Opportunistic Learning: Budgeted Cost-Sensitive Learning from Data Streams
Authors: Mohammad Kachuee, Orpaz Goldstein, Kimmo Kärkkäinen, Sajad Darabi, Majid Sarrafzadeh
ICLR 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The solution is evaluated on three different datasets including the well-known MNIST dataset as a benchmark as well as two cost-sensitive datasets: Yahoo Learning to Rank and a dataset in the medical domain for diabetes classification. According to the results, the proposed method is able to efficiently acquire features and make accurate predictions. |
| Researcher Affiliation | Academia | Mohammad Kachuee, Orpaz Goldstein, Kimmo Karkkainen, Sajad Darabi, Majid Sarrafzadeh Department of Computer Science University of California, Los Angeles (UCLA) Los Angeles, CA 90095, USA {mkachuee,orpgol,kimmo,sajad.darabi,majid}@cs.ucla.edu |
| Pseudocode | Yes | Algorithm 1 summarizes the procedures for cost-sensitive feature acquisition and training the networks. |
| Open Source Code | Yes | A version of the source code and the health dataset preproccessing code for this paper is available at: https://github.com/mkachuee/Opportunistic |
| Open Datasets | Yes | The MNIST dataset is used as it is a widely used benchmark. ... Regarding the LTRC dataset, we use feature acquisition costs provided by Yahoo! ... The national health and nutrition examination survey (NAHNES) data (nha, 2018) was used for this purpose. ... URL https://www.cdc.gov/nchs/nhanes. |
| Dataset Splits | Yes | Each dataset was randomly splitted to 15% for test, 15% for validation, and the rest for train. |
| Hardware Specification | No | The experiments took between a few hours to a couple days on a GPU server, depending on the experiment. The paper mentions 'GPU server' but does not specify any particular model of GPU, CPU, or other hardware components used for the experiments. |
| Software Dependencies | No | In this paper, Py Torch numerical computational library (Paszke et al., 2017) is used for the implementation of the proposed method. ... The Adam optimization algorithm Kingma & Ba (2014) was used throughout this work for training the networks. The paper mentions PyTorch and Adam but does not provide specific version numbers for these or any other software libraries or dependencies. |
| Experiment Setup | Yes | Cross-entropy and mean squared error (MSE) loss functions were used as the objective functions for the P and Q networks, respectively. ... We used dropout with the probability of 0.5 for all hidden layers of the P-Network and no dropout for the Q-Network. The target Q-Network was updated softly with the rate of 0.001. We update P, Q, and target Q networks every 1 + nfe / 100 experiences, where nfe is the total number of features in an experiment. In addition, the replay memory size is set to store 1000 nfe most recent experiences. The random exploration probability is decayed such that eventually it reaches the probability of 0.1. |