|
287 | 287 | "source": [ |
288 | 288 | "### Performance compared among different implementation\n", |
289 | 289 | "\n", |
| 290 | + "We conduct performance comparison experiments across different implementations of IRT.\n", |
| 291 | + "The implementations we chose are those high-stared in github\n", |
| 292 | + "and easy to be applied in new datasets with clear tutorials.\n", |
290 | 293 | "[pyirt](https://github.com/17zuoye/pyirt),\n", |
291 | | - "[bayesian-irt](https://github.com/jamalex/bayesian-irt),\n", |
292 | | - "[irt_parameter_estimation](https://github.com/pluralsight/irt_parameter_estimation),\n", |
293 | | - "[pypsy](https://github.com/inuyasha2012/pypsy),\n", |
294 | 294 | "[vibo](https://github.com/mhw32/variational-item-response-theory-public)\n", |
295 | | - "and [irt](https://github.com/aimir/irt)\n", |
296 | 295 | "are high-stared implementation of IRT in github.\n", |
297 | 296 | "We run them on the cdbd dataset and report the result (AUC and MAP is reported) in the following tables,\n", |
298 | 297 | "which also includes our solution:\n", |
299 | 298 | "\n", |
300 | 299 | "#### AUC (auroc)\n", |
301 | 300 | "\n", |
302 | | - "| | cdbd-a0910 |\n", |
303 | | - "| ------------------------ | ---------- |\n", |
304 | | - "| pyirt | 0.501737 |\n", |
305 | | - "| bayesian-irt | |\n", |
306 | | - "| irt_parameter_estimation | |\n", |
307 | | - "| irt | |\n", |
308 | | - "| XCD-IRT-MXIRT | 0.703092 |\n", |
309 | | - "| XCD-IRT-MXIRT-with-tricks | 0.739607 |\n", |
| 301 | + "| | cdbd-a0910 | cdbd-lsat |\n", |
| 302 | + "| ------------------------ | ---------- | ---------- |\n", |
| 303 | + "| pyirt | 0.501737 | |\n", |
| 304 | + "| bayesian-irt | | |\n", |
| 305 | + "| irt_parameter_estimation | | |\n", |
| 306 | + "| irt | | |\n", |
| 307 | + "| XCD-IRT-MXIRT | 0.703092 | |\n", |
| 308 | + "| XCD-IRT-MXIRT-with-tricks | 0.739607 | 0.673990 |\n", |
310 | 309 | "\n", |
311 | 310 | "#### MAP(aupoc)\n", |
312 | | - "| | cdbd-a0910 |\n", |
313 | | - "| ------------------------ | ---------- |\n", |
314 | | - "| pyirt | 0.674690 |\n", |
315 | | - "| bayesian-irt | |\n", |
316 | | - "| irt_parameter_estimation | |\n", |
317 | | - "| irt | |\n", |
318 | | - "| XCD-IRT-MXIRT | 0.788603 |\n", |
319 | | - "| XCD-IRT-MXIRT-with-tricks | 0.826107 |\n", |
| 311 | + "| | cdbd-a0910 | cdbd-lsat |\n", |
| 312 | + "| ------------------------ | ---------- | ---------- |\n", |
| 313 | + "| pyirt | 0.674690 | |\n", |
| 314 | + "| bayesian-irt | | |\n", |
| 315 | + "| irt_parameter_estimation | | |\n", |
| 316 | + "| irt | | |\n", |
| 317 | + "| XCD-IRT-MXIRT | 0.788603 | |\n", |
| 318 | + "| XCD-IRT-MXIRT-with-tricks | 0.826107 | 0.868745 |\n", |
320 | 319 | "\n", |
321 | 320 | "The scripts we used to test the performance of the baseline implementation could be found in:\n", |
322 | 321 | "* [pyirt](../scripts/reproduction/pyirt)\n", |
|
0 commit comments