# QSVM [Hackmd Fast Tutorial](https://hackmd.io/c/tutorials-tw/%2Fs%2Fquick-start-tw) ## Paper * [Enhancing Quantum Support Vector Machines through Variational Kernel Training](https://arxiv.org/pdf/2305.06063.pdf) ## Dataset ## Method ## Result ### Classical #### PCA + Linear * MSE | | Training | Testing | | -------- | -------- | -------- | | C | 0.00024 | 0.00020 * Fitting Figure | C-Training | C-Testing | | -------- | -------- | | ![](https://hackmd.io/_uploads/ryKwoaV82.png)| ![](https://hackmd.io/_uploads/Bk-uj6VU2.png) | ### IBM # Quera ## PCA + Simple Encoding + Tuning * Dataset * output06_24.csv * MSE | | Training | Testing | | -------- | -------- | -------- | | Q | 4e-8 | 0.00017 | | C | 0.00024|0.00020 * Fitting Figure | Q-Training | Q-Test | | --------------------------------------------- | --------------------------------------------- | | ![](https://hackmd.io/_uploads/SJ0bcpVLh.png) | ![](https://hackmd.io/_uploads/H1pzqTEL2.png) | ## PCA + Chain + Tuning ### Case1 * Dataset * output08_21_BSBM.csv * train : 180 * test : 60 * Fitting Figure | Q-Training | Q-Test | | --------------------------------------------- | --------------------------------------------- | | ![](https://hackmd.io/_uploads/HJzpeFCy6.png) | ![](https://hackmd.io/_uploads/H1AplKCka.png) | ### Case2 * Dataset * output08_21_BSBM.csv * train : 500 * test : 150 * Fitting Figure - epsilon = 0 | Q-Training | Q-Test | | --------------------------------------------- | --------------------------------------------- | | ![](https://hackmd.io/_uploads/B1JLa9Cyp.png) |![](https://hackmd.io/_uploads/SJUVTqC16.png)| * Fitting Figure - tuned | Q-Training | Q-Test | | --------------------------------------------- | --------------------------------------------- | | ![](https://hackmd.io/_uploads/HytsxjCka.png) |![](https://hackmd.io/_uploads/Hkl0ljAkT.png)| * Fitting Figure - tuned | C-Training | C-Test | | --------------------------------------------- | --------------------------------------------- | | ![](https://hackmd.io/_uploads/H1o-eoRkT.png) |![](https://hackmd.io/_uploads/rk5fesRJ6.png)| * Q-Kernel Matrix ![](https://hackmd.io/_uploads/SJJ-1iCJa.png) * C-Kernel Matrix ![](https://hackmd.io/_uploads/rJdgyoA16.png) ## Chain + Tuning (1009) ### Case1 - 700 data * Dataset * output08_21_BSBM.csv * train : 700 * test : 210 * epsilon : 0.022 * C : 0.1 * $V_{ij}t=\pi$ * Fitting Figure - epsilon = 0 | Q-Training | Q-Test | | --------------------------------------------- | --------------------------------------------- | | ![](https://hackmd.io/_uploads/HJV7CFMZT.png)|![](https://hackmd.io/_uploads/SJAMAYMba.png) | * Fitting Figure - tuned | Q-Training | Q-Test | | --------------------------------------------- | --------------------------------------------- | | ![](https://hackmd.io/_uploads/BkrqpKM-a.png)|![](https://hackmd.io/_uploads/ByIjTYzWT.png)| * Fitting Figure - tuned | C-Training | C-Test | | --------------------------------------------- | --------------------------------------------- | | ![](https://hackmd.io/_uploads/Hyopatz-T.png)|![](https://hackmd.io/_uploads/SklTptGW6.png) | * Q-Kernel Matrix ![](https://hackmd.io/_uploads/B14wRYMbp.png) * C-Kernel Matrix ![](https://hackmd.io/_uploads/SJYI0Yfb6.png) ### Case2 - 500 data * Dataset * output08_21_BSBM.csv * train : 700 * test : 210 * epsilon : 0.019 * C : 0.1 * $V_{ij}t=\pi$ * Fitting Figure - tuned | Q-Training | Q-Test | | --------------------------------------------- | --------------------------------------------- | | ![](https://hackmd.io/_uploads/H1cm1cfW6.png)|![](https://hackmd.io/_uploads/BkI719GWa.png) | * Fitting Figure - tuned | C-Training | C-Test | | --------------------------------------------- | --------------------------------------------- | |![](https://hackmd.io/_uploads/Sy1NJczb6.png) |![](https://hackmd.io/_uploads/ByQ4JcfWp.png) | * Q-Kernel Matrix ![](https://hackmd.io/_uploads/Byvzk9GZT.png) * C-Kernel Matrix ![](https://hackmd.io/_uploads/ry7fJ9fWT.png) # Case3 - 100 data * Dataset * output08_21_BSBM.csv * train : 100 * test : 30 * **epsilon : 0.008** * C : 0.1 * $V_{ij}t=\pi$ * MSE | | Training | Testing | | -------- | -------- | -------- | | Q | 5.5e-5 | 0.00014 | | C |1.2e-5 |0.00011 | * Fitting Figure - epsilon = 0 | Q-Training | Q-Test | | --------------------------------------------- | --------------------------------------------- | | | | * Fitting Figure - tuned | Q-Training | Q-Test | | --------------------------------------------- | --------------------------------------------- | | ![](https://hackmd.io/_uploads/rkH_7jmZp.png)|![](https://hackmd.io/_uploads/rJuFQjmbT.png) | * Fitting Figure - tuned | C-Training | C-Test | | --------------------------------------------- | --------------------------------------------- | | ![](https://hackmd.io/_uploads/rkHiQo7-T.png)|![](https://hackmd.io/_uploads/Bk6iXom-p.png) | * Q-Kernel Matrix ![](https://hackmd.io/_uploads/H1ORmi7-a.png) * C-Kernel Matrix ![](https://hackmd.io/_uploads/r1-AXjXbp.png) # Sample * Dataset * output08_21_BSBM.csv * train : 700 * test : 210 * epsilon : 0.019 * C : 0.1 * $V_{ij}t=\pi$ * MSE | | Training | Testing | | -------- | -------- | -------- | | Q | | | | C | | * Fitting Figure - epsilon = 0 | Q-Training | Q-Test | | --------------------------------------------- | --------------------------------------------- | | | | * Fitting Figure - tuned | Q-Training | Q-Test | | --------------------------------------------- | --------------------------------------------- | | || * Fitting Figure - tuned | C-Training | C-Test | | --------------------------------------------- | --------------------------------------------- | | || * Q-Kernel Matrix * C-Kernel Matrix