28 28 binary images with 60,000 training and 10,000 test instances. and Omniglot (Lake, Salakhutdinov, and Tenenbaum 2015): Omniglot dataset consists of 105 105 binary images across 1628 classes with just 20 images per class.
Dataset Splits Yes MNIST (Le Cun et al. 2001): This dataset consists of 28 28 binary images with 60,000 training and 10,000 test instances. and The performance are evaluated with 5-fold cross-validation.
Hardware Specification No The paper does not provide specific details about the hardware used for running its experiments.
Software Dependencies No The paper mentions software like 'mlpack toolbox (Curtin et al. 2013)' and 'SWIProlog (Wielemaker et al. 2012)' but does not provide specific version numbers for these software dependencies.
Experiment Setup Yes The dictionary sizes are set at |D| = 20, 50, 100, 200, respectively. These sizes are not very large because we believe the effective dimension of handwritten characters should be small, involving some different strokes, their combinations and spacial relations. The hyper-parameters (turn limit and error threshold) of Algorithm 1 in the experiments are determined by cross-validation on training data.