Mr.Siolas requested our apologies for cheating, in the overview section. ## Labels were public First off, the labels were not hidden. Instead, Mr.Siolas encouraged all participants to take a look at the actual dataset(which contained the labels) by showing us [this](https://fluent.ai/fluent-speech-commands-a-dataset-for-spoken-language-understanding-research/) link numerous times. ## Our mistake We downloaded the labels and used them to **evaluate** our model. Instead of submitting our actual results, we submitted the labels, by mistake. **We had no intention whatsoever to either cheat or disrespect the organizers, participants, Kaggle and the teacher** . We immediately contacted the organizers and requested our submission to be deleted. They were very cooperative and understanding but Mr.Siolas refused to even talk to us numerous times, while the organisers tried their best to persuade him. We really appreciate their effort as well as them listening carefully our explanations. ## The winning submission In the meantime, we still managed to achieve an accuracy of 0.95 which would put us on the top of the leaderboard, using the [pre-trained model we shared in the discussion section](https://www.kaggle.com/c/2020-athens-eestech-challenge/discussion/155141) as well as auxiliary .wav data. However, Mr.Siolas repeatedly avoided speaking to us, never gave us the chance to explain the situation and resolve the misunderstanding, even though we were willing to describe in detail our approach as well as every single line of code we wrote. On top of that, he reported the incident as cheating in order for us to get banned from Kaggle. We also e-mailed Mr.Siolas but he did not respond. Of course, we did not select our 1.0 submission for the leaderboard, but instead the 0.95 one which was fairly generated. ## The inappropriate response We decided not to express our complaints in the final meeting, as requested by the organizers for the contest to end smoothly, out of respect to them and the other teams. On the contrary, Mr.Siolas(who did not even attend the meeting, not to face us and our complaints) acted differently. He decided to humiliate us in the Overview Section, accusing us of cheating, explicitly pointing out our names in public, without even letting us talk to him and explain the situation. We consider his behaviour to be unprofessional and insulting. ## Our code We will soon provide a link to our code as well as the pre-trained model and data needed to reproduce our submission. We will try and automate this process as much as possible. Once it is ready, please follow the instruction in the README.md file to reproduce our submission in a Unix-like system. ## Last notes We entered this competition in order to learn and meet people as passionate as we are when it comes to data science. We were actually excited to share our progress with Mr.Siolas. He was supposed to be our mentor. After insulting us like that in public, repeatedly calling cheaters, we want nothing to do with him anymore. We are ML enthusiasts. **Students**, not thieves. On behalf of SKRT, Konstantinos Koyias. ## Links * labels & dataset: [link](https://fluent.ai/fluent-speech-commands-a-dataset-for-spoken-language-understanding-research/) * pre-trained model: [link](https://github.com/lorenlugosch/end-to-end-SLU) * our code: [soon to be uploaded]