Hierarchical Class-Based Curriculum Loss

Authors: Palash Goyal, Divya Choudhary, Shalini Ghosh

IJCAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We test our loss function on real world image data sets, and show that it significantly outperforms state-of-the-art baselines.
Researcher Affiliation Industry 1Samsung Research America 2Amazon Alexa AI {palash.goyal, d.choudhary}@samsung.com, ghoshsha@amazon.com
Pseudocode Yes Algorithm 1: Class Selection for Hierarchical Class Based Curriculum Learning.
Open Source Code No The paper does not provide any statement or link indicating the release of open-source code for the described methodology.
Open Datasets Yes We evaluate our loss function on four real world image data sets (i) IMCLEF [Dimitrovski et al., 2011], (ii) Wipo [Rousu et al., 2006], (iii) Reuters [Lewis et al., 2004], and (iv) i Naturalist [Van Horn et al., 2018].
Dataset Splits Yes We select the hyperparameters of the neural network using evaluation on a validation set with binary cross entropy loss.
Hardware Specification Yes We performed our experiments on 2 Nvidia Ge Force RTX 2080 Ti with 12 GB memory with 3.30 GHz CPU clock speed.
Software Dependencies No The paper mentions using a multi-layer perceptron and ResNet-18, and Adam optimizer, but does not provide specific version numbers for any software dependencies.
Experiment Setup Yes For evaluation on i Naturalist, we used a Res Net-18 architecture (pre-trained on Image Net). We use Adam optimizer and a learning rate of 10 5. ... Based on this, we get a structure with 800 hidden neurons and a dropout of 0.25.