projects

A growing collection of my cool projects.

planning
project thumbnail

pyRDDLGym

The pyRDDLGym package is a Python toolkit to automatically generate OpenAI Gym environments from Relational Dynamic Influence Diagram Language (RDDL) description files. It serves as the official parser, simulator, and evaluation system for RDDL in Python, making it easier to create and work with dynamic stochastic environments in any pipeline where an OpenAI gym environment is required. pyRDDLGym includes customizable visualization and recording tools for debugging and interpreting plans, and supports a growing ecosystem of planning tools that aims to fill the niche on standardized benchmarking and baseline problems for the planning community.

project thumbnail

RDDL Repository

The rddlrepository hosts a collection of RDDL (Relational Dynamic Influence Diagram Language) description files. These files cover problems from a wide range of disciplines and include archived problems from previous International Planning Competitions. The repository is integrated with the pyRDDLGym ecosystem, providing a variety of environments for testing and developing planning algorithms. It also includes custom visualizers for a select subset of domains.

project thumbnail

JaxPlan

pyRDDLGym-jax (or JaxPlan as it is referred to in the literature) is an extension of the pyRDDLGym eco-system, leveraging the JAX library to automatically build differentiable simulators for any RDDL problem and do gradient-based planning. It also supports various planning methods, including deep reactive policy networks and straight-line planning, risk-aware planning, and provides tools for automatic hyper-parameter tuning.

machine learning
project thumbnail

Cascade-Correlation

The cascade-correlation-neural-networks package provides a general framework for building and training constructive feed-forward neural networks such as the Cascade-Correlation (CCNN) architecture in Python, which dynamically adds hidden units to the network during training. The package also includes wrappers for TensorFlow, Keras, Scipy, and Scikit-learn, and supports custom topologies, training algorithms, and loss functions.

miscellaneous
project thumbnail

bboptpy

The bboptpy package is a powerful and scalable suite of state-of-the-art black-box and meta-heuristic optimization algorithms (algorithms for optimizing functions without derivatives) written in C++ with a user-friendly Python interface. It offers faithful reproductions of algorithms in the literature, and robust improvements and variations on well-known algorithms, such as Covariance Matrix Adaptation Evolutionary Strategy (CMA-ES), Differential Evolution, and Particle Swarm Optimization (PSO). The package provides transparent implementations with no dependencies for easy installation and extension.