Symbolic Regression via Deep Reinforcement Learning Enhanced Genetic Programming Seeding
What this work does
Uses a recurrent expression generator to seed short genetic-programming runs and trains the generator using its own samples plus selected high-performing evolved expressions.
Connection to discovery
Shows how trajectories or candidates improved by a non-neural search process can become training data for a discovery proposal policy.
Evidence
Inspected the sample-sharing algorithm and comparisons with random-restart GP and neural-only search.
Scope and limitations
Selected GP samples are off-policy relative to the generator. Symbolic-regression recovery within a fixed operator grammar does not establish general scientific autonomy.
THE RESEARCH CONVERSATION
Discussion & reactions
Share a question, an insight, or a replication experience. Sign in with GitHub to join the conversation.