![]()
Zero-Sum Games and AI A player's utility gain or loss is exactly balanced by the combined gain or loss of opponents → 어느 한 플레이어에게 이득인것은 반대편 플레이어에게는 손해여야한다.This is a powerful concept important to AI development for measuring the cost/benefit of a particular moveNash Equilibrium Games and AI Traditional strategy - Minimax:Attempt to minimize opponent's maximum reward at each state (Nash Equilibriu..
![]()
Memory mapped I/OAllows an application to map a file into memoryProgrammers can access the file directly through memory어떤 file들을 메모리 공간에 mapping시켜서 pointer들으로 접근할 수 있게 해준다. mmap void *mmap(void *addr, size_t len, int prot, int flags, int fd, off_t offset)Maps len bytes of the object represented by the file descriptor fd, starting at offset, into memoryaddrOffers a suggestion to the kernel of whe..
File descriptorRegular filesSpecial filesCharacter devices (e.g., Keyboard)Block devicesPipesSockets우리가 평소에 다루던 진짜 file들은 Regular file들이고, Character devices나 pipe와 같은 file들을 Special file들이라 부른다. 그리고 Multiplexed I/O는 special file들에 유용하게 쓰인다. A signle process cannot reasonably block on more than one file descriptor at the same timeif a read() system call is issued, and there is not yet any data, t..
![]()
Traversal Visit Nodes of a Graph in SOME orderMay go through a node several timeWhile doing calculation Any order traversalStart at a node s (put s into BOX)while BOX is not empty:Take one node from BOXIf node not Marked:Mark nodeDo computation on nodeput Every Adjacent node This solves ReachabilitySolution: start at s, check if t was visited with any order traversal Proof (귀류법) 위에서 제시한 any orde..
![]()
https://arxiv.org/abs/1409.0473 Neural Machine Translation by Jointly Learning to Align and TranslateNeural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the traarxiv.org Abstract. 기존의 통계적인 기계 번역기와 달리, ..
![]()
Process Specification shows process details which are implied but not shown in a DFDSpecifying the input, output, and algorithm of a module in a DFDNormally written in pseudo-code or table formatSpecifying all (upper/lower) processes in DFD hierachies프로세스 스펙 또한 DFD, SA에 포함되어야한다. 이는 DFD에 나타나지 않은 프로세스의 디테일을 보여준다. Example: Left Sensor Interface Data Dictionary Defines data elements to avoid differe..
![]()
File Offset Location in the file at which the next read() or write() will commenceExpressed as an ordinal byte position relative to the start of the fileThe first byte of the file is at offset 0Set to point to the start of the file when the file is openedAutomatically adjusted by each subsequent call to read() or write()Successive read() and write() calls progress sequentially through a file Rep..
![]()
2024.09.09 - [[Deep daiv.]/[Deep daiv.] NLP] - [Deep daiv.] NLP, WIL - 9. ELMo and GPT (1) [Deep daiv.] NLP, WIL - 9. ELMo and GPT (1)https://chloamme.github.io/2021/12/08/illustrated-gpt2-korean.html#h-%EC%9E%85%EB%A0%A5-encoding [번역] 그림으로 설명하는 GPT-2 (Transformer Language Model 시각화)이 글은 Jay Alammar님의 글을 번역한 글입니다. [추가정보] Thishw-hk.tistory.com에서 이어지는 글입니다. Self-Attention 이 글의 앞 부분에서 단어 it을 처리하는 레..
