AI Engineerhttps://sunghogigio.com/https://sunghogigio.com/blog/en/minecraft-social-simulation-rebuild/https://sunghogigio.com/blog/en/minecraft-social-simulation-rebuild/Evidence first this time.Wed, 27 May 2026 01:00:00 GMThttps://sunghogigio.com/blog/en/ai-token-spending-knowledge-system/https://sunghogigio.com/blog/en/ai-token-spending-knowledge-system/12 weeks, 21.4 billion tokens.Thu, 26 Mar 2026 01:00:00 GMThttps://sunghogigio.com/blog/en/blog-migration-hugo-ghost-astro/https://sunghogigio.com/blog/en/blog-migration-hugo-ghost-astro/From sitemap debugging to AI-agent-driven blogging.Tue, 24 Mar 2026 09:00:00 GMThttps://sunghogigio.com/blog/en/astro-agent-skill/https://sunghogigio.com/blog/en/astro-agent-skill/Agents kept getting Astro 6 patterns wrong, so I put together a skill.Mon, 23 Mar 2026 05:00:00 GMThttps://sunghogigio.com/blog/en/unity-official-mcp-broken-analysis/https://sunghogigio.com/blog/en/unity-official-mcp-broken-analysis/CoplayDev/unity-mcp kept disconnecting, so I tried the official package. The relay binary returned empty tools/list. Back to the community alternative.Sun, 22 Mar 2026 09:00:00 GMThttps://sunghogigio.com/blog/en/agent-library-review-with-mcp/https://sunghogigio.com/blog/en/agent-library-review-with-mcp/Hands-on review of OpenAI Agents SDK, LangGraph, Gemini Gen AI SDK, and MCP for building AI agents. Conclusion: just use LLM Provider SDKs directly.Tue, 01 Apr 2025 17:19:12 GMThttps://sunghogigio.com/blog/en/openai-agents-sdk-review/https://sunghogigio.com/blog/en/openai-agents-sdk-review/Hands-on review of OpenAI Agents SDK comparing it to LangChain, covering tools, guardrails, tracing, and the agent loop architecture.Wed, 12 Mar 2025 17:19:12 GMThttps://sunghogigio.com/blog/en/rethinking-agent-and-tools/https://sunghogigio.com/blog/en/rethinking-agent-and-tools/The bottleneck is outside the agent.Sat, 08 Mar 2025 06:00:00 GMThttps://sunghogigio.com/blog/en/m1-gpu-acceleration/https://sunghogigio.com/blog/en/m1-gpu-acceleration/Setup guide for GPU acceleration on Apple Silicon M1 with PyTorch MPS backend and HuggingFace Transformers.Fri, 20 May 2022 00:00:00 GMThttps://sunghogigio.com/blog/en/mac-settings/https://sunghogigio.com/blog/en/mac-settings/Essential Mac setup checklist covering Rosetta 2, Homebrew for ARM and Intel, oh-my-zsh, language toggle key remapping, and menu bar customization.Thu, 19 May 2022 00:00:00 GMThttps://sunghogigio.com/blog/en/retrieval-augmented-diffusion-model/https://sunghogigio.com/blog/en/retrieval-augmented-diffusion-model/A review of the paper that enhances diffusion models using retrieval-based approaches from NLP, achieving high-fidelity image generation with fewer parameters.Wed, 18 May 2022 00:00:00 GMThttps://sunghogigio.com/blog/en/med-bert/https://sunghogigio.com/blog/en/med-bert/A review of Med-BERT, a BERT-based model pretrained on structured EHR data for disease prediction.Tue, 17 May 2022 00:00:00 GMThttps://sunghogigio.com/blog/en/condenser-cocondenser/https://sunghogigio.com/blog/en/condenser-cocondenser/Paper review of Condenser and coCondenser for dense retrieval pre-training.Mon, 16 May 2022 00:00:00 GMThttps://sunghogigio.com/blog/en/grad-cache/https://sunghogigio.com/blog/en/grad-cache/An approach that enables large-batch contrastive learning under memory constraints, similar to gradient accumulation.Tue, 12 Apr 2022 00:00:00 GMThttps://sunghogigio.com/blog/en/nv-tabular/https://sunghogigio.com/blog/en/nv-tabular/Introduction to NVTabular, NVIDIA's GPU-accelerated library for tabular data feature engineering and preprocessing.Tue, 12 Apr 2022 00:00:00 GMThttps://sunghogigio.com/blog/en/airflow-and-kubernetes/https://sunghogigio.com/blog/en/airflow-and-kubernetes/Comparison of two approaches for running Airflow on Kubernetes: deploying Airflow components as pods vs. using KubernetesExecutor with KubernetesPodOperator.Fri, 08 Apr 2022 00:00:00 GMThttps://sunghogigio.com/blog/en/split-vram-of-gpu-on-kubernetes/https://sunghogigio.com/blog/en/split-vram-of-gpu-on-kubernetes/Exploring options for splitting GPU VRAM across pods in Kubernetes, including replicas, extended resources, and GPU virtualization.Fri, 08 Apr 2022 00:00:00 GMThttps://sunghogigio.com/blog/en/githubio-basic-setting/https://sunghogigio.com/blog/en/githubio-basic-setting/Setting up a GitHub Pages blog with Jekyll: installation, configuration, YAML front matter, and theme customization.Wed, 06 Apr 2022 00:00:00 GMThttps://sunghogigio.com/blog/en/jekyll-page/https://sunghogigio.com/blog/en/jekyll-page/Guide to creating non-date-based Jekyll pages with custom permalinks, layouts, and a custom 404 page for GitHub Pages.Wed, 06 Apr 2022 00:00:00 GMThttps://sunghogigio.com/blog/en/open-graph-protocol/https://sunghogigio.com/blog/en/open-graph-protocol/Overview of Open Graph Protocol for standardized metadata in HTML, with tips on Facebook's Sharing Debugger and cache management.Wed, 06 Apr 2022 00:00:00 GMThttps://sunghogigio.com/blog/en/search-engine-registration/https://sunghogigio.com/blog/en/search-engine-registration/Step-by-step guide to registering a blog with Google Search Console, Google Analytics, and Naver Search Advisor.Wed, 06 Apr 2022 00:00:00 GMThttps://sunghogigio.com/blog/en/programmers-tuple/https://sunghogigio.com/blog/en/programmers-tuple/Solution for the Programmers tuple problem using string parsing, regex, and length-based sorting.Thu, 24 Feb 2022 10:54:54 GMThttps://sunghogigio.com/blog/en/programmers-lotto-max-min-rank/https://sunghogigio.com/blog/en/programmers-lotto-max-min-rank/C++ solution for the Programmers lotto best/worst rank problem with notes on find and vector methods.Mon, 31 Jan 2022 08:04:48 GMThttps://sunghogigio.com/blog/en/programmers-get-report-results/https://sunghogigio.com/blog/en/programmers-get-report-results/C++ solution for the Programmers report results problem using unordered_map, set, and stringstream.Sun, 30 Jan 2022 17:51:20 GMThttps://sunghogigio.com/blog/en/programmers-printer/https://sunghogigio.com/blog/en/programmers-printer/Solution for the Programmers printer queue problem using priority-based simulation.Thu, 20 Jan 2022 09:56:01 GMThttps://sunghogigio.com/blog/en/automl/https://sunghogigio.com/blog/en/automl/Introduction to AutoML and hyperparameter optimization using Bayesian Optimization with Gaussian Process Regression.Sun, 28 Nov 2021 14:04:07 GMThttps://sunghogigio.com/blog/en/model-compression-competition/https://sunghogigio.com/blog/en/model-compression-competition/Key considerations for model compression competitions, covering FLOPs, memory access cost, and speed optimization guidelines from ShuffleNet v2.Sun, 28 Nov 2021 13:26:00 GMThttps://sunghogigio.com/blog/en/model-compression-overview/https://sunghogigio.com/blog/en/model-compression-overview/Overview of model compression techniques: NAS, network pruning, knowledge distillation, tensor decomposition, and quantization.Sun, 28 Nov 2021 13:07:01 GMThttps://sunghogigio.com/blog/en/hackerrank-roads-and-libraries/https://sunghogigio.com/blog/en/hackerrank-roads-and-libraries/Solution for the HackerRank Roads and Libraries problem using DFS to minimize the cost of building roads and libraries across connected city groups.Fri, 19 Nov 2021 12:04:28 GMThttps://sunghogigio.com/blog/en/hackerrank-sherlock-and-cost/https://sunghogigio.com/blog/en/hackerrank-sherlock-and-cost/Dynamic programming solution for the HackerRank Sherlock and Cost problem, maximizing the sum of absolute differences in a sequence.Fri, 19 Nov 2021 09:34:40 GMThttps://sunghogigio.com/blog/en/hackerrank-climbing-the-leaderboard/https://sunghogigio.com/blog/en/hackerrank-climbing-the-leaderboard/Solving HackerRank's Climbing the Leaderboard problem efficiently using sorted traversal instead of repeated search.Thu, 18 Nov 2021 08:48:12 GMThttps://sunghogigio.com/blog/en/programmers-string-compression/https://sunghogigio.com/blog/en/programmers-string-compression/Solution for the Programmers string compression problem using n-gram style chunking.Wed, 17 Nov 2021 17:09:49 GMThttps://sunghogigio.com/blog/en/programmers-farthest-node/https://sunghogigio.com/blog/en/programmers-farthest-node/Solution for the Programmers farthest node problem using BFS on an undirected graph.Fri, 12 Nov 2021 08:07:04 GMThttps://sunghogigio.com/blog/en/baekjoon-z/https://sunghogigio.com/blog/en/baekjoon-z/Solution for Baekjoon 1074 (Z) using recursive quadrant division and cumulative counting.Thu, 11 Nov 2021 03:15:39 GMThttps://sunghogigio.com/blog/en/kaggle-dataset/https://sunghogigio.com/blog/en/kaggle-dataset/Quick shell script for batch-downloading Kaggle competition datasets using the Kaggle API for local server training.Wed, 10 Nov 2021 15:31:55 GMThttps://sunghogigio.com/blog/en/baekjoon-picnic/https://sunghogigio.com/blog/en/baekjoon-picnic/Solution for Baekjoon 2026 (Picnic) using DFS on a friendship graph with adjacency matrix and adjacency list.Wed, 10 Nov 2021 11:35:47 GMThttps://sunghogigio.com/blog/en/baekjoon-password-creation/https://sunghogigio.com/blog/en/baekjoon-password-creation/Solution for Baekjoon 1759 (Password Creation) using combinations with vowel and consonant constraints.Mon, 08 Nov 2021 08:12:58 GMThttps://sunghogigio.com/blog/en/programmers-filename-sort/https://sunghogigio.com/blog/en/programmers-filename-sort/Solution for the Programmers filename sort problem using regex-based splitting and custom sort keys.Tue, 02 Nov 2021 02:51:54 GMThttps://sunghogigio.com/blog/en/programmers-english-word-chain/https://sunghogigio.com/blog/en/programmers-english-word-chain/Solution for the Programmers word chain game problem using stack-based simulation and modular arithmetic.Fri, 29 Oct 2021 22:50:29 GMThttps://sunghogigio.com/blog/en/nlp-trends/https://sunghogigio.com/blog/en/nlp-trends/Survey of NLP trends including reinforcement learning for summarization and QA, and multimodal NLP+CV approaches like Show and Tell and GLAC Net.Fri, 29 Oct 2021 09:55:54 GMThttps://sunghogigio.com/blog/en/grep-process-and-kill/https://sunghogigio.com/blog/en/grep-process-and-kill/One-liner to find and kill all matching processes using ps, grep, awk, and xargs.Fri, 29 Oct 2021 08:57:49 GMThttps://sunghogigio.com/blog/en/python3-container/https://sunghogigio.com/blog/en/python3-container/Overview of Python collections containers: UserDict, OrderedDict, and deque with practical usage notes.Wed, 27 Oct 2021 00:53:46 GMThttps://sunghogigio.com/blog/en/baekjoon-gear/https://sunghogigio.com/blog/en/baekjoon-gear/Solution for Baekjoon 14891 (Gear/Cogwheel) using recursive state checking and deque-based rotation.Tue, 26 Oct 2021 20:12:15 GMThttps://sunghogigio.com/blog/en/programmers-stepping-stones/https://sunghogigio.com/blog/en/programmers-stepping-stones/Solution for the Programmers stepping stones problem using binary search on minimum distance.Fri, 22 Oct 2021 17:34:24 GMThttps://sunghogigio.com/blog/en/naver-boost-camp-ai-tech-2nd-report-eleventh-week/https://sunghogigio.com/blog/en/naver-boost-camp-ai-tech-2nd-report-eleventh-week/Week 11 retrospective of Naver Boostcamp AI Tech covering the Relation Extraction competition, model experiments, and leaderboard submissions.Fri, 22 Oct 2021 11:27:28 GMThttps://sunghogigio.com/blog/en/programmers-lifeboat/https://sunghogigio.com/blog/en/programmers-lifeboat/Solution for the Programmers lifeboat problem using a greedy two-pointer approach.Fri, 22 Oct 2021 03:45:53 GMThttps://sunghogigio.com/blog/en/programmers-h-index/https://sunghogigio.com/blog/en/programmers-h-index/Solution for the Programmers H-Index problem using sorting and linear scan.Fri, 22 Oct 2021 02:44:44 GMThttps://sunghogigio.com/blog/en/programmers-disguise/https://sunghogigio.com/blog/en/programmers-disguise/Solution to the Programmers 'Disguise' hash problem using combinatorics to count clothing outfit combinations.Thu, 21 Oct 2021 17:35:00 GMThttps://sunghogigio.com/blog/en/programmers-phone-book/https://sunghogigio.com/blog/en/programmers-phone-book/Solution for the Programmers phone book problem using sorting and prefix comparison.Thu, 21 Oct 2021 14:57:35 GMThttps://sunghogigio.com/blog/en/qa-with-phrase-retrieval/https://sunghogigio.com/blog/en/qa-with-phrase-retrieval/Overview of phrase retrieval for open-domain QA, using dense-sparse representations to bypass the retriever-reader pipeline.Wed, 20 Oct 2021 06:13:25 GMThttps://sunghogigio.com/blog/en/programmers-immigration/https://sunghogigio.com/blog/en/programmers-immigration/Solution for the Programmers immigration problem using binary search on time cost.Wed, 20 Oct 2021 02:11:09 GMThttps://sunghogigio.com/blog/en/negative-in-batch/https://sunghogigio.com/blog/en/negative-in-batch/An approach that enables large-batch contrastive learning under memory constraints by using in-batch negatives instead of explicit negative sampling.Mon, 18 Oct 2021 17:53:53 GMThttps://sunghogigio.com/blog/en/reducing-training-bias/https://sunghogigio.com/blog/en/reducing-training-bias/Types of training bias in MRC and ODQA, and techniques to mitigate them including negative sampling and annotation bias handling.Mon, 18 Oct 2021 03:44:02 GMThttps://sunghogigio.com/blog/en/linking-mrc-and-retrieval/https://sunghogigio.com/blog/en/linking-mrc-and-retrieval/Deep dive into Open Domain Question Answering, covering the Retriever-Reader approach, distant supervision, and passage granularity strategies.Sun, 17 Oct 2021 08:45:26 GMThttps://sunghogigio.com/blog/en/passage-retrieval-scaling-up/https://sunghogigio.com/blog/en/passage-retrieval-scaling-up/Scaling passage retrieval with approximate similarity search techniques: scalar quantization, inverted file indexing, and FAISS.Sun, 17 Oct 2021 08:32:29 GMThttps://sunghogigio.com/blog/en/dense-embedding/https://sunghogigio.com/blog/en/dense-embedding/Dense embedding for passage retrieval: training bi-encoders with contrastive learning and negative sampling strategies.Thu, 14 Oct 2021 06:43:17 GMThttps://sunghogigio.com/blog/en/passage-retrieval/https://sunghogigio.com/blog/en/passage-retrieval/Fundamentals of passage retrieval for open-domain QA, covering sparse embedding, TF-IDF, and similarity search in vector space.Wed, 13 Oct 2021 15:00:36 GMThttps://sunghogigio.com/blog/en/baekjoon-2048-easy/https://sunghogigio.com/blog/en/baekjoon-2048-easy/Solution for Baekjoon 12100 (2048 Easy) using brute-force simulation with block merging and four-directional movement.Wed, 13 Oct 2021 14:59:23 GMThttps://sunghogigio.com/blog/en/generation-based-mrc/https://sunghogigio.com/blog/en/generation-based-mrc/Generation-based MRC using seq2seq models like BART: comparison with extraction-based approaches, preprocessing, and decoding strategies.Wed, 13 Oct 2021 04:58:10 GMThttps://sunghogigio.com/blog/en/extraction-based-mrc/https://sunghogigio.com/blog/en/extraction-based-mrc/Extraction-based machine reading comprehension: span prediction with BERT, preprocessing, fine-tuning, and post-processing.Wed, 13 Oct 2021 04:09:10 GMThttps://sunghogigio.com/blog/en/latest-models/https://sunghogigio.com/blog/en/latest-models/Survey of post-BERT NLP models including XLNet, RoBERTa, BART, T5, Meena, and controllable language models like PPLM.Wed, 13 Oct 2021 01:49:14 GMThttps://sunghogigio.com/blog/en/gpt-language-model/https://sunghogigio.com/blog/en/gpt-language-model/Evolution of GPT language models from GPT-1 to GPT-3: architecture changes, zero/few-shot learning, and scaling effects.Tue, 12 Oct 2021 05:01:34 GMThttps://sunghogigio.com/blog/en/unicode-tokenization/https://sunghogigio.com/blog/en/unicode-tokenization/Basics of Unicode handling in Python and subword tokenization methods including BPE (Byte-Pair Encoding).Tue, 12 Oct 2021 02:26:14 GMThttps://sunghogigio.com/blog/en/mrc/https://sunghogigio.com/blog/en/mrc/Introduction to Machine Reading Comprehension covering dataset types, key challenges like coreference resolution and multi-hop reasoning, and evaluation metrics.Tue, 12 Oct 2021 01:31:41 GMThttps://sunghogigio.com/blog/en/naver-boost-camp-ai-tech-2nd-report-tenth-week/https://sunghogigio.com/blog/en/naver-boost-camp-ai-tech-2nd-report-tenth-week/Week 10 of Naver Boostcamp AI Tech with a retrospective on the Relation Extraction competition.Fri, 08 Oct 2021 08:42:13 GMThttps://sunghogigio.com/blog/en/nvidia-driver/https://sunghogigio.com/blog/en/nvidia-driver/Troubleshooting NVIDIA driver installation issues on Ubuntu, resolved by following the official NVIDIA documentation.Wed, 06 Oct 2021 21:59:12 GMThttps://sunghogigio.com/blog/en/linux-server-sleep/https://sunghogigio.com/blog/en/linux-server-sleep/Troubleshooting and fixing unexpected auto-sleep shutdowns on a Linux server by disabling systemd sleep targets.Sun, 03 Oct 2021 21:06:55 GMThttps://sunghogigio.com/blog/en/naver-boost-camp-ai-tech-2nd-report-ninth-week/https://sunghogigio.com/blog/en/naver-boost-camp-ai-tech-2nd-report-ninth-week/Week 9 of Naver Boostcamp AI Tech covering NLP competition setup, model experimentation, and leaderboard submissions.Fri, 01 Oct 2021 08:50:31 GMThttps://sunghogigio.com/blog/en/sentence-token-classification/https://sunghogigio.com/blog/en/sentence-token-classification/Token-level classification tasks with BERT, covering Named Entity Recognition (NER) and POS tagging for Korean text.Tue, 28 Sep 2021 12:22:48 GMThttps://sunghogigio.com/blog/en/klue-dependency-parsing-single-sentence-classification/https://sunghogigio.com/blog/en/klue-dependency-parsing-single-sentence-classification/Overview of Korean dependency parsing and single sentence classification tasks in KLUE, including BERT-based model architecture and training pipeline.Tue, 28 Sep 2021 08:46:25 GMThttps://sunghogigio.com/blog/en/bert-sentence-pair-classification/https://sunghogigio.com/blog/en/bert-sentence-pair-classification/BERT-based sentence pair classification covering NLI, semantic text pair similarity, and information retrieval QA.Tue, 28 Sep 2021 08:35:45 GMThttps://sunghogigio.com/blog/en/training-bert/https://sunghogigio.com/blog/en/training-bert/BERT training pipeline details: tokenizer creation, dataset construction with NSP, masking strategies, segment handling, and truncation.Tue, 28 Sep 2021 02:22:24 GMThttps://sunghogigio.com/blog/en/bert-applications/https://sunghogigio.com/blog/en/bert-applications/Practical applications of BERT for sentiment analysis, relation extraction, semantic similarity, NER, and machine reading comprehension.Mon, 27 Sep 2021 16:09:51 GMThttps://sunghogigio.com/blog/en/regex-practice/https://sunghogigio.com/blog/en/regex-practice/Hands-on regex practice dissecting a password validation pattern with lookaheads and character classes.Mon, 27 Sep 2021 12:44:21 GMThttps://sunghogigio.com/blog/en/language-model-benchmark-summary/https://sunghogigio.com/blog/en/language-model-benchmark-summary/Summary of language model benchmarks from GLUE to multilingual evaluations, covering the evolution from RNN to ELMo, BERT, and beyond.Sun, 26 Sep 2021 22:12:41 GMThttps://sunghogigio.com/blog/en/ai-and-ethics/https://sunghogigio.com/blog/en/ai-and-ethics/Discussion of AI bias, fairness, and privacy issues, including the COMPAS case study and COVID contact tracing privacy analysis.Sun, 26 Sep 2021 10:10:36 GMThttps://sunghogigio.com/blog/en/full-stack-ml-engineer/https://sunghogigio.com/blog/en/full-stack-ml-engineer/What it means to be a full-stack ML engineer: roles, ML product lifecycle, team structure, and trade-offs.Sun, 26 Sep 2021 09:41:46 GMThttps://sunghogigio.com/blog/en/ai-copyright-law/https://sunghogigio.com/blog/en/ai-copyright-law/Overview of copyright and licensing considerations for AI projects, covering CCL, fair use, and the legal gray areas of AI-generated content.Sun, 26 Sep 2021 08:46:19 GMThttps://sunghogigio.com/blog/en/naver-boost-camp-ai-tech-2nd-report-eighth-week/https://sunghogigio.com/blog/en/naver-boost-camp-ai-tech-2nd-report-eighth-week/Week 8 retrospective of Naver Boostcamp AI Tech covering NLP lectures, MLOps tooling research, and validation set distribution discussions.Fri, 24 Sep 2021 09:24:46 GMThttps://sunghogigio.com/blog/en/quant-trading-researcher/https://sunghogigio.com/blog/en/quant-trading-researcher/Introduction to quantitative trading strategies including arbitrage, market making, and statistical arbitrage.Thu, 23 Sep 2021 09:51:47 GMThttps://sunghogigio.com/blog/en/kaggle-tip/https://sunghogigio.com/blog/en/kaggle-tip/Practical tips for Kaggle competitions covering ranking systems, competition types, validation strategies, ensemble methods, and hardware recommendations.Thu, 23 Sep 2021 06:35:03 GMThttps://sunghogigio.com/blog/en/ai-model-as-service/https://sunghogigio.com/blog/en/ai-model-as-service/End-to-end process of deploying AI models as services, from requirements analysis and dataset construction to modeling, testing, and team organization.Thu, 23 Sep 2021 05:09:46 GMThttps://sunghogigio.com/blog/en/mlops-summary/https://sunghogigio.com/blog/en/mlops-summary/Curated list of MLOps tools and frameworks including Wandb, Hydra, DVC, ONNX, and CI/CD options for ML project workflows.Wed, 22 Sep 2021 17:27:56 GMThttps://sunghogigio.com/blog/en/nlp-confusing-points/https://sunghogigio.com/blog/en/nlp-confusing-points/Clarifying commonly confused NLP concepts: argmax vs. multinomial sampling, beam search vs. sampling, and Pre-LN vs. Post-LN in GPT and Transformer.Mon, 20 Sep 2021 05:49:40 GMThttps://sunghogigio.com/blog/en/recent-trends-of-nlp/https://sunghogigio.com/blog/en/recent-trends-of-nlp/Survey of recent NLP trends covering QA, open-domain chatbots, unsupervised machine translation, text style transfer, and in-context learning.Sun, 19 Sep 2021 13:41:16 GMThttps://sunghogigio.com/blog/en/advanced-self-supervised-pre-training-model/https://sunghogigio.com/blog/en/advanced-self-supervised-pre-training-model/Overview of GPT-2, GPT-3, and ALBERT covering zero/few-shot learning, model scaling, and efficient pre-training approaches.Sat, 18 Sep 2021 11:08:52 GMThttps://sunghogigio.com/blog/en/self-supervised-pre-training-models/https://sunghogigio.com/blog/en/self-supervised-pre-training-models/Overview of self-supervised pre-training models including GPT-1 and BERT, covering masked language modeling and transfer learning.Sat, 18 Sep 2021 10:51:48 GMThttps://sunghogigio.com/blog/en/naver-boost-camp-ai-tech-2nd-report-fifth-week/https://sunghogigio.com/blog/en/naver-boost-camp-ai-tech-2nd-report-fifth-week/Week 7 retrospective of Naver Boostcamp AI Tech covering Transformer and BERT studies, mentoring Q&A on attention mechanisms and positional encoding.Fri, 17 Sep 2021 11:34:08 GMThttps://sunghogigio.com/blog/en/transformer-introduction/https://sunghogigio.com/blog/en/transformer-introduction/Introduction to the Transformer architecture: self-attention with Query/Key/Value, scaled dot-product attention, and how it overcomes RNN limitations.Mon, 13 Sep 2021 09:25:25 GMThttps://sunghogigio.com/blog/en/boostcamp-ai-tech-2nd-week6-study-summary/https://sunghogigio.com/blog/en/boostcamp-ai-tech-2nd-week6-study-summary/Week 6 study summary of Naver Boostcamp AI Tech covering NLP lectures, bucketing for data batching, and peer session discussions.Fri, 10 Sep 2021 10:21:43 GMThttps://sunghogigio.com/blog/en/bleu/https://sunghogigio.com/blog/en/bleu/BLEU score for machine translation evaluation: precision, recall, F1, n-gram overlap, and brevity penalty.Fri, 10 Sep 2021 05:52:17 GMThttps://sunghogigio.com/blog/en/beam-search/https://sunghogigio.com/blog/en/beam-search/Explanation of beam search decoding as a trade-off between greedy decoding and exhaustive search in sequence generation.Wed, 08 Sep 2021 10:39:16 GMThttps://sunghogigio.com/blog/en/attention/https://sunghogigio.com/blog/en/attention/Explanation of the attention mechanism in Seq2Seq models, covering attention vectors, context vectors, teacher forcing, and similarity functions.Wed, 08 Sep 2021 05:27:10 GMThttps://sunghogigio.com/blog/en/lstm-gru/https://sunghogigio.com/blog/en/lstm-gru/Detailed explanation of LSTM and GRU architectures, covering gate mechanisms, cell state, and how they solve RNN's long-term dependency problem.Tue, 07 Sep 2021 06:49:00 GMThttps://sunghogigio.com/blog/en/rnns/https://sunghogigio.com/blog/en/rnns/Deep dive into RNN types (one-to-one, one-to-many, many-to-many), character-level language models, and LSTM/GRU architectures.Tue, 07 Sep 2021 04:06:29 GMThttps://sunghogigio.com/blog/en/pathlib/https://sunghogigio.com/blog/en/pathlib/Introduction to Python's pathlib module for cleaner, object-oriented file path manipulation as an alternative to os.path.Tue, 07 Sep 2021 01:10:10 GMThttps://sunghogigio.com/blog/en/nlp-preprocessing/https://sunghogigio.com/blog/en/nlp-preprocessing/NLP text preprocessing techniques: stopword removal, lemmatization, and punctuation handling with NLTK and spaCy.Mon, 06 Sep 2021 23:56:25 GMThttps://sunghogigio.com/blog/en/word-embedding/https://sunghogigio.com/blog/en/word-embedding/Introduction to word embedding and Word2Vec: how words are mapped to vector spaces to capture semantic relationships.Mon, 06 Sep 2021 11:04:54 GMThttps://sunghogigio.com/blog/en/first-ai-competition-review/https://sunghogigio.com/blog/en/first-ai-competition-review/Retrospective on a mask/age/gender image classification competition: model selection, augmentation, CutMix, and lessons learned.Mon, 06 Sep 2021 11:01:54 GMThttps://sunghogigio.com/blog/en/transformationalbumentation/https://sunghogigio.com/blog/en/transformationalbumentation/Using the Albumentations library for image augmentation in PyTorch, with example pipelines for training and TTA.Mon, 06 Sep 2021 10:17:15 GMThttps://sunghogigio.com/blog/en/bag-of-words/https://sunghogigio.com/blog/en/bag-of-words/Introduction to Bag-of-Words text representation with one-hot encoding and Naive Bayes classification.Mon, 06 Sep 2021 07:04:30 GMThttps://sunghogigio.com/blog/en/nlp-overview/https://sunghogigio.com/blog/en/nlp-overview/Overview of NLP fundamentals including tokenization, NER, sentiment analysis, machine translation, and the impact of Transformer on the field.Mon, 06 Sep 2021 04:45:14 GMThttps://sunghogigio.com/blog/en/reinforcement-learning-study-materials/https://sunghogigio.com/blog/en/reinforcement-learning-study-materials/Curated list of reinforcement learning study resources including lectures, books, and blog posts.Fri, 03 Sep 2021 14:10:16 GMThttps://sunghogigio.com/blog/en/programmers-word-conversion/https://sunghogigio.com/blog/en/programmers-word-conversion/Solution for the Programmers word conversion problem using BFS with single-character diff detection.Thu, 02 Sep 2021 20:28:41 GMThttps://sunghogigio.com/blog/en/baekjoon-messi-gimossi/https://sunghogigio.com/blog/en/baekjoon-messi-gimossi/Solution for Baekjoon 17297 (Messi Gimossi) using Fibonacci-based string search with a binary search-like approach.Thu, 02 Sep 2021 20:17:18 GMThttps://sunghogigio.com/blog/en/final-score-improvement/https://sunghogigio.com/blog/en/final-score-improvement/Last-minute competition score boosting: test-time augmentation (TTA) with soft voting and half-precision training.Thu, 02 Sep 2021 01:27:24 GMThttps://sunghogigio.com/blog/en/confusing-numpy/https://sunghogigio.com/blog/en/confusing-numpy/Common NumPy pitfalls: dtype=object issues with mixed-type lists and axis behavior in np.mean.Wed, 01 Sep 2021 20:33:06 GMThttps://sunghogigio.com/blog/en/ensemble/https://sunghogigio.com/blog/en/ensemble/Ensemble methods for AI competitions: hard voting, soft voting, and weighted voting to improve model performance.Tue, 31 Aug 2021 18:22:32 GMThttps://sunghogigio.com/blog/en/training-proecss/https://sunghogigio.com/blog/en/training-proecss/How gradient accumulation works in PyTorch for effective large-batch training on limited GPU memory.Mon, 30 Aug 2021 18:02:28 GMThttps://sunghogigio.com/blog/en/additional-training-techniques/https://sunghogigio.com/blog/en/additional-training-techniques/Practical training techniques including AMP, label smoothing, ArcFace loss, class pivot adjustment, and Wandb logging for image classification.Mon, 30 Aug 2021 15:13:38 GMThttps://sunghogigio.com/blog/en/confusing-training-methods/https://sunghogigio.com/blog/en/confusing-training-methods/Clarifying common mistakes in PyTorch training loops, validation ordering, and K-fold cross validation usage.Sun, 29 Aug 2021 01:31:27 GMThttps://sunghogigio.com/blog/en/cutmix-vertical/https://sunghogigio.com/blog/en/cutmix-vertical/Applying vertical CutMix augmentation for face mask classification to focus patches on facial regions.Sat, 28 Aug 2021 14:48:06 GMThttps://sunghogigio.com/blog/en/cutmix/https://sunghogigio.com/blog/en/cutmix/CutMix data augmentation in PyTorch: loss computation, accuracy, and F1 score calculation.Fri, 27 Aug 2021 23:02:28 GMThttps://sunghogigio.com/blog/en/naver-boost-camp-ai-tech-2nd-report-fourth-week/https://sunghogigio.com/blog/en/naver-boost-camp-ai-tech-2nd-report-fourth-week/Week 4 retrospective of Naver Boostcamp AI Tech covering image classification competition, CNN model comparisons, and CutMix augmentation.Fri, 27 Aug 2021 08:55:22 GMThttps://sunghogigio.com/blog/en/ti-coding-20210825-27/https://sunghogigio.com/blog/en/ti-coding-20210825-27/TIL notes on wandb logging, fine-tuning pre-trained models, early stopping, and ensemble learning during an image classification competition.Thu, 26 Aug 2021 23:45:48 GMThttps://sunghogigio.com/blog/en/til-train-plan-20210824/https://sunghogigio.com/blog/en/til-train-plan-20210824/Day 2 competition log covering ensemble strategies (bagging vs. boosting) and CNN model selection for image classification.Tue, 24 Aug 2021 15:53:05 GMThttps://sunghogigio.com/blog/en/data-generation/https://sunghogigio.com/blog/en/data-generation/Tips on efficient data feeding, transform ordering for performance, and using Albumentations for image augmentation.Tue, 24 Aug 2021 02:35:24 GMThttps://sunghogigio.com/blog/en/preprocessing/https://sunghogigio.com/blog/en/preprocessing/Image preprocessing and data augmentation techniques for deep learning competitions, including resizing, bounding box cropping, and torchvision transforms.Tue, 24 Aug 2021 02:04:59 GMThttps://sunghogigio.com/blog/en/til-train-plan-20210823/https://sunghogigio.com/blog/en/til-train-plan-20210823/Training plan for a mask classification competition: per-feature models with ensemble learning, plus Jupyter vs. CLI workflow notes.Mon, 23 Aug 2021 16:57:30 GMThttps://sunghogigio.com/blog/en/ai-competition/https://sunghogigio.com/blog/en/ai-competition/Guide to approaching AI competitions with EDA strategies for image classification tasks.Mon, 23 Aug 2021 04:42:14 GMThttps://sunghogigio.com/blog/en/facet/https://sunghogigio.com/blog/en/facet/Using matplotlib facets with GridSpec and subplots to visualize datasets from multiple perspectives.Sun, 22 Aug 2021 21:04:31 GMThttps://sunghogigio.com/blog/en/color/https://sunghogigio.com/blog/en/color/Color usage in data visualization: sequential and diverging colormaps, HSI color space, and color palettes.Sun, 22 Aug 2021 20:58:50 GMThttps://sunghogigio.com/blog/en/trouble-shooting/https://sunghogigio.com/blog/en/trouble-shooting/Common PyTorch troubleshooting tips for GPU memory issues, OOM errors, tensor accumulation, and inference best practices.Sun, 22 Aug 2021 20:10:07 GMThttps://sunghogigio.com/blog/en/hyperparameter-tuning/https://sunghogigio.com/blog/en/hyperparameter-tuning/Overview of hyperparameter tuning methods including grid search, random search, and Bayesian optimization, with an introduction to Ray for parallel tuning.Sun, 22 Aug 2021 19:43:58 GMThttps://sunghogigio.com/blog/en/tqdm-with-epoch-statics/https://sunghogigio.com/blog/en/tqdm-with-epoch-statics/Quick tip on displaying training accuracy and loss inside tqdm progress bars in PyTorch.Sun, 22 Aug 2021 13:36:09 GMThttps://sunghogigio.com/blog/en/naver-boost-camp-ai-tech-2nd-report-thrid-week/https://sunghogigio.com/blog/en/naver-boost-camp-ai-tech-2nd-report-thrid-week/Week 3 of Naver Boostcamp AI Tech covering PyTorch custom models, datasets, dataloaders, and data visualization.Fri, 20 Aug 2021 08:16:21 GMThttps://sunghogigio.com/blog/en/multi-gpu/https://sunghogigio.com/blog/en/multi-gpu/Guide to multi-GPU training in PyTorch covering model parallelism, DataParallel, and DistributedDataParallel with code examples.Fri, 20 Aug 2021 08:14:42 GMThttps://sunghogigio.com/blog/en/pytorch-transfer/https://sunghogigio.com/blog/en/pytorch-transfer/Transfer learning in PyTorch: freezing layers, stepping frozen strategies, and model file formats.Fri, 20 Aug 2021 04:11:55 GMThttps://sunghogigio.com/blog/en/pandas/https://sunghogigio.com/blog/en/pandas/Quick reference for pandas essentials: DataFrame, Series, read_csv, indexing with loc and iloc.Thu, 19 Aug 2021 02:17:47 GMThttps://sunghogigio.com/blog/en/pytorch-dataset-dataloader/https://sunghogigio.com/blog/en/pytorch-dataset-dataloader/Guide to PyTorch Dataset and DataLoader: data flow, tensor conversion, samplers, collate_fn, and transforms.Thu, 19 Aug 2021 01:53:16 GMThttps://sunghogigio.com/blog/en/pytorch-apply/https://sunghogigio.com/blog/en/pytorch-apply/How PyTorch's nn.Module.apply() maps a function across all submodules via postorder traversal.Thu, 19 Aug 2021 00:07:50 GMThttps://sunghogigio.com/blog/en/pytorch-hook/https://sunghogigio.com/blog/en/pytorch-hook/How to use PyTorch hooks for inspecting and modifying forward/backward passes in nn.Module.Wed, 18 Aug 2021 23:40:41 GMThttps://sunghogigio.com/blog/en/pytorch-template-tip/https://sunghogigio.com/blog/en/pytorch-template-tip/PyTorch project tips: template structure, getattr for dynamic config, and abstract methods.Tue, 17 Aug 2021 13:22:09 GMThttps://sunghogigio.com/blog/en/pytorch-basic-syntax/https://sunghogigio.com/blog/en/pytorch-basic-syntax/Essential PyTorch operations: view vs reshape, squeeze, matrix multiplication, index_select, gather, and more.Tue, 17 Aug 2021 13:07:14 GMThttps://sunghogigio.com/blog/en/bar-plot/https://sunghogigio.com/blog/en/bar-plot/Guide to bar plot visualization techniques in matplotlib, covering stacked, overlapped, and grouped bar charts with best practices.Mon, 16 Aug 2021 12:14:26 GMThttps://sunghogigio.com/blog/en/programmers-network-python/https://sunghogigio.com/blog/en/programmers-network-python/Solution for the Programmers network problem using DFS to count connected components.Sun, 15 Aug 2021 19:49:33 GMThttps://sunghogigio.com/blog/en/programmers-target-number-python/https://sunghogigio.com/blog/en/programmers-target-number-python/Solution for the Programmers target number problem with DFS, BFS, and itertools.product approaches.Sun, 15 Aug 2021 15:52:19 GMThttps://sunghogigio.com/blog/en/generative-model-2/https://sunghogigio.com/blog/en/generative-model-2/Latent variable models explained: variational inference, ELBO, VAE, and adversarial auto-encoders (AAE).Sat, 14 Aug 2021 12:01:37 GMThttps://sunghogigio.com/blog/en/generative-models/https://sunghogigio.com/blog/en/generative-models/Introduction to generative models: probability distributions, independence assumptions, chain rule, and auto-regressive models.Fri, 13 Aug 2021 15:08:09 GMThttps://sunghogigio.com/blog/en/transformer-practice/https://sunghogigio.com/blog/en/transformer-practice/Hands-on PyTorch implementation of Scaled Dot-Product Attention and Multi-Head Attention for the Transformer.Fri, 13 Aug 2021 13:09:33 GMThttps://sunghogigio.com/blog/en/transformer/https://sunghogigio.com/blog/en/transformer/Core concepts of the Transformer model: encoder self-attention, Query/Key/Value embeddings, and multi-head attention mechanics.Fri, 13 Aug 2021 11:44:01 GMThttps://sunghogigio.com/blog/en/naver-boost-camp-ai-tech-2nd-report-second-week/https://sunghogigio.com/blog/en/naver-boost-camp-ai-tech-2nd-report-second-week/Week 2 of Naver Boostcamp AI Tech covering deep learning basics, data visualization, and sample variance with degrees of freedom.Fri, 13 Aug 2021 11:33:19 GMThttps://sunghogigio.com/blog/en/parameter-count/https://sunghogigio.com/blog/en/parameter-count/Discussion on how the relationship between model parameter count and generalization performance shifted with scaling law findings.Fri, 13 Aug 2021 09:58:27 GMThttps://sunghogigio.com/blog/en/convolution-practice/https://sunghogigio.com/blog/en/convolution-practice/Practical CNN implementation in PyTorch: add_module, training loops, and batch normalization.Thu, 12 Aug 2021 05:33:51 GMThttps://sunghogigio.com/blog/en/reference-blog/https://sunghogigio.com/blog/en/reference-blog/A curated list of reference blogs for machine learning and engineering study.Wed, 11 Aug 2021 08:57:12 GMThttps://sunghogigio.com/blog/en/cnn-key-concept/https://sunghogigio.com/blog/en/cnn-key-concept/Key CNN architectures from ILSVRC: AlexNet, VGGNet, GoogLeNet, and ResNet, with analysis of receptive fields and 1x1 convolutions.Wed, 11 Aug 2021 06:50:47 GMThttps://sunghogigio.com/blog/en/weight-init/https://sunghogigio.com/blog/en/weight-init/Why weight initialization matters in deep learning and why zero initialization should be avoided.Wed, 11 Aug 2021 06:35:45 GMThttps://sunghogigio.com/blog/en/convolution/https://sunghogigio.com/blog/en/convolution/Convolution fundamentals: stride, padding, parameter counting, and 1x1 convolutions.Wed, 11 Aug 2021 02:38:17 GMThttps://sunghogigio.com/blog/en/optimizer-practice/https://sunghogigio.com/blog/en/optimizer-practice/Practical comparison of SGD, Momentum, and Adam optimizers on function approximation with noisy data in PyTorch.Tue, 10 Aug 2021 07:58:31 GMThttps://sunghogigio.com/blog/en/optimization/https://sunghogigio.com/blog/en/optimization/Deep learning optimization fundamentals: generalization, overfitting, cross-validation, bias-variance tradeoff, bootstrapping, bagging, and boosting.Tue, 10 Aug 2021 01:35:15 GMThttps://sunghogigio.com/blog/en/data-visualization/https://sunghogigio.com/blog/en/data-visualization/Data visualization concepts: data types, marks, channels, and pre-attentive attributes.Mon, 09 Aug 2021 13:06:49 GMThttps://sunghogigio.com/blog/en/rnn/https://sunghogigio.com/blog/en/rnn/Fundamentals of RNN including sequence data handling, latent autoregressive models, BPTT, and truncated backpropagation.Fri, 06 Aug 2021 07:23:27 GMThttps://sunghogigio.com/blog/en/naver-boost-camp-ai-tech-2nd-report-first-week/https://sunghogigio.com/blog/en/naver-boost-camp-ai-tech-2nd-report-first-week/Week 1 retrospective of Naver Boostcamp AI Tech covering AI math fundamentals, Python basics, and gradient descent implementation.Fri, 06 Aug 2021 05:37:46 GMThttps://sunghogigio.com/blog/en/proof-of-gradient-descent/https://sunghogigio.com/blog/en/proof-of-gradient-descent/Mathematical derivation of gradient descent for linear regression, from L2 norm cost functions to SGD.Wed, 04 Aug 2021 06:19:39 GMThttps://sunghogigio.com/blog/en/gradient-descent-basic/https://sunghogigio.com/blog/en/gradient-descent-basic/Gradient descent basics: differentiation, gradient vectors, partial derivatives, and the nabla operator.Tue, 03 Aug 2021 18:23:37 GMThttps://sunghogigio.com/blog/en/matrix-vector/https://sunghogigio.com/blog/en/matrix-vector/Linear algebra refresher covering vector norms, inner products, matrix multiplication, inverse matrices, and Moore-Penrose pseudo-inverse with NumPy examples.Tue, 03 Aug 2021 17:51:44 GMThttps://sunghogigio.com/blog/en/python-nested-function/https://sunghogigio.com/blog/en/python-nested-function/Using Python decorators to compose nested functions for cleaner code structure.Tue, 03 Aug 2021 16:47:56 GMThttps://sunghogigio.com/blog/en/python-decorator-property/https://sunghogigio.com/blog/en/python-decorator-property/How to use Python's @property decorator for clean getter/setter patterns and backward compatibility.Tue, 03 Aug 2021 16:41:17 GMThttps://sunghogigio.com/blog/en/bayesian-statistics/https://sunghogigio.com/blog/en/bayesian-statistics/Fundamentals of Bayesian statistics covering Bayes' theorem, conditional probability, posterior updating, and causality interpretation.Fri, 01 Jan 2021 00:00:00 GMThttps://sunghogigio.com/blog/en/cnn/https://sunghogigio.com/blog/en/cnn/CNN fundamentals: convolution operations, kernel mechanics, multi-dimensional convolutions, and backpropagation.Fri, 01 Jan 2021 00:00:00 GMThttps://sunghogigio.com/blog/en/dl/https://sunghogigio.com/blog/en/dl/Deep learning fundamentals: key components (data, model, loss, optimizer) and a brief history from AlexNet to GPT-3.Fri, 01 Jan 2021 00:00:00 GMThttps://sunghogigio.com/blog/en/neural-network/https://sunghogigio.com/blog/en/neural-network/Introduction to neural networks covering linear regression, softmax classification, activation functions, and why deep layers are preferred.Fri, 01 Jan 2021 00:00:00 GMThttps://sunghogigio.com/blog/en/nn-multi-layer-perceptron/https://sunghogigio.com/blog/en/nn-multi-layer-perceptron/Neural networks as function approximators: linear models, activation functions, multi-layer perceptrons, universal approximation theorem, and loss functions.Fri, 01 Jan 2021 00:00:00 GMThttps://sunghogigio.com/blog/en/numpymatmul-numpydot/https://sunghogigio.com/blog/en/numpymatmul-numpydot/Difference between numpy.matmul and numpy.dot for high-dimensional matrix multiplication.Fri, 01 Jan 2021 00:00:00 GMThttps://sunghogigio.com/blog/en/probability-theroy/https://sunghogigio.com/blog/en/probability-theroy/Probability theory foundations for machine learning: probability distributions, joint and conditional distributions, Bayes' rule, and expectation.Fri, 01 Jan 2021 00:00:00 GMThttps://sunghogigio.com/blog/en/pytorch/https://sunghogigio.com/blog/en/pytorch/Practical PyTorch tips covering parameter initialization, model.eval(), tensor views, and the training loop.Fri, 01 Jan 2021 00:00:00 GMThttps://sunghogigio.com/blog/en/statistics/https://sunghogigio.com/blog/en/statistics/Foundations of statistical modeling: parametric vs. nonparametric methods, probability distributions, MLE, and log-likelihood.Fri, 01 Jan 2021 00:00:00 GMT