Directory
Published directory entries matching your search.
PyVarInf is a Python package that facilitates the use of Bayesian deep learning methods with variational inference for PyTorch, providing facilities to train PyTorch neural network models using variational inference. It approximates the posterior distribution over network weights with a parametric distribution.
Statsmodels is a Python package that complements scipy for statistical computations, including descriptive statistics and estimation and inference for statistical models. It provides linear regression models, generalized linear models, discrete models, time series analysis, survival analysis, multivariate methods and a range of statistical tests.
Superpowers is an agentic skills framework and software development methodology for coding agents. It prompts the agent to clarify a specification, shows the design in chunks for sign-off, then builds an implementation plan emphasizing red/green TDD, YAGNI, and DRY. A subagent-driven-development process has agents work through engineering tasks and review their work. Skills trigger automatically.
TextCL is a text preprocessing package intended for cleaning text data before use in natural language processing tasks. It can be an initial step in text analysis and in predictive, classification, or text generation models. Documented preprocessing tasks include splitting texts into sentences, filtering by language, perplexity filtering, duplicate sentence filtering using Jaccard similarity, and unsupervised outlier detection with methods such as TONMF, RPCA, and SVD.
The case against AI-generated users is an IDEO journal article that explores the use of AI-generated synthetic users for design research. It discusses alternative methods to meet research goals, compares the value of traditional design research to outputs from large language models, and offers forward-looking perspectives on design research practices.
Your daily dose of AI research from AK.
TokenVerse is a method for multi-concept personalization that uses a pre-trained text-to-image diffusion model. It can disentangle complex visual elements and attributes from a single image and combine concepts extracted from multiple images, supporting objects, accessories, materials, pose and lighting. It uses a DiT-based text-to-image model and learns modulation-space directions for each word.
TPOT (Tree-based Pipeline Optimization Tool) is a Python automated machine learning tool that optimizes machine learning pipelines using genetic programming. The project underwent a major refactoring, merging what was once called TPOT2 into the main package, with new features including genetic feature selection, multi-objective optimization and an expanded method of defining search spaces.
Twitter Text Java is the Java implementation of the twitter-text parsing library. It provides methods and classes to parse Tweets, calculate length, check validity, and parse @mentions, #hashtags, and URLs. Artifacts are published to Maven Central.
Unsloth is a desktop application for running and training LLMs and diffusion models locally. It supports model formats including GGUF and MLX, and offers fine-tuning methods such as LoRA, QLoRA, full fine tuning and reinforcement learning, along with dataset building and model export. Native builds are offered for Windows, macOS and Linux, with additional installation options.
Word-As-Image for Semantic Typography is a research method that automatically creates word-as-image illustrations, visualizing a word's meaning while preserving readability. It optimizes letter outlines guided by a pretrained Stable Diffusion model and handles a large variety of semantic concepts and any font.
