In order to study different subatomic particles, particle physicists need to use particle accelerators, like the Large Hadron Collider. Opposing beams of protons are accelerated to almost the speed of light and are smashed against each other, resulting in a cone-shaped spray of subatomic particles called jets.
There are several kinds of jets - in particular we observe single charmed quarks (QCD_c), single bottom quarks (QCD_b), the Higgs Boson which decays into two bottom quarks (H_bb), two charmed quarks (QCD_cc), two bottom quarks (QCD_bb), and all other jets (QCD_other). These categories depend on the decay and spray of the particles, as seen on the image to the right.
Previously, we trained a model to classify the Higgs boson against all other jets. Now, we want to expand our classifier to all jets.
Pipeline for jet multiclassification process
Before adding weights, our models struggled with poor performance. Even so, we can tell that the Higgs boson jet has the best classification amongst all the jets.
After adding more datasets and incorporating class weights into the model, we see a huge jump in improvement for our models. Once again, we notice that the Higgs boson jet has a far more impressive performance from all three models compared to the other jets.
created with
Static Website Generators .