- Searching computational models and execution engines registered in cybercompute.
- Defining experiments by coupling computational models with execution engines, and defining their choice of parameter values.
- Defining larger experiments through composition of smaller experiments.
- Defining experiment collections by grouping a set of experiments with common observables.
- Comparing and contrasting observables within/across experiment collections.
- System should find which systems out of N systems are compatible, and suggest them to users
- System should type-match parameters/observables when coupling models and engines, and when composing larger experiments.
- system should validate first (before execution) and throw error
Inputs of scientific interest to an experiment
- initial values for ODE
- weights of inference-mode NNs
Outputs of scientific interest from an experiment
- variables (e.g. time)
- constants (e.g., total energy)
Q: Given a "solar system" with "parameters" and "initial conditions", what is the gravitational force X between planet X and planet Y at time T?
Q: Given a "neuronal network" with "parameters" "initial conditions", when does neuron X exhibit a membrane voltage < Y?
Q: Given a "double pendulum" with "parameters" and "initial conditions", what is the trajectory (X, Y) followed by the second bob?
- Neuroscience (Giri)
- Deep Learning (Giri, Chris)
- Quantum Chemistry (Sudhakar)
- Geoscience (Dimuthu)
- Molecular Dynamics (Sudhakar)