9th place solution in "Santa 2020 - The Candy Cane Contest"

Overview

Santa 2020 - The Candy Cane Contest

My solution in this Kaggle competition "Santa 2020 - The Candy Cane Contest", 9th place.

Basic Strategy

In this competition, the reward was decided by comparing the threshold and random generated number. It was easy to calculate the probability of getting reward if we knew the thresholds. But the agents can't see the threshold during the game, we had to estimate it.

Like other teams, I also downloaded the history by Kaggle API and created a dataset for supervised learning. We can see the true value of threshold at each round in the response of API. So, I used it as the target variable.

In the middle of the competition, I found out that quantile regression is much better than conventional L2 regression. I think it can adjust the balance between Explore and Exploit by the percentile parameter.

Features

        #         Name Explanation
#1 round number of round in the game (0-1999)
#2 last_opponent_chosen whether the opponent agent chose this machine in the last step or not
#3 second_last_opponent_chosen whether the opponent agent chose this machine in the second last step or not
#4 third_last_opponent_chosen whether the opponent agent chose this machine in the third last step or not
#5 opponent_repeat_twice whether the opponent agent continued to choose this machine in the last two rounds (#2 x #3)
#6 opponent_repeat_three_times whether the opponent agent continued to choose this machine in the last three rounds (#2 x #3 x #4)
#7 num_chosen how many times the opponent and my agent chose this machine
#8 num_chosen_mine how many times my agent chose this machine
#9 num_chosen_opponent how many time the opponent agent chose this machine (#7 - #8)
#10 num_get_reward how many time my agent got rewards from this machine
#11 num_non_reward how many time my agent didn't get rewarded from this machine
#12 rate_mine ratio of my choices against the total number of choices (#8 / #7)
#13 rate_opponent ratio of opponent choices against the total number of choices (#9 / #7)
#14 rate_get_reward ratio of my rewarded choices against the total number of choices (#10 / #7)
#15 empirical_win_rate posterior expectation of threshold value based on my choices and rewords
#16 quantile_10 10% point of posterior distribution of threshold based on my choices and rewords
#17 quantile_20 20% point of posterior distribution of threshold based on my choices and rewords
#18 quantile_30 30% point of posterior distribution of threshold based on my choices and rewords
#19 quantile_40 40% point of posterior distribution of threshold based on my choices and rewords
#20 quantile_50 50% point of posterior distribution of threshold based on my choices and rewords
#21 quantile_60 60% point of posterior distribution of threshold based on my choices and rewords
#22 quantile_70 70% point of posterior distribution of threshold based on my choices and rewords
#23 quantile_80 80% point of posterior distribution of threshold based on my choices and rewords
#24 quantile_90 90% point of posterior distribution of threshold based on my choices and rewords
#25 repeat_head how many times my agent chose this machine before the opponent agent chose this agent for the first time
#26 repeat_tail how many times my agent chose this machine after the opponent agent chose this agent last time
#27 repeat_get_reward_head how many times my agent got reward from this machine before my agent didn't get rewarded or the opponent agent chose this agent for the first time
#28 repeat_get_reward_tail how many times my agent got reward from this machine after my agent didn't get rewarded or the opponent agent chose this agent last time
#29 repeat_non_reward_head how many times my agent didn't get rewarded from this machine before my agent got reward or the opponent agent chose this agent for the first time
#30 repeat_non_reward_tail how many times my agent didn't get rewarded from this machine after my agent got reward or the opponent agent chose this agent last time
#31 opponent_repeat_head how many times the opponent agent chose this machine before my agent chose this machine for the first time
#32 opponent_repeat_tail how many times the opponent agent chose this machine after my agent chose this machine last time

Software

  • Python 3.7.8
  • numpy==1.18.5
  • pandas==1.0.5
  • matplotlib==3.2.2
  • lightgbm==3.1.1
  • catboost==0.24.4
  • xgboost==1.2.1
  • tqdm==4.47.0

Usage

  1. download data from Kaggle by /src/01_downlaod/download.py

  2. create a dataset by /src/02_[regressor]/preprocess.py

  3. train a model by /src/02_[regressor]/train.py

Top Agents

Regressor Loss NumRound LearningRate LB Score SubmissionID
LightBGM Quantile (0.65) 4000 0.05 1449.4 19318812
LightBGM Quantile (0.65) 4000 0.10 1442.1 19182047
LightBGM Quantile (0.65) 3000 0.03 1438.8 19042049
LightBGM Quantile (0.66) 3500 0.04 1433.9 19137024
CatBoost Quantile (0.65) 4000 0.05 1417.6 19153745
CatBoost Quantile (0.67) 3000 0.10 1344.5 19170829
LightGBM MSE 4000 0.03 1313.3 19093039
XGBoost Pairwised 1500 0.10 1173.5 19269952
Owner
toshi_k
toshi_k
An Easy-to-use, Modular and Prolongable package of deep-learning based Named Entity Recognition Models.

DeepNER An Easy-to-use, Modular and Prolongable package of deep-learning based Named Entity Recognition Models. This repository contains complex Deep

Derrick 9 May 30, 2022
Scalable machine learning based time series forecasting

mlforecast Scalable machine learning based time series forecasting. Install PyPI pip install mlforecast Optional dependencies If you want more functio

Nixtla 145 Dec 24, 2022
AI创造营 :Metaverse启动机之重构现世,结合PaddlePaddle 和 Wechaty 创造自己的聊天机器人

paddle-wechaty-Zodiac AI创造营 :Metaverse启动机之重构现世,结合PaddlePaddle 和 Wechaty 创造自己的聊天机器人 12星座若穿越科幻剧,会拥有什么超能力呢?快来迎接你的专属超能力吧! 现在很多年轻人都喜欢看科幻剧,像是复仇者系列,里面有很多英雄、超

105 Dec 22, 2022
Deep Reinforcement Learning by using an on-policy adaptation of Maximum a Posteriori Policy Optimization (MPO)

V-MPO Simple code to demonstrate Deep Reinforcement Learning by using an on-policy adaptation of Maximum a Posteriori Policy Optimization (MPO) in Pyt

Nugroho Dewantoro 9 Jun 06, 2022
Image to Image translation, image generataton, few shot learning

Semi-supervised Learning for Few-shot Image-to-Image Translation [paper] Abstract: In the last few years, unpaired image-to-image translation has witn

yaxingwang 49 Nov 18, 2022
High-Fidelity Pluralistic Image Completion with Transformers (ICCV 2021)

Image Completion Transformer (ICT) Project Page | Paper (ArXiv) | Pre-trained Models | Supplemental Material This repository is the official pytorch i

Ziyu Wan 243 Jan 03, 2023
FCAF3D: Fully Convolutional Anchor-Free 3D Object Detection

FCAF3D: Fully Convolutional Anchor-Free 3D Object Detection This repository contains an implementation of FCAF3D, a 3D object detection method introdu

SamsungLabs 153 Dec 29, 2022
An Agnostic Computer Vision Framework - Pluggable to any Training Library: Fastai, Pytorch-Lightning with more to come

IceVision is the first agnostic computer vision framework to offer a curated collection with hundreds of high-quality pre-trained models from torchvision, MMLabs, and soon Pytorch Image Models. It or

airctic 789 Dec 29, 2022
Implementation of UNet on the Joey ML framework

Independent Research Project - Code Joey can be cloned from here https://github.com/devitocodes/joey/. Devito and other dependencies such as PyTorch a

Navjot Kukreja 1 Oct 21, 2021
This is a TensorFlow implementation for C2-Rec

This is a TensorFlow implementation for C2-Rec We refer to the repo SASRec. Requirements requirement.txt Datasets This repo includes Amazon Beauty dat

7 Nov 14, 2022
Bunch of different tools which helps visualizing and annotating images for semantic/instance segmentation tasks

Data Framework for Semantic/Instance Segmentation Bunch of different tools which helps visualizing, transforming and annotating images for semantic/in

Bruno Fernandes Carvalho 5 Dec 21, 2022
Official implementation of VQ-Diffusion

Vector Quantized Diffusion Model for Text-to-Image Synthesis Overview This is the official repo for the paper: [Vector Quantized Diffusion Model for T

Microsoft 592 Jan 03, 2023
Final project code: Implementing MAE with downscaled encoders and datasets, for ESE546 FA21 at University of Pennsylvania

546 Final Project: Masked Autoencoder Haoran Tang, Qirui Wu 1. Training To train the network, please run mae_pretraining.py. Please modify folder path

Haoran Tang 0 Apr 22, 2022
Simple converter for deploying Stable-Baselines3 model to TFLite and/or Coral

Running SB3 developed agents on TFLite or Coral Introduction I've been using Stable-Baselines3 to train agents against some custom Gyms, some of which

Gary Briggs 16 Oct 11, 2022
Scheduling BilinearRewards

Scheduling_BilinearRewards Requirement Python 3 =3.5 Structure main.py This file includes the main function. For getting the results in Figure 1, ple

junghun.kim 0 Nov 25, 2021
A-SDF: Learning Disentangled Signed Distance Functions for Articulated Shape Representation (ICCV 2021)

A-SDF: Learning Disentangled Signed Distance Functions for Articulated Shape Representation (ICCV 2021) This repository contains the official implemen

81 Dec 14, 2022
Multi-modal Text Recognition Networks: Interactive Enhancements between Visual and Semantic Features

Multi-modal Text Recognition Networks: Interactive Enhancements between Visual and Semantic Features | paper | Official PyTorch implementation for Mul

48 Dec 28, 2022
[CVPR 2022] Official Pytorch code for OW-DETR: Open-world Detection Transformer

OW-DETR: Open-world Detection Transformer (CVPR 2022) [Paper] Akshita Gupta*, Sanath Narayan*, K J Joseph, Salman Khan, Fahad Shahbaz Khan, Mubarak Sh

Akshita Gupta 127 Dec 27, 2022
RetinaFace: Deep Face Detection Library in TensorFlow for Python

RetinaFace is a deep learning based cutting-edge facial detector for Python coming with facial landmarks.

Sefik Ilkin Serengil 512 Dec 29, 2022
A library built upon PyTorch for building embeddings on discrete event sequences using self-supervision

pytorch-lifestream a library built upon PyTorch for building embeddings on discrete event sequences using self-supervision. It can process terabyte-si

Dmitri Babaev 103 Dec 17, 2022