Accelerating the Horizon of Autonomous Discovery
with NIMO
NIMO is an open-source initiative designed to drive autonomous discovery.
Offering a wide array of specialized exploration algorithms, we provide the essential toolkit for building the self-driving laboratories of tomorrow.
Implemented Algorithms
PHYSBO
Conventional Bayesian optimization method. Highly effective for finding the best objective values.
- ✔ Training data required
- ✔ Real-valued objectives
- ✔ Single objective
- ✔ Multiple objectives
BOMP
Bayesian optimization method under fixed parameter constraints. Highly effective for batch experiments of multiple materials within the same process.
- ✔ Training data required
- ✔ Real-valued objectives
- ✔ Single objective
- ✔ Multiple objectives
PTR
Optimization method focused on target range attainment. Highly effective when meeting specific property thresholds is more critical than finding the best one.
- ✔ Training data required
- ✔ Real-valued objectives
- ✔ Single objective
- ✔ Multiple objectives
BLOX
Objective free exploration method. Highly effective for gaining a bird's-eye view of the distribution within the property space.
- ✔ Training data required
- ✔ Real-valued objectives
- ✔ Single objective
- ✔ Multiple objectives
PDC
Phase diagram construction method. Highly effective for rapidly identifying boundaries of categorical objectives, beyond simple phase diagrams.
- ✔ Training data required
- ✔ Categorical objective
- ✔ Single objective
RE
Random sampling. Effecive for preparing the initial data.
- ✔ No training data required
DOE
Desing of experiments method. Effecive for preparing the initial data.
- ✔ No training data required
ES
Exahstive search method. Highly effective for guiding sequential experimentation.
- ✔ No training data required
SDL (Self-Driving Labs) Integrations
Developers
Ryo Tamura
Lead Developer
Shoichi Matsuda
Experimental Advisor
Naruki Yoshikawa
SDL Designer
Koji Tsuda
Algirithmic Advisor