Hi, in the DeepLense common task, we're given two sets: train and val. Can we do hyperparameter tuning on val (assuming our solution will be tested on an external set not given to us) or should we consider the existing val set as the test set and do hyperparameter tuning on a subset of the train set?
For now I'm getting an AUC of 99% on the train set and 98.5% on the val set (without any hyperparameter tuning). Is this enough for the common task?
Also, how will the submissions be judged? Based on AUC? Code clarity? Code structure?
Hi everyone. Absolutely newcomer to ML4SCI community
I am currently a final year undergraduate student at Veermata Jijabai Technological Institute, India. I have experience in real-time object detection, classification, and tracking algorithms. I am also doing a research internship on a topic related to leaf classification and disease detection using fewer computational resources.
I came across the project list for this year’s Google Summer of Code and found the project titled “Vision Transformers for End-to-End Particle Reconstruction for the CMS Experiment” quite interesting to me as my final year project is on similar lines where we are aiming to detect deepfakes, by using an ensemble of low latency algorithms followed by temporal feature extraction by using optical flow and vision transformers
I am currently working on the tasks that are given.
@Shra1-25 I was trying to solve the given tasks for the project - 'graph neural network for end to end particle identification with CMS experiment'
while solving the 1st task I am facing problem while training the dataset, I have given the command for more than 12 hours. Is it any other way to do the same?
For the 2nd task, I am unable to read the dataset
This is showing - "ArrowMemoryError: malloc of size 13602000000 failed"
I would be glad if anyone can help to resolve my doubts so that I can proceed in solving these doubts.
As I am interested in the Quantum CNN project, I found the previous year's work on this project by Eraraya Ricardo Muten. I wanted to how this year's project is different from the previous year as he also implemented a QCNN using TensorFlow quantum.
Mentors please resolve this confusion.