Supervised Learning with General Risk Functionals
Authors: Liu Leqi, Audrey Huang, Zachary Lipton, Kamyar Azizzadenesheli
ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In experiments, we demonstrate the efficacy of our learning procedure, both in settings where uniform convergence results hold and in high-dimensional settings with deep networks. |
| Researcher Affiliation | Academia | 1Machine Learning Department, Carnegie Mellon University 2Department of Computer Science, University of Illinois Urbana Champaign 3Department of Computer Science, Purdue University. |
| Pseudocode | No | The paper does not include any pseudocode or clearly labeled algorithm blocks. The optimization procedure is described in text and mathematical equations. |
| Open Source Code | No | The paper does not provide any explicit statements or links indicating that source code for the described methodology is open-source or publicly available. |
| Open Datasets | Yes | We perform risk assessments on pretrained Pytorch models for Image Net classification. ... After showing that the classifier learned under different risk objectives behave differently in a toy example, we learn risk-sensitive models for CIFAR-10. |
| Dataset Splits | Yes | The models are VGG-11 (Simonyan & Zisserman, 2014), Goog Le Net (Szegedy et al., 2015), Shuffle Net (Ma et al., 2018), Inception (Szegedy et al., 2016) and Res Net-18 (He et al., 2016) and the accuracy of these models evaluated on the validation set are around 69% (Table 1). |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper mentions 'Pytorch models' but does not specify version numbers for any software dependencies used in the experiments. |
| Experiment Setup | Yes | We have trained VGG-16 models on CIFAR-10 through minimizing the empirical risks for expected loss, CVa R.05, CVa R.7 and HRM.3,.4 (Leqi et al., 2019) using the gradient descent step presented in (7). The models are trained over 150 epochs and the learning rate is chosen to be 0.005. |