Memory Network

[Memory Networks (Chainer implementation of End-to-End Memory Networks) https://www.slideshare.net/shuyo/memory-networks-endtoend-memory-networks- chainer] Comparison with [LSTM

Paper Description Memory Networks (MemNN) - Deep Learning Blog

OLD TITLE: Memory Network

• Natural Language Processing with Deep Learning p.99 arXiv: Memory Networks
• Strongly supervised memory network

Memory addition: $m_N\leftarrow x$ Memory retrieval:. $o_1={\mathrm{argmax}}_i{S}_O(x,m_i)$ $o_2={\mathrm{argmax}}_i{S}_O((x,m_{o_1},m_i)$

• Isn’t $o1\ne o2$ guaranteed in this form? $r={\mathrm{argmax}}_{w\in \mathcal{V}}{S}_R((x,m_{o_1},m_{o_2}),w)$

SO and SR, respectively. $s(x,y)=\Phi (X{)}^T{U}^{T}U\Phi (y)$ $\Phi$ is embedded in the D dimension The real problem is whether or not we can prepare the teacher data necessary for the training of SO and SR.

So end-to-end memory network end-to-end memory networks

---

This page is auto-translated from [/nishio/Memory Network](https://scrapbox.io/nishio/Memory Network) using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I’m very happy to spread my thought to non-Japanese readers.