4 ms·
Scalable and Sustainable Deep Learning via Randomized Hashing
- dang 9y agoUrl changed from https://www.sciencedaily.com/releases/2017/06/170601135633.htm https://www.sciencedaily.com/releases/2017/06/170601135633.h..., which points to this.
- captaindiego 9y agoLooks like a github repo here for this paper: https://github.com/rdspring1/LSH_DeepLearning https://github.com/rdspring1/LSH_DeepLearning
- JacobiX 9y agoLocality sensitive hash is well known technique for approximate nearest search and dimensions reduction. Combining it with dnn may reduce the complexity of the dataset. But I don't know if in this setting the parameters of the lsh are learnable via backprop?
- xiphias 9y agoThey used CPU for both sparse and dense approach. It would be interesting to see price/performance comparision for dense GPU vs sparse CPU, especially as more specialized architectures for dense matrix operations are coming out