home

epstein-data
Research ▼
🔍 SearchFull-text document search 🤖 Ask AIAI research assistant 🔎 Evidence MapFBI serial resolution 📷 Reverse Image SearchCLIP + face across 614K images 🧑 Find Face BETASearch 29K faces by photo 💻 Run Your OwnDownload & search locally
Explore ▼
📚 Full Text Corpus1.39M docs, 2.77M pages 🌎 Global Heatmap145 countries mentioned 📈 Coverage MapWhat's here 🌌 AtlasSemantic map · 1.29M docs ⚖ Cases53 federal & state cases · per-case briefings 🎤 DepositionsTranscribed audio & video 💬 Hear from the SurvivorsSurvivors in their own words 📖 Cover to Cover-Up24-hour public reading, synced to the video ✉ Wolff–Epstein Emails2,009 messages · 2009–2019
📷 Images92K analyzed photographs 🔍 Multi-DB SearchSearch all databases individually 🗃 All Databases14 searchable databases
Reports
News ▼
📰 NewsCoverage & reporting ⚖ Justice MonitorArrests, charges, lawsuits, firings
Source ▼
🏛 DOJ ProductionOfficial EFTA disclosures 📜 EFTA Law TextPublic Law 119-38 📁 Source Data (GitHub)Open source databases
🌐 Community ResourcesCurated external projects ✉ ContactGeneral · privacy · DMCA · press
❤️ Donate 🎧 Podcast

Research

🔍 Search Documents 🤖 Ask AI 🔎 Evidence Map 📷 Reverse Image Search 🧑 Find Face BETA 💻 Run Your Own Investigator

Explore

📚 Full Text Corpus 🌎 Global Heatmap 📈 Coverage Map 🌌 Atlas ⚖ Cases 🎤 Depositions 💬 Hear from the Survivors 📖 Cover to Cover-Up ✉ Wolff–Epstein Emails 📷 Images 🔍 Multi-DB Search 🗃 All Databases

Reports

Browse All Reports 📰 News ⚖ Justice Monitor

Source

🏛 DOJ Production 📜 EFTA Law 📁 Source Data (GitHub) 🌐 Community Resources ✉ Contact
🎧 Podcast & Newsletter ❤️ Donate Privacy Policy

EFTA02586434

← Prev Next →
Loading document…

From: Misha Gromov < Sent: Wednesday, October 11, 2017 7:47 PM To: Jeffrey E. Subject: Re: Fwd: Like Bach's comments:) On Wed, 11 Oct 2017 20:01:46 +0200, Jeffrey E. wrote: Forwarded message Fro=: Joscha Bach < Date: Wed, Oct 11, 2017 at 7:55 PM Subject: Re: To: Jeffrey Eps=ein <[email protected] <mailto:[email protected]> =gt; After skimming their paper, the idea seemed unexcitin= to me at first: basically, if we have enough feature dimensions we can al=ost always find a linear separation. This is also related to how Support V=ctor Machines work: they project the data into an extremely high-dimension=l space, find a separating hyperplane with linear regression, and then pro=ect that plane back into the original space as the separator. A similar id=a is behind Echo State networks, which use a randomly wired recurrent neur=l network and then only train the output layer with a single linear regres=ion. The authors take an existing trained neural network, and whenev=r it makes a mistake, they train a linear classifier on the network state =nd data, i.e. they try to find out when the network goes wrong. Instead of=improving the network (which is also likely to make it worse in other case=), they add an additional layer to it. For engineering, this makes a lot o= sense, because large neural networks are cheap to use and deploy but expe=sive to train. On a more philosophical level, it is tempting t= ask if that might be a general learning principle for brains: when you do='t perform well, add more control structure on top. It probably makes sens= whenever you are confident that training the existing structure won't imp=ove it that much, but unless training the weights in an existing network, =t also adds quite a few milliseconds to the processing time. There is prob=bly an optimal tradeoff for this. The other thing is that the new layer is=a linear classifier only (at least in this paper), and it is creating a lo=al override on the s

Suggest a category
Misclassified? Pick a better fit.
Community Notes
▸ People Mentioned
▸ Interest Level
Routine Notable Significant
▸ Dates Mentioned
▸ Related Topics
▸ Places & Organizations
▸ Transcription Correction
Related documents
Source Data Investigation Reports DOJ EFTA CC BY-NC-SA 4.0 Contact
Independent research project. Not affiliated with the U.S. Department of Justice, FBI, any government agency, or Anthropic. All analytical text on this site is AI-generated (Claude, Anthropic) and iteratively fact-checked against source documents, but may contain errors. Verify all claims against linked EFTA sources before citing.
Powered by Datasette  ·  ❤️ Buy me a coffee

You are leaving epstein-data.com

You are being redirected to an external website not operated by this project. We are not responsible for the content or privacy practices of external sites.

Powered by Datasette