A Joyful AI Research Journey🌳😊

아나콘다: 호환 되는 프로그램을 깔아준다. 하나하나 설치X 프로그램 설치anaconda 설치https://ldjwj.github.io/kagglebook/pdf_html/env01_anaconda_install_220208.htmlAnaconda : https://www.anaconda.com/download/Visual Studio Code 설치https://ldjwj.github.io/kagglebook/pdf_html/env03_VisualStudioCode_install.htmlVisual Studio : https://code.visualstudio.com/Download

Teacher: 복습하시는 분들을 위해 제가 공유한 notion에 실습하고 찾아볼 내용을 다음과 같이 추가해 두었습니다. 참고해 주세요.https://platform.openai.com/tokenizer토큰 GPT 모델에서의 토큰(token)은 매우 간단한 개념이다. 일상 언어에서 우리가 사용하는 단어나 문장 부호와 같은 것을 생각하면 된다. GPT 모델에서는 이러한 단어나 문장 부호를 더 작은 단위로 쪼갠 것을 토큰이라고 부른다. 예를 들어, 영어에서 "cat"이라는 단어는 하나의 토큰으로 처리될 수 있지만, 더 긴 단어나 복잡한 문자는 여러 개의 토큰으로 나뉠 수 있다.https://wikidocs.net/223125입력 혹은 출력하는데 사용되는 텍스트 데이터 내 ‘의미를 가진 최소 말의 단위(형태소..

241030 Wed 1st class 오늘 배운 것 중 기억할 것을 정리했다. 6개월의 압축판이다. LLM은 메인이다. 논문 수준의 실력은 회사에 개발자, AI 엔지니어로 취업해야 가능 1개월 동안 성장하는 모습으로 포트폴리오 만든다-> 취업에 도움 된다. 프로젝트를 보고 열심히했다고 평가한다 긍정적인 언어가 중요하다 (안 좋은 말은 금방 번진다 -> 의욕 다운) 부딪혀서 넘어가야 한다 할 수 있는 것에 집중한다 모르는 것은 개인톡으로 질문, 상담 필요하면 디코 디엠 보냄 비디오 켜기 -> 배움의 속도가 다르다 과제 충실히 하기복습은 필수다.40-45분 수업, 5분은 QA, 개인 정리 시간 오전 3시간 시간, 1시간 점심, 오후 4시간 수업멘토 -> 프로젝트 업그레이드 늦은 건 없다. 4년 동안 열심히 ..

Google Scholar: 다양한 학술 자료를 검색할 수 있는 대표적인 사이트입니다. 특정 연구 분야나 논문을 찾기에 유용하며, 다양한 필터링 기능을 제공합니다.ArXiv: AI, 머신러닝, 컴퓨터 비전, 자연어 처리와 같은 컴퓨터 과학 연구에 대한 최신 논문을 공개하는 사이트입니다. 특히, RAG와 같은 최신 기술에 대한 논문을 쉽게 찾을 수 있습니다.Semantic Scholar: AI 기반 학술 논문 검색 엔진으로, 논문 추천, 인용 수 분석, 관련 연구 자료 등을 효과적으로 제공합니다.PubMed: 의학과 생물학 연구에 중점을 두고 있지만, AI와 관련된 일부 논문도 찾을 수 있습니다.미션 - 나에 맞는 GPTs 3-5개 찾기. 45분 정도에 올리기1. Thesis Buddy: 논문 작성 ste..

https://github.com/yjyuwisely/MovieSense_NLP GitHub - yjyuwisely/MovieSense_NLP: MovieSense, an NLP project that provides sentiment analysis, translation, summarization, andMovieSense, an NLP project that provides sentiment analysis, translation, summarization, and text generation services for movie reviews. - yjyuwisely/MovieSense_NLPgithub.com Aug 22 - Sep 2Page ScreenshotsBelow are some scree..

ChatGPT, OpenAICO₂ flux inversion is a method used in atmospheric science to estimate the sources and sinks of carbon dioxide (CO₂) based on observed atmospheric CO₂ concentrations. This technique combines observations of CO₂ concentrations from various platforms (such as satellites, ground stations, and aircraft) with atmospheric transport models to "invert" the observed data and deduce where C..

Using the Retrieval-Augmented Generation (RAG) method, I experimented with GPT-3.5-turbo to generate both positive and negative movie reviews.RAG enhances the model's ability to produce relevant text by retrieving the most pertinent documents before generating the output.For example, the negative review generated highlighted issues like 'a movie version of a paint-by-numbers picture' and 'a loud..

ChatGPT, OpenAIAll the optimization techniques you listed are relevant to your project, MovieSense, and can help improve the performance and efficiency of the models and methods you are using. Here's how each technique relates to your project:Relevance of Optimization Techniques to Your ProjectModel Quantization to Reduce Model Size and Speed Up InferenceRelevance: Quantization reduces the memor..

ChatGPT, OpenAIFor text generation, the evaluation metric often depends on the specific task and desired outcomes. However, some common evaluation metrics used in NLP for text generation tasks include:Perplexity:Definition: Perplexity measures how well a probability model predicts a sample. In the context of language models, lower perplexity indicates a better predictive model.Usage: It is widel..

ChatGPT, OpenAIYes, using Retrieval-Augmented Generation (RAG) would indeed be a better choice for the scenario where you want to write prompts like "write a positive review about a certain movie" or "write a negative review about a certain movie." Here’s why RAG is more suitable for this task:1. Contextual Relevance and Specificity:RAG can retrieve specific reviews or information related to the..

ChatGPT, OpenAIPretraining GPT-2 with Rotten Tomatoes data and incorporating Retrieval-Augmented Generation (RAG) with the same data are two different approaches with distinct goals and outcomes. Here’s a breakdown of the differences:1. Pretraining or Fine-Tuning GPT-2 with Rotten Tomatoes DataWhat It Is:Pretraining: Training GPT-2 from scratch using a large corpus like Rotten Tomatoes data (not..

The * in zip(*combined_dataset) is the "unpacking" operator in Python. It takes a list of tuples (in this case, combined_dataset, which consists of pairs like (review_text, label)) and "unzips" them into two separate tuples: one for texts and one for labels.In other words:texts will contain all the review texts.labels will contain all the corresponding labels.The * operator effectively transpose..

Join two tuples together:a = ("John", "Charles", "Mike")b = ("Jenny", "Christy", "Monica")x = zip(a, b)#use the tuple() function to display a readable version of the result:print(tuple(x))(('John', 'Jenny'), ('Charles', 'Christy'), ('Mike', 'Monica'))https://www.w3schools.com/python/ref_func_zip.asp W3Schools.comW3Schools offers free online tutorials, references and exercises in all the major la..

The model bert-base-uncased is used because it converts all text to lowercase before processing, ignoring case differences. This is particularly useful when case sensitivity is not important for the task, such as sentiment analysis, where "Happy" and "happy" should be treated the same. The "uncased" version is generally more efficient and performs well when the distinction between uppercase and ..

ChatGPT, OpenAINaive Bayes in Sentiment Analysis:Pros:Simplicity: Easy to implement and interpret.Efficiency: Works well with smaller datasets and requires less computational power.Baseline: Provides a strong baseline for comparison with more complex models.Cons:Assumption of Independence: Assumes features (words) are independent, which is often not true in language processing.Limited Understand..