Collective Model Fusion for Multiple Black-Box Experts
Authors: Minh Hoang, Nghia Hoang, Bryan Kian Hsiang Low, Carleton Kingsford
ICML 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | This section evaluates and reports the empirical performance of our collective fusion frameworks CIGAR (light-weight, ephemeral inference fusion) and COLBI (persistent surrogate model fusion) on three real-world datasets |
| Researcher Affiliation | Collaboration | 1Carnegie Mellon University 2MIT-IBM Watson AI Lab, IBM Research Cambridge 3National University of Singapore. |
| Pseudocode | No | The paper describes the proposed algorithms (CIGAR, COLBI) in narrative text but does not provide formal pseudocode blocks or algorithm boxes. |
| Open Source Code | No | The paper does not provide any explicit statement about releasing source code or a link to a code repository for the described methodology. |
| Open Datasets | Yes | The DIABETES dataset (Efron et al., 2004) containing 442 diabetes patient records with 10 features. The target output is a quantitative measure of disease progression one year after baseline. (b) The AIMPEAK dataset (Chen et al., 2013a) containing 41850 instances of traffic measured along 775 road segments of an urban road network, each comprises of 5 variables that describe the corresponding segment. The target output is the averaged vehicle speed on the segment (km/h). (c) The PROTEIN dataset (Rana, 2013) featuring physicochemical properties of protein tertiary structure with 45730 instances, each containing 9 variables that describe various properties of the structure. |
| Dataset Splits | No | The paper specifies the number of data points used for training and testing for each dataset (e.g., '30 (DIABETES), 500 (AIMPEAK) and 500 (PROTEIN) data points' for training and '35 (DIABETES), 500 (AIMPEAK) and 500 (PROTEIN) data points' for testing), but does not explicitly mention a separate validation split or explicit percentages for data partitioning. |
| Hardware Specification | No | The paper does not provide specific details about the hardware (e.g., CPU, GPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper does not specify any software dependencies, libraries, or their version numbers used in the experiments. |
| Experiment Setup | No | The paper mentions general training concepts like 'sufficiently small learning rate' and 'number of fusion iterations' but does not provide specific hyperparameter values (e.g., learning rate value, batch size, specific optimizer settings, or explicit number of epochs) for the experimental setup. |