Skip to content

Instantly share code, notes, and snippets.

View znxkznxk1030's full-sized avatar
๐ŸŒฎ
Taco

Youngsoo Kim znxkznxk1030

๐ŸŒฎ
Taco
View GitHub Profile
import os
import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import DataLoader
import torchvision
import torchvision.transforms as transforms
from torchvision.utils import make_grid, save_image
import matplotlib.pyplot as plt
import torch.nn.functional as F

Sliding Window

๊ฐœ์š”

์‚ฌ์šฉ์ฒ˜

Template

n = len(nums)

Binary Search

๊ฐœ์š”

  • ์ •๋ ฌ๋œ ๋ฐฐ์—ด์—์„œ ํŠน์ • ๊ฐ’์„ ์ฐพ๋Š” ์•Œ๊ณ ๋ฆฌ์ฆ˜.
  • ์ค‘๊ฐ„ ๊ฐ’์„ ๊ธฐ์ค€์œผ๋กœ ํƒ์ƒ‰ ๋ฒ”์œ„๋ฅผ ์ ˆ๋ฐ˜์”ฉ ์ค„์—ฌ๊ฐ€๋ฉฐ ํšจ์œจ์ ์œผ๋กœ ํƒ์ƒ‰. ( ์‹œ๊ฐ„ ๋ณต์žก๋„: $O(logN)$ )

์‚ฌ์šฉ์ฒ˜

  1. upper bound/lower bound ์ฐพ๊ธฐ
#1.
์•ˆ๋…•ํ•˜์„ธ์š”. 10์กฐ ๋ฐœํ‘œ๋ฅผ ์‹œ์ž‘ํ•˜๊ฒ ์Šต๋‹ˆ๋‹ค.
์ €ํฌ๋Š” "์Œ์„ฑ-์ด๋ฏธ์ง€ ๋ฉ€ํ‹ฐ๋ชจ๋‹ฌ์„ ํ™œ์šฉํ•˜์—ฌ ์Œ์„ฑ๊ณผ ๋งค์นญ๋˜๋Š” ์ธ๋ฌผ์„ ์ถ”๋ก ํ•˜๋Š” AI ๋ชจ๋ธ"์ด๋ผ๋Š” ์ฃผ์ œ๋กœ ์„ธ๋ฏธ๋‚˜ ์—ฐ๊ตฌ๋ฅผ ์ง„ํ–‰ํ•˜๊ณ ์ž ํ•ฉ๋‹ˆ๋‹ค.
#2.
๋จผ์ € ํ”„๋กœ์ ํŠธ ๊ฐœ์š”๋ฅผ ์„ค๋ช…๋“œ๋ฆฐ ํ›„, ๋‚ด์šฉ๊ณผ ์‘์šฉ, ์ผ์ • ์ˆœ์œผ๋กœ ๋ฐœํ‘œ๋ฅผ ์ง„ํ–‰ํ•˜๊ฒ ์Šต๋‹ˆ๋‹ค.
#3.

๊ธฐ๊ณ„ํ•™์Šต๊ณผ์ •๋ณด์ด๋ก  - ๊ณผ์ œ2 ๋…ผ๋ฌธ ์ดˆ์•ˆ ๊ฐœ์„ ํ•˜๊ธฐ

2025451021
์ธ๊ณต์ง€๋Šฅํ•™๊ณผ
๊น€์˜์ˆ˜

Title

On the Effect of Negative-Pair Variance in Contrastive Learning and a VRN-Based Solution

์ค‘๊ฐ„๊ณ ์‚ฌ ์˜ˆ์ƒ ๋ฌธ์ œ

1. (Finite) Markov Decision Process

1. ๊ฐ•ํ™”ํ•™์Šต(Reinforcement Learning)์˜ ์ •์˜๋ฅผ ์„œ์ˆ ํ•˜๊ณ , ์ง€๋„ํ•™์Šต(Supervised Learning)๊ณผ์˜ ์ฐจ์ด์ ์„ ์˜ˆ์‹œ์™€ ํ•จ๊ป˜ ์„ค๋ช…ํ•˜์‹œ์˜ค

A goal-directed learning from interaction

์ค‘๊ฐ„๊ณ ์‚ฌ ์˜ˆ์ƒ ๋ฌธ์ œ

Introduction

1. ๋‹ค์Œ ๊ฐœ๋…๋“ค: ์ธ๊ณต์ง€๋Šฅ(AI), ๋จธ์‹ ๋Ÿฌ๋‹(ML), ๋”ฅ๋Ÿฌ๋‹ (DL)์˜ ๊ด€๊ณ„๋ฅผ ์„ค๋ช…ํ•˜๊ณ , ๊ฐ๊ฐ ๋Œ€ํ‘œ์ ์ธ ์˜ˆ์‹œ๋ฅผ ํ•˜๋‚˜์”ฉ ๋“ค์–ด ์„œ์ˆ ํ•˜์‹œ์˜ค

$$ ๋”ฅ๋Ÿฌ๋‹ \subset ๋จธ์‹ ๋Ÿฌ๋‹ \subset ์ธ๊ณต์ง€๋Šฅ$$3

์ธ๊ณต์ง€๋Šฅ์€ ์ธ๊ฐ„์ฒ˜๋Ÿผ ์‚ฌ๊ณ ํ•˜๊ณ  ํ–‰๋™ํ•˜๋Š” ๊ธฐ๊ณ„๋ฅผ ๋งŒ๋“œ๋Š” ๊ธฐ์ˆ  ์ „๋ฐ˜์„ ์˜๋ฏธํ•œ๋‹ค.

๊ฐ•์˜ ๋‚ด์šฉ ์š”์•ฝ ๊ณผ์ œ

2025451021
์ธ๊ณต์ง€๋Šฅํ•™๊ณผ
๊น€์˜์ˆ˜

1. Entropy ์ •์˜

Entropy๋ž€ ์–ด๋–ค ํ™•๋ฅ  ๋ณ€์ˆ˜์— ๋Œ€ํ•ด ์ •๋ณด์˜ ์–‘์„ ์ธก์ •ํ•˜๋Š” ๊ฐœ๋…์ด๋‹ค. ์—ฌ๊ธฐ์—์„œ ์ •๋ณด๋Š” ๋ถˆํ™•์‹ค์„ฑ์„ ์˜๋ฏธํ•˜๊ณ  ํ•ด๋‹น ํ™•๋ฅ  ๋ณ€์ˆ˜์˜ ๋ถˆํ™•์‹ค์„ฑ์˜ ์ •๋„๋ฅผ ์˜๋ฏธํ•œ๋‹ค.

import random
import numpy as np
from visualize_train import draw_value_image, draw_policy_image
# left, right, up, down
ACTIONS = [np.array([0, -1]),
np.array([0, 1]),
np.array([-1, 0]),