AI Paper Atlas

Research map / Probabilistic and scientific modeling / Variational and generative sampling

Goal-directed Generation of Discrete Structures with Conditional Generative Models

Amina Mollaysa, Brooks Paige, Alexandros Kalousis

NeurIPS 2020 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Score-based generative modeling analysis · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Train with reward-based sampled targets instead of maximum likelihoodLSTM conditional generator · Python expressionsMAE: 13.917 -> 11.823Improved
Train with reward-based sampled targets instead of maximum likelihoodLSTM conditional generator · Python expressionsaccuracy within +-3: 0.596 -> 0.682Improved
Train with reward-based sampled targets instead of maximum likelihoodLSTM conditional generator · Python expressionsexact accuracy: 0.166 -> 0.166No clear effect
Train with reward-based sampled targets instead of maximum likelihoodLSTM conditional generator · Python expressionsnegative log-likelihood: 1.830 -> 1.986Hurt
Train with reward-based sampled targets instead of maximum likelihoodLSTM conditional generator · QM9total MSE: 10.7237 -> 7.6398Improved
Train with reward-based sampled targets instead of maximum likelihoodLSTM conditional generator · ChEMBLtotal MSE: 204.4400 -> 302.5634Mixed
Train with reward-based sampled targets instead of maximum likelihoodLSTM conditional generator · QM9valid fraction: 0.962 -> 0.989Improved
Train with reward-based sampled targets instead of maximum likelihoodLSTM conditional generator · ChEMBLvalid fraction: 0.895 -> 0.945Improved
Train with reward-based sampled targets instead of maximum likelihoodLSTM conditional generator · QM9novel fraction: 0.366 -> 0.261Hurt
Replace data augmentation (RAML-like) with entropy-regularised reward trainingLSTM conditional generator · QM9novel fraction: 0.747 -> 0.274Hurt
Replace classic data augmentation with entropy-regularised reward trainingLSTM conditional generator · QM9validity: 0.937 -> 0.977Improved
Use warm-started REINFORCE instead of reward-based sampled trainingLSTM conditional generator · QM9validity: 0.9886 -> 0.4013Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy