# 2022 All Photonic Neural Net

## Scope

This folder is a talk family for all-photonic artificial neural network processing via nonlinear optics, paired with context from room-temperature photonic logical qubits via second-order nonlinearities. The top-level `HEADER.md` carries the archive metadata, links to arXiv 2002.07193 and 2205.08608, and paper-style abstracts. At the time of writing there is no top-level deck HTML; the actual presentation entry points are in the venue-specific subfolders.

## Talk Versions

- `2022-NTT_talk_all_photonic_NN_and_quantum/presentation.html`: Jun 30, 2022 at NTT.
- `2022-Erlangen_talk_all_photonic_NN_and_quantum/index.html`: Nov 15, 2022 at MPG, Erlangen.

The two versions are effectively the same deck with duplicated assets and identical `custom.css`; their HTML differs only in the title-slide venue/date line. Keep common content edits synchronized between both copies unless intentionally making a venue-specific variant.

## Content Themes

- Room-temperature optical quantum computing with multimode nonlinear cavities, `chi(2)` interactions, bosonic logical qubits, and coherent feedback-free error correction.
- Device realism and material progress, especially the second-harmonic-generation figure-of-merit slide.
- All-photonic linear algebra acceleration using frequency modes as neurons and pump amplitudes as weights.
- Reversible/unitary matrix multiplication, loss-driven expressivity limits, and simple classification examples.
- Optical activation functions using subharmonic and second-harmonic processes, including complex/negative-valued amplitudes.
- MNIST and iris case studies, plus the UMass postdoc/research-program closing slide.

## Assets And Patterns

- Core quantum optics assets include `cavity_drawing_optical.png`, `cavity_drawing_optical_spec.png`, `pulses_and_pops_only_pump.png`, `circuit_drawing.png`, and `literature_shg.png`.
- Neural-network assets include `ring_resonator.png`, `ring_resonator_spec.png`, `expressivity.png`, `expressivity_loss.png`, `iris.png`, `nonlin_arch.png`, `sigmoid_subharm.png`, `sigmoid_secondharm.png`, and `mnist.png`.
- The decks use reveal.js with local `dist/` and `plugin/` copies, black theme, KaTeX math, fragments, vertical slide stacks, and several `data-auto-animate` sequences.
- `custom.css` defines reusable layout and citation helpers: `.flex-h`, `figure`, `.bib`, `.border`, `.border-bottom`, plus `.bc0`-`.bc9` and `.c0`-`.c9` color classes used with math highlighting.

## Editing Guidance

- Edit the concrete venue deck, not generated/vendor files under `dist/` or `plugin/`.
- Preserve relative asset paths; each talk folder is meant to be self-contained.
- For content review, run `../tools/reveal_text.py 2022-NTT_talk_all_photonic_NN_and_quantum/presentation.html --max-chars 20000` or the analogous command on the Erlangen `index.html`.
- Use existing reveal idioms: nested `<section>` blocks for vertical stacks, `class="fragment"` for staged reveals, `ol.bib` for fixed-bottom citations, and KaTeX delimiters for equations.
- When adding a shared slide, copy the required asset into both version folders and update both HTML files. When adding a venue-only slide, note that difference in the version's local `AGENTS.md`.
- Keep the entry filenames stable: external archive links may expect `presentation.html` for NTT and `index.html` for Erlangen.