Parameter Prediction for Unseen Deep Architectures · HackerPapers discussions

Parameter Prediction for Unseen Deep Architectures — mentioned by _bobm

Mon, 02 Nov 2026 20:57:19 GMT

I think the subfield has been there for a long time. See a small section of papers on the topic of generating dynamic weights during inference, predicting weights from network architecture, and learning the distribution of previously trained network parameters, just to name a few keywords: Parameter Prediction for Unseen Deep Architectures [1] NeuroGen: Neural Network Parameter Generation via Large Language Models [2] Diffusion-Based Neural Network Weights Generation [3] The critical thing to me is that Astra is capable of doing it. I don't know if it was trained to do so. All this does not…

Paper and discussion history. A mention is not an endorsement.

Parameter Prediction for Unseen Deep Architectures — HN submission

Thu, 10 Mar 2022 19:13:23 GMT

Parameter Prediction for Unseen Deep Architectures

2 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Parameter Prediction for Unseen Deep Architectures — mentioned by cs702

Tue, 25 Jan 2022 17:57:50 GMT

The paper is well-written, and I would recommend reading its abstract and introduction over the OP.[a] The big idea is that a graph neural network (GNN) can be trained to predict the weights of any other neural network for performing a task, given the architecture as an input. For example, if we give the trained GNN a ResNet-50 architecture, the GNN predicts the trained weights in one forward pass, and voilà, we have a ResNet-50 that is now ready for use -- and you can always finetune it to get better performance out of it. Remarkably, the prediction of weights works for previously unseen…

Paper and discussion history. A mention is not an endorsement.

Parameter Prediction for Unseen Deep Architectures — HN submission

Sun, 28 Nov 2021 02:42:32 GMT

Hypernetworks for prediction of neural network parameters

1 point; 1 discussion reply on HN. Paper and source history. A submission is not an endorsement.