ML-KULeuven/deepproblog

DeepProbLog is an extension of ProbLog that integrates Probabilistic Logic Programming with deep learning by introducing the neural predicate.

DeepProbLog – Neural Probabilistic Logic Programming

What it is – DeepProbLog extends the ProbLog probabilistic logic programming language by adding neural predicates. These are facts whose probabilities are produced by a neural network, letting you combine symbolic reasoning with deep‑learning perception in a single model.

Key components

  • ProbLog integration – builds on the existing ProbLog system for defining probabilistic logical programs.
  • Neural predicates – wrap PyTorch models (e.g., CNNs from TorchVision) so that the truth‑value of a predicate is a learned probability.
  • Inference – supports exact inference via the underlying ProbLog engine and optional approximate inference when the extra dependencies (PySwip + SWI‑Prolog) are installed.
  • Examples – the repository ships a src/deepproblog/examples folder containing the experiments described in the accompanying research papers.

Installation

pip install deepproblog   # pulls in ProbLog, PySDD, PyTorch, TorchVision, etc.

For testing you also need pytest and can run the built‑in test suite with:

python -m deepproblog test

If you want to use the approximate inference mode, install the extra packages pyswip and have SWI‑Prolog available on your system.

Use case – Write a ProbLog program that reasons over uncertain facts, and let some of those facts be classified by a neural network (e.g., image recognition). DeepProbLog will train the neural components while performing probabilistic inference over the whole program.

Publications

  1. DeepProbLog: Neural Probabilistic Logic Programming (NeurIPS 2018)
  2. Neural Probabilistic Logic Programming in DeepProbLog (AIJ)
  3. Approximate Inference for Neural Probabilistic Logic Programming (KR 2021)

License – Apache 2.0 (KU Leuven, DTAI Research Group).

All information above is taken directly from the repository’s README.

Related

  • Project
    AThevon/genjutsu

    genjutsu is a Claude‑AI plugin that lets you ask an LLM to add or redesign motion‑heavy UI elements. Two main skills – `/genjutsu:cast` for small, targeted animations and `/genjutsu:paint` for full‑scale redesigns – automatically detect your stack (React/Vue/Svelte, Compose, SwiftUI), propose an interaction thesis, generate the code (GSAP, Framer Motion, Compose, SwiftUI, etc.), and run a quick audit for accessibility and performance. Install via a single ZIP bundle or individual skill ZIPs on Claude AI, Claude Code, or Cowork.

  • Project
    InternLM/InternBootcamp

    InternAgentHarness (InternBootcamp) is an open‑source framework for creating, running, and evaluating multi‑turn, tool‑augmented LLM agents. It lets researchers define synthetic tasks, expose verified external tools, manage multi‑round interactions, compute rewards, and log full trajectories for data generation or RL training.

  • Project
    ntd4996/agentpet

    AgentPet is a cross‑platform desktop app that monitors the live state of AI coding agents and gamifies the experience with a tamagotchi‑style pet that levels up based on token usage and completed sessions. It works offline, supports many agents, offers per‑project dashboards, and provides optional web sync and leaderboards.

  • Project
    kaanozhan/Frame

    Frame is an Electron‑based IDE that gives AI coding agents a persistent, git‑anchored context. It creates a standard `.frame/` folder with rules, a module map, notes, tasks and markdown‑based specs (spec/plan/tasks/outcome). Agents start each session with this context, and a built‑in conductor can run multiple specs in parallel isolated git worktrees, enforcing non‑overlapping footprints and requiring human approval before merging. Features include a multi‑terminal grid, task/spec panels, GitHub integration, fast intent‑index lookup, and support for Claude Code, Codex CLI, and Gemini CLI. Installation is via `npm install` or pre‑built binaries; the project is Apache‑2.0 licensed.