JSAT: Java Statistical Analysis Tool, a Library for Machine Learning

Authors: Edward Raff

JMLR 2017 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental A small benchmark on MNIST is given in Table 1, where JSAT is compared against the same algorithm in other libraries, with JSAT adjusted to match the default parameters used by others. Error rate was measured from one run on the standard training and testing split of MNIST.
Researcher Affiliation Academia Edward Raff EMAIL Department of Computer Science and Electrical Engineering University of Maryland, Baltimore County
Pseudocode No The paper describes the functionalities of the JSAT library and provides a code example in Listing 1, but it does not contain explicit pseudocode or algorithm blocks.
Open Source Code Yes JSAT is made available under the GNU GPL license here: https://github.com/Edward Raff/JSAT.
Open Datasets Yes A small benchmark on MNIST is given in Table 1, where JSAT is compared against the same algorithm in other libraries, with JSAT adjusted to match the default parameters used by others. Error rate was measured from one run on the standard training and testing split of MNIST.
Dataset Splits Yes Error rate was measured from one run on the standard training and testing split of MNIST.
Hardware Specification No The paper does not provide specific hardware details (like CPU/GPU models or memory) used for running the experiments. It only mentions 'single-threaded executions'.
Software Dependencies Yes It is written in Java 6 and has no dependencies, making it easy to integrate into any Java project without conflict.
Experiment Setup No JSAT adjusted to match the default parameters used by others. The paper describes the framework for parameter search within JSAT but does not provide specific hyperparameter values or detailed training configurations used for the benchmark results in Table 1.