what do you mean? if you don't change the Hamiltonian, the pushforward is indeed again canonical, no?
November 6, 2026 at 5:38 AM
what do you mean? if you don't change the Hamiltonian, the pushforward is indeed again canonical, no?
should not be canonical, but the pushforward of canonical under that flow. It is a remarkable fact that the complicated result might as well be canonical again, as far as all the average values of quantities we care about are concerned. That's wild!
November 6, 2026 at 2:27 AM
should not be canonical, but the pushforward of canonical under that flow. It is a remarkable fact that the complicated result might as well be canonical again, as far as all the average values of quantities we care about are concerned. That's wild!
I may have been confusing. I mean "canonical" here like "canonical ensemble."
We know that a thermal system "in" a canonical ensemble can be moved to another higher-entropy canonical ensemble. Any pushforward measure under symplectic dynamics would have the same (Gibbs) entropy, however.
We know that a thermal system "in" a canonical ensemble can be moved to another higher-entropy canonical ensemble. Any pushforward measure under symplectic dynamics would have the same (Gibbs) entropy, however.
November 6, 2026 at 4:49 PM
I may have been confusing. I mean "canonical" here like "canonical ensemble."
We know that a thermal system "in" a canonical ensemble can be moved to another higher-entropy canonical ensemble. Any pushforward measure under symplectic dynamics would have the same (Gibbs) entropy, however.
We know that a thermal system "in" a canonical ensemble can be moved to another higher-entropy canonical ensemble. Any pushforward measure under symplectic dynamics would have the same (Gibbs) entropy, however.
Post-Quantum Time-Lock Puzzle from Isogenies (Shweta Agrawal, Andrea Basso, Sikhar Patranabis) ia.cr/2026/2306
November 9, 2026 at 1:26 AM
Post-Quantum Time-Lock Puzzle from Isogenies (Shweta Agrawal, Andrea Basso, Sikhar Patranabis) ia.cr/2026/2306
מפות תיקון דטרמיניסטיות המוחלות על דגימות קוונטיות מצברות הסתברות על פני תמונות קדם, ומשפרות ביצועי אופטימיזציה. CE-QAOA חד-שכבתי שחזר אופטימום TSP בסביב 1,000 זריקות לעומת 10^6 בעבר, עם אימות על מעבד סופרמוליך של IBM בעל 127 קוביטים...
#אופטימיזציהקוונטית #יתרוןקוונטי #מחקר
#אופטימיזציהקוונטית #יתרוןקוונטי #מחקר
יתרון קוונטי באופטימיזציה מוגבלת דרך Pushforward Repair וצבירת Basin
iq.fp2.dev
November 3, 2026 at 5:18 AM
מפות תיקון דטרמיניסטיות המוחלות על דגימות קוונטיות מצברות הסתברות על פני תמונות קדם, ומשפרות ביצועי אופטימיזציה. CE-QAOA חד-שכבתי שחזר אופטימום TSP בסביב 1,000 זריקות לעומת 10^6 בעבר, עם אימות על מעבד סופרמוליך של IBM בעל 127 קוביטים...
#אופטימיזציהקוונטית #יתרוןקוונטי #מחקר
#אופטימיזציהקוונטית #יתרוןקוונטי #מחקר
Deterministische Reparaturkarten, die auf Quantenproben angewendet werden, aggregieren Wahrscheinlichkeit über Urbilder und verbessern die Optimierungsleistung. Single-Layer-CE-QAOA wiederhergestellte TSP-Optima in ~1.000 Schüssen gegenüber 10^6 zuvor, mit Validierung auf eine...
Quantenvorteil bei eingeschränkter Optimierung durch Pushforward-Reparatur und Basin-Aggregation
iq.fp2.dev
November 3, 2026 at 5:17 AM
Deterministische Reparaturkarten, die auf Quantenproben angewendet werden, aggregieren Wahrscheinlichkeit über Urbilder und verbessern die Optimierungsleistung. Single-Layer-CE-QAOA wiederhergestellte TSP-Optima in ~1.000 Schüssen gegenüber 10^6 zuvor, mit Validierung auf eine...
Deterministic repair maps applied to quantum samples aggregate probability across preimages, enhancing optimization performance. Single-layer CE-QAOA recovered TSP optima in ~1,000 shots versus 10^6 previously, with validation on 127-qubit IBM superconducting processor using u...
Quantum Advantage in Constrained Optimization Through Pushforward Repair and Basin Aggregation
arxiv.org
November 3, 2026 at 5:16 AM
Deterministic repair maps applied to quantum samples aggregate probability across preimages, enhancing optimization performance. Single-layer CE-QAOA recovered TSP optima in ~1,000 shots versus 10^6 previously, with validation on 127-qubit IBM superconducting processor using u...
Mapas de reparo determinísticos aplicados a amostras quânticas agregam probabilidade entre pré-imagens, melhorando o desempenho da otimização. CE-QAOA de camada única recuperou ótimas do TSP em ~1.000 disparos versus 10^6 anteriormente, com validação em processador supercondut...
Vantagem Quântica em Otimização com Restrições Através de Reparo Pushforward e Agregação de Bacia
iq.fp2.dev
November 3, 2026 at 5:19 AM
Mapas de reparo determinísticos aplicados a amostras quânticas agregam probabilidade entre pré-imagens, melhorando o desempenho da otimização. CE-QAOA de camada única recuperou ótimas do TSP em ~1.000 disparos versus 10^6 anteriormente, com validação em processador supercondut...
Les cartes de réparation déterministes appliquées aux échantillons quantiques agrègent la probabilité sur les préimages, améliorant les performances d'optimisation. Le CE-QAOA monocouche a récupéré les optima TSP en ~1 000 coups contre 10^6 auparavant, avec validation sur un p...
Avantage Quantique en Optimisation Contrainte par Réparation Pushforward et Agrégation de Bassin
iq.fp2.dev
November 3, 2026 at 5:17 AM
Les cartes de réparation déterministes appliquées aux échantillons quantiques agrègent la probabilité sur les préimages, améliorant les performances d'optimisation. Le CE-QAOA monocouche a récupéré les optima TSP en ~1 000 coups contre 10^6 auparavant, avec validation sur un p...
Los mapas de reparación determinista aplicados a muestras cuánticas agregan probabilidad entre preimágenes, mejorando el rendimiento de optimización. CE-QAOA de una sola capa recuperó óptimos de TSP en ~1,000 disparos versus 10^6 previamente, con validación en un procesador su...
Ventaja Cuántica en Optimización Restringida a Través de Reparación Pushforward y Agregación de Cuenca
arxiv.org
November 3, 2026 at 5:17 AM
Los mapas de reparación determinista aplicados a muestras cuánticas agregan probabilidad entre preimágenes, mejorando el rendimiento de optimización. CE-QAOA de una sola capa recuperó óptimos de TSP en ~1,000 disparos versus 10^6 previamente, con validación en un procesador su...
Các bản đồ sửa chữa xác định được áp dụng cho mẫu lượng tử để tập hợp xác suất trên các hình ảnh trước, nâng cao hiệu suất tối ưu hóa. CE-QAOA một lớp đã khôi phục tối ưu TSP chỉ trong ~1.000 lần quay so với 10^6 trước đó, với xác thực trên bộ xử lý siêu dẫn IBM 127 qubit sử d...
Lợi thế Lượng tử trong Tối ưu hóa Có ràng buộc Thông qua Sửa chữa Pushforward và Tập hợp Lưu vực
iq.fp2.dev
November 3, 2026 at 5:18 AM
Các bản đồ sửa chữa xác định được áp dụng cho mẫu lượng tử để tập hợp xác suất trên các hình ảnh trước, nâng cao hiệu suất tối ưu hóa. CE-QAOA một lớp đã khôi phục tối ưu TSP chỉ trong ~1.000 lần quay so với 10^6 trước đó, với xác thực trên bộ xử lý siêu dẫn IBM 127 qubit sử d...
Research shows deep generative models struggle with heavy-tailed or multimodal distributions due to transport issues. The authors propose bijective reparametrizations to improve model adaptation in finance and climate science. https://arxiv.org/abs/2610.34484
Intrinsic Gradient Barriers in Pushforward Deep Generative Models
ArXiv link for Intrinsic Gradient Barriers in Pushforward Deep Generative Models
arxiv.org
October 28, 2026 at 9:50 AM
Research shows deep generative models struggle with heavy-tailed or multimodal distributions due to transport issues. The authors propose bijective reparametrizations to improve model adaptation in finance and climate science. https://arxiv.org/abs/2610.34484
Edric Tam, Qingyang Liu, Barbara Engelhardt, Ray Bai: Intrinsic Gradient Barriers in Pushforward Deep Generative Models https://arxiv.org/abs/2610.34484 https://arxiv.org/pdf/2610.34484 https://arxiv.org/html/2610.34484
October 27, 2026 at 7:03 AM
Edric Tam, Qingyang Liu, Barbara Engelhardt, Ray Bai: Intrinsic Gradient Barriers in Pushforward Deep Generative Models https://arxiv.org/abs/2610.34484 https://arxiv.org/pdf/2610.34484 https://arxiv.org/html/2610.34484
Excited to share our preprint "Intrinsic Gradient Barriers in Pushforward Deep Generative Models" with Edric Tam, Qingyang Liu, and Barbara Engelhardt! We unveil intrinsic limitations of pushforward DGMs in learning heavy-tailed and multimodal densities. Link 👉 arxiv.org/abs/2610.34484
October 27, 2026 at 2:01 AM
Excited to share our preprint "Intrinsic Gradient Barriers in Pushforward Deep Generative Models" with Edric Tam, Qingyang Liu, and Barbara Engelhardt! We unveil intrinsic limitations of pushforward DGMs in learning heavy-tailed and multimodal densities. Link 👉 arxiv.org/abs/2610.34484
or be one more pullback/pushforward?
October 25, 2026 at 2:15 AM
or be one more pullback/pushforward?
I'm giving a talk "Intrinsic Gradient Barriers in Pushforward Generative Models" at the ASA Virginia Chapter 2026 Annual Meeting next Fri., Oct. 30! It's from 9:20-9:40 am (first of four in Session I). The paper associated with this talk should be on arXiv by early next week!
October 23, 2026 at 6:57 PM
I'm giving a talk "Intrinsic Gradient Barriers in Pushforward Generative Models" at the ASA Virginia Chapter 2026 Annual Meeting next Fri., Oct. 30! It's from 9:20-9:40 am (first of four in Session I). The paper associated with this talk should be on arXiv by early next week!
What is the pushforward of the convention?
October 12, 2026 at 1:00 PM
What is the pushforward of the convention?
PushForward Opens Steam Page For MillionCube, A 30-Player 3D Maze Game
PushForward opened the Steam page for MillionCube on October 10 and began taking wishlists. Up to 30 players search one goal room across labyrinths of up to 1,000,000 rooms, with four stages and a streamer mode.
PushForward opened the Steam page for MillionCube on October 10 and began taking wishlists. Up to 30 players search one goal room across labyrinths of up to 1,000,000 rooms, with four stages and a streamer mode.
October 11, 2026 at 4:52 AM
PushForward Opens Steam Page For MillionCube, A 30-Player 3D Maze Game
PushForward opened the Steam page for MillionCube on October 10 and began taking wishlists. Up to 30 players search one goal room across labyrinths of up to 1,000,000 rooms, with four stages and a streamer mode.
PushForward opened the Steam page for MillionCube on October 10 and began taking wishlists. Up to 30 players search one goal room across labyrinths of up to 1,000,000 rooms, with four stages and a streamer mode.
100万の部屋からたった一つのゴールを見つけ出せ！最大30人で挑める耐久系3D迷路ゲーム『ミリオンキューブ』Steamストアページを公開！
100万の部屋からたった一つのゴールを見つけ出せ！最大30人で挑める耐久系3D迷路ゲーム『ミリオンキューブ』Steamストアページを公開！
株式会社PushForwardのプレスリリース 『みんなで推理』のPushForward新作。ドアの番号と景色がヒントになり、ゴールに近づくほど演出が熱くなる。配信者モードも搭載 株式会社PushForward（本社：東 ...
gamehack.jp
October 11, 2026 at 2:56 AM
100万の部屋からたった一つのゴールを見つけ出せ！最大30人で挑める耐久系3D迷路ゲーム『ミリオンキューブ』Steamストアページを公開！
最大100万の部屋からゴールを探す『ミリオンキューブ』。PushForwardが10月11日に紹介した、最大30人で遊べる3D迷路ゲームです。10月末発売予定。
気になったのは、1万7576室ある水族館ステージでは、魚の種類がゴールへのヒントになること。大規模な迷路の話から、目の前を泳ぐ魚の観察へ。急いで出口を探すときほど、景色をよく見る必要があるのが面白い。水槽の前で立ち止まる人が、いちばん頼りになるかもしれない。
prtimes.jp/main/html/rd...
気になったのは、1万7576室ある水族館ステージでは、魚の種類がゴールへのヒントになること。大規模な迷路の話から、目の前を泳ぐ魚の観察へ。急いで出口を探すときほど、景色をよく見る必要があるのが面白い。水槽の前で立ち止まる人が、いちばん頼りになるかもしれない。
prtimes.jp/main/html/rd...
100万の部屋からたった一つのゴールを見つけ出せ！最大30人で挑める耐久系3D迷路ゲーム『ミリオンキューブ』Steamストアページを公開！
株式会社PushForwardのプレスリリース（2026年10月11日 11時00分）100万の部屋からたった一つのゴールを見つけ出せ！最大30人で挑める耐久系3D迷路ゲーム『ミリオンキューブ』Steamストアページを公開！
prtimes.jp
October 11, 2026 at 10:38 AM
最大100万の部屋からゴールを探す『ミリオンキューブ』。PushForwardが10月11日に紹介した、最大30人で遊べる3D迷路ゲームです。10月末発売予定。
気になったのは、1万7576室ある水族館ステージでは、魚の種類がゴールへのヒントになること。大規模な迷路の話から、目の前を泳ぐ魚の観察へ。急いで出口を探すときほど、景色をよく見る必要があるのが面白い。水槽の前で立ち止まる人が、いちばん頼りになるかもしれない。
prtimes.jp/main/html/rd...
気になったのは、1万7576室ある水族館ステージでは、魚の種類がゴールへのヒントになること。大規模な迷路の話から、目の前を泳ぐ魚の観察へ。急いで出口を探すときほど、景色をよく見る必要があるのが面白い。水槽の前で立ち止まる人が、いちばん頼りになるかもしれない。
prtimes.jp/main/html/rd...
MillionCube #Windows が、2026/10配信と発表されました。
Steam：MillionCube
『ミリオンキューブ』は、最大30人マルチプレイ対応の最大100万の部屋からたったひとつの正解の部屋を探す迷路探索ゲームです。ドアの文字、景色に隠れた手がかりをもとに、ゴールの部屋を目指しましょう！
store.steampowered.com
October 10, 2026 at 6:49 AM
MillionCube #Windows が、2026/10配信と発表されました。
Xin Geng, Andrey Losev, Vyacheslav Lysov: Classical BV Pushforward https://arxiv.org/abs/2609.34533 https://arxiv.org/pdf/2609.34533 https://arxiv.org/html/2609.34533
September 29, 2026 at 6:54 AM
Xin Geng, Andrey Losev, Vyacheslav Lysov: Classical BV Pushforward https://arxiv.org/abs/2609.34533 https://arxiv.org/pdf/2609.34533 https://arxiv.org/html/2609.34533
Candace Bethea, Kirsten Wickelgren: A Quadratically Enriched Pushforward for Deligne--Mumford Stacks https://arxiv.org/abs/2609.19423 https://arxiv.org/pdf/2609.19423 https://arxiv.org/html/2609.19423
September 18, 2026 at 6:46 AM
Candace Bethea, Kirsten Wickelgren: A Quadratically Enriched Pushforward for Deligne--Mumford Stacks https://arxiv.org/abs/2609.19423 https://arxiv.org/pdf/2609.19423 https://arxiv.org/html/2609.19423
This makes it clear that (linear transformation associated to) the adjacency matrix is in fact of reasonable provenance: pullback along source, pushforward along target — it's a transfer operator!
September 14, 2026 at 3:14 AM
This makes it clear that (linear transformation associated to) the adjacency matrix is in fact of reasonable provenance: pullback along source, pushforward along target — it's a transfer operator!