Joint Data-Task Generation for Auxiliary Learning

Authors: Hong Chen, Xin Wang, Yuwei Zhou, Yijian Qin, Chaoyu Guan, Wenwu Zhu

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments show that our proposed DTG-Aux L framework consistently outperforms existing methods in various auxiliary learning scenarios, particularly when the manually collected auxiliary data and tasks are unhelpful.
Researcher Affiliation Academia Hong Chen1, Xin Wang1,2 , Yuwei Zhou1, Yijian Qin1, Chaoyu Guan1, Wenwu Zhu1,2 1Department of Computer Science and Technology, Tsinghua University 2Beijing National Research Center for Information Science and Technology, Tsinghua
Pseudocode Yes We summarize the complete algorithm in Appendix 2
Open Source Code No The paper does not contain any explicit statements about releasing source code or links to a code repository.
Open Datasets Yes We conduct our experiments on two scenarios... (i) CUB [25]: ... (ii) CIFAR100 [26]: ... We choose the widely used Amazon Toys and Movies [27] datasets... (i) CIFAR10-100... (ii) Pet-CUB... on the Pet [28] dataset
Dataset Splits No There is also a validation dataset Dv which is used to evaluate the model performance on the primary task. The paper mentions the use of a validation set but does not provide specific split percentages or sample counts for reproduction.
Hardware Specification No The paper does not specify any particular hardware (e.g., CPU, GPU models, memory) used for running the experiments, only model architectures and datasets.
Software Dependencies No The paper mentions software components like 'Res Net18' and 'Auto INT' and 'MLP' but does not specify version numbers for any programming languages, libraries, or frameworks used for implementation.
Experiment Setup Yes In the generator, the embedding dimension mn is searched from {32, 64}, and the layer number of the MLP is searched from {2, 3, 4}. For the head of each task, we adopt Multi-Layer Perceptron(MLP) whose layer is searched from {1, 2}. N is fixed to 3 in all our experiments.