CLAP: Collaborative Adaptation for Patchwork Learning
Authors: Sen Cui, Abudukelimu Wuerkaixi, Weishen Pan, Jian Liang, Lei Fang, Changshui Zhang, Fei Wang
ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In extensive experiments, we demonstrate the superiority of the proposed method compared to existing related methods on benchmark data sets and a real-world clinical data set. |
| Researcher Affiliation | Collaboration | 1 Institute for Artificial Intelligence, Tsinghua University (THUAI) Beijing National Research Center for Information Science and Technology (BNRist) Department of Automation, Tsinghua University, Beijing, P.R.China 2 Department of Population Health Sciences, Weill Cornell Medical College, Cornell University, New York, USA 3 Independent Researcher 4 Data Canvas Technology Co., Ltd. |
| Pseudocode | Yes | Algorithm 1 The training of CLAP |
| Open Source Code | Yes | The source codes of our framework are made publicly available at https://github.com/zaocan666/CLAP. |
| Open Datasets | Yes | Datasets. Following the work (Sutter et al., 2020b), we evaluate our method with baselines on benchmark datasets with various modalities, including Poly MNIST, MNIST-SVHN-TEXT, Celeb A and CUB. More importantly, the practicability of our method is validated on a real-world clinical distributed dataset e ICU (Pollard et al., 2018). |
| Dataset Splits | No | The paper explicitly mentions training and test sets but does not provide details on validation dataset splits. |
| Hardware Specification | Yes | In the experiments, we conduct all methods on a local Linux server that has two physical CPU chips (Intel(R) Xeon(R) CPU E5-2640 v4 @ 2.40GHz) and 32 logical kernels. All methods are implemented using Pytorch framework and all models are trained on Ge Force RTX 2080 Ti GPUs. |
| Software Dependencies | No | The paper mentions using the 'Pytorch framework' but does not specify a version number for it or other software dependencies. |
| Experiment Setup | Yes | We use Adam optimizer for training and the training batch size of all experiments is set as 256. The architecture of classifiers for the coherence test is the same as the unimodal encoder architecture except for the last layer. The model architecture and training hyperparameters are the same for different methods. Bimodal Celeb A. The learning rate is set as 0.0005, β is set as 1.0, dimension of latent vectors is set as 64. The models are trained for 200 epochs. |