560 field photographs of five durian disease categories, collected from commercial orchards across Peninsular Malaysia between July 2025 and June 2026, with per-image capture-session identifiers. The session identifiers are the point of this release. The 560 images come from only 73 independent capture sessions. A symptomatic leaf is normally photographed several times in a few seconds from slightly different angles, and those frames are not independent observations. Split this data at image level and near-identical views of one specimen land on both sides of the train/test boundary. In our own initial partition, 79.6% of images fell in sessions that straddled a split. Re-running the identical experiment with sessions kept whole lowered macro F1 by 12.2 points on average across nine architectures, positive in all nine and as much as 18.4 in one. Group your partitions by the session column in sessions.csv. Contents. images_fullres/ — the 560 originals as captured, in class folders. images_512/ — the same images at 512 px maximum edge, which is what the models were trained and evaluated on. sessions.csv — class, filename and session for every image. splits/session_level/ — the partition reported in the paper (446/58/56). splits/image_level/ — the control partition used to measure leakage (446/54/60). Evaluate at the resolution you train at. Every figure in the paper is computed on images_512. Running the same checkpoint over images_fullres through an identical Resize(256) and CenterCrop(224) pipeline gives 77.6% instead of 72.0% on the held-out set, because the two resampling paths to 224 px are not the same. The originals are included so the collection is complete, not because they are the working copy. Classes. Algal Leaf Spot (Cephaleuros virescens), Leaf Rot (Colletotrichum spp.), Phomopsis Fruit and Stem Blight (Phomopsis durionis), Pink Disease (Erythricium salmonicolor), Root Disease (Phytophthora spp.). Pink_disease is represented by three capture sessions in the entire collection; its per-class metrics are not interpretable at that support, and it is what bounds grouped cross-validation at k = 3. Annotation. Labels were assigned by the author under the guidance of growers and extension staff with field experience in these orchards. There was no second independent rater, so no inter-rater agreement statistic is available. Consent. Images were collected on site with the orchard owner's permission, or contributed by growers who were told at the time that the images would be released publicly for research. No images contain identifiable persons. A small number show a hand holding a leaf; that framing is part of the field condition being modelled. No location is published at finer resolution than district.
Lin Ding Shan· Zenodo (CERN European Organi...· 0 citations
Computer vision---making machines interpret images---traveled from blocks-world edge finders to deep convolutional networks matching human benchmarks, and its history is AI's most complete case of representation learning's triumph. This article presents a narrative review of the field's canonical line: Roberts's 1963 machine perception of solids, Marr's 1982 computational vision, Viola and Jones's 2001 face detection, Lowe's 2004 SIFT features, Dalal and Triggs's 2005 HOG descriptors, Felzenszwalb and colleagues' 2010 deformable part models, Szeliski's 2010 synthesis, Girshick's 2015 Fast R-CNN, Long, Shelhamer, and Darrell's 2015 fully convolutional nets, Simonyan and Zisserman's 2015 VGG, He and colleagues' 2016 ResNet, and Redmon and colleagues' 2016 YOLO. The synthesis is organized around three themes: representation, in which hand-engineered features gave way to learned hierarchies; architecture, in which convolution, regions, and residual depth solved recognition's geometry; and tasks, in which classification widened into detection, segmentation, and real-time video. It is concluded that vision's deep learning settlement reorganized the field around data and compute---and that its open problems, robustness and embodiment, define the current frontier.
Zen Revista, 10 IA· Zenodo (CERN European Organi...· 0 citations
This record contains Corrected Implementation Freeze v0.1.2 for HLV-R-MECH-001. The controlling scientific protocol remains: Krūger, M. (2026). HLV-R-MECH-001: Prospective Triangle-Matched Mechanism Test of the Surviving Degree-Preserving Rewire Spectral Residual — Pre-Execution Protocol Freeze v0.1.0. Zenodo. DOI: 10.5281/zenodo.22166283 The public corrected predecessor implementation is: Krūger, M. (2026). HLV-R-MECH-001: Deterministic One-Click Engine for Triangle-Matched Rewire Mechanism Testing — Corrected Implementation Freeze v0.1.1 [Computer software]. Zenodo. DOI: 10.5281/zenodo.22170138 Version v0.1.2 corrects only the numerical-runtime bootstrap of the One-Click Colab launcher. The scientific engine itself is unchanged and remains byte-identical to the engine used in the earlier implementation freezes. Frozen scientific engine SHA-256: 317df650991120f686768ffc07d12f044f58e38ce8f2c47c083901bf1d7a8a14 The need for v0.1.2 arose after the v0.1.1 launcher correctly detected that the assigned Colab host environment did not match the frozen numerical environment but then failed during creation of an isolated Python virtual environment. The v0.1.1 execution stopped before the scientific engine began. Therefore no confirmatory HLV-R-MECH-001 spectrum was evaluated, no target QSPEC or RRESP score was computed, and no scientific mechanism verdict was exposed. The exact lower-level cause of the managed Colab virtual-environment failure is not asserted beyond the observed failure at the venv-creation stage. Version v0.1.2 removes dependence on Python venv. The corrected runtime bootstrap now operates as follows: 1. reconstruct and SHA-256 verify the byte-identical frozen scientific engine; 2. inspect the assigned host NumPy and SciPy versions; 3. if the host environment already provides the exact frozen versions, use that environment directly; 4. otherwise install exact binary packages NumPy 2.3.5 and SciPy 1.17.0 into a private target directory using pip --target; 5. launch a fresh Python subprocess with the private target directory placed first on PYTHONPATH; 6. verify that the subprocess reports exactly NumPy 2.3.5 and SciPy 1.17.0; 7. verify that both NumPy and SciPy are physically imported from the private target directory; 8. only after those checks execute the unchanged frozen HLV-R-MECH-001 scientific engine. This correction is restricted entirely to the external runtime-bootstrap layer. No scientific element of HLV-R-MECH-001 is changed. In particular, v0.1.2 does not modify: - the DG-001 target identity; - the target graph; - the R_DEG family; - the R_TRI family; - the confirmatory seed streams; - candidate ordering; - accepted-swap counts; - proposal caps; - structural admission criteria; - exact degree-sequence preservation; - exact triangle-count preservation T = 6960 in R_TRI; - the 40–45% edge-replacement-depth requirement; - the 31-control family size; - the between-family rewiring-depth gate; - QSPEC; - RRESP; - spectral-band definitions; - leave-one-out scoring; - the robust-margin threshold; - numerical scientific hard gates; - or scientific verdict logic. The prospective mechanism design therefore remains exactly the design specified by DOI 10.5281/zenodo.22166283. The two confirmatory control families remain: R_DEG: fresh degree-preserving structural rewires of the fixed DG-001 target graph. R_TRI: fresh rewires preserving both the exact labelled target degree sequence and the exact global target triangle count T = 6960. Each family requires 31 accepted controls. The scientific structural firewall is also unchanged. The complete R_DEG and R_TRI control banks must be generated, structurally validated, written to disk, and hash-fixed before any confirmatory spectral evaluation may begin. No control may be admitted or rejected using eigenvalues, QSPEC, RRESP, target-control spectral distances, band scores, or scientific verdict information. The primary signatures remain the prospectively frozen QSPEC and RRESP observables inherited from the DS-SPEC-001R chain. The underlying mechanism motivation also remains unchanged. For a simple graph Laplacian L = D - A, Tr(L) = sum_i d_i Tr(L^2) = sum_i d_i^2 + sum_i d_i Tr(L^3) = sum_i d_i^3 + 3 sum_i d_i^2 - 6T. Thus the R_TRI controls match the target exactly in the first three raw Laplacian spectral moments through simultaneous preservation of the exact degree sequence and exact global triangle count T = 6960. This does not imply matching of the complete spectrum, lambda_max, QSPEC, RRESP, local triangle structure, four-cycle structure, or higher-order incidence organization. The v0.1.2 correction does not inspect or optimize any of those scientific outcomes. No confirmatory seed stream was used while preparing this correction. No confirmatory target spectrum was computed. No confirmatory QSPEC or RRESP score was computed. No scientific HLV-R-MECH-001 verdict was generated during correction preparation. The corrected v0.1.2 One-Click notebook SHA-256 is: c360d318d97575b56d8fb65327bfd4cb34255325fa2d5509794311d5fa5622cf The corrected implementation-freeze PDF SHA-256 is: fb826a84eee955236c8922559ceeb8e2587854b3c8d9ef5b539a3bb0695469a2 The unchanged scientific engine SHA-256 is: 317df650991120f686768ffc07d12f044f58e38ce8f2c47c083901bf1d7a8a14 The complete corrected implementation package SHA-256 is: 724837c73506bbd001082d9b9b6aec0412304fd467ed55c3705c70d59acd719b This record supersedes corrected implementation freeze v0.1.1 only with respect to the runtime-bootstrap mechanism. It does not alter or supersede the scientific protocol. HLV-R-MECH-001 remains a finite graph-mechanism test within the Helix–Light–Vortex Framework (HLV), positioned as a Cut-and-Project and Incidence-Spectral Research Programme. Neither this corrected implementation nor any later HLV-R-MECH-001 result can by itself establish unique HLV geometry, physical selection of the golden ratio, extra dimensions, spacetime, particle physics, an absolute energy scale, gravity, dark matter, dark energy, cosmology, or experimental validation. The sole purpose of this corrected implementation freeze is to make the already prospectively frozen scientific engine executable in a reproducible numerical environment despite restrictions of the externally managed Colab runtime.
Marcel Krüger· Zenodo (CERN European Organi...· 0 citations
Reach audiences
Advertise in front of researchers, engineers, and readers.
Ramsey’s theorem says that if it is large enough, order must appear. This paper asks how large “large enough” is──the answer is the theorem does not say. No new mathematical theorem and no new law is claimed. Scope of this paper (scope note): No new mathematical theorem and no new law is claimed──Ramsey’s theorem, the known values and ranges of Ramsey numbers, van der Waerden numbers, and Erdos’s probabilistic lower bound are all standard. We do not build combinatorics──all we use is one binomial coefficient and two logarithms. We do not prove Ramsey’s theorem──we merely quote it. We do not compute Ramsey numbers──no attempt is made to find R(5,5). We count only the effort it would take. We assert no values for the ranges──R(5,5) in [43,48] and R(6,6) in [102,160] are ranges known at one epoch and may be improved. The claim is that the range is not empty, not the endpoints. We do not say exhaustion is the only route──actual searches cut enormously by symmetry and pruning. 2^903 is a naive upper bound, not the necessary work. The point is the order of magnitude by which it remains out of reach. We do not use “Ackermann type” strictly──it indicates that the pre-Gowers bound was of a tower-of-growing-height kind. This paper asserts no classification of the bound, only its divergence from the true value. Relation to earlier papers: Paper 163 separated “a good code exists” from “here is a good code”──that is existence against construction; this is existence against quantity. It treats the case where one can construct yet cannot reach, so the cut differs. Paper 179 showed that “cannot be constructed” has distinct roots──this paper treats what can be constructed yet is out of reach. Paper 190 measured “rare” on a logarithmic scale──the 271.8 orders here are read logarithmically too. Paper 259 showed there are two roads to independence──this paper likewise asks what kinds of showing there are. What is added is computing the exhaustion for R(5,5) as 10^271.8 and measuring the gap to the atoms of the universe as 191.8 orders, lining up the jumps at R(3,3), R(4,4), R(5,5), computing that the probabilistic lower bound is one 24.94th at k=10, and placing the separator at existence against quantity. First, only two values are settled. R(3,3)=6 and R(4,4)=18; R(5,5) is pinned only to the range [43,48] (Section 2). Second, this is the core of the paper. Checking R(5,5)=43 by exhaustion needs 2^903=10^271.8 colourings, 191.8 orders more than the atoms in the universe (Section 3). Third, each step up jumps. R(3,3) is 32768 colourings, within reach by hand; R(4,4) is 1.14x10^46, which computers barely reached (Section 3). Fourth, the lower-bound proof does not give the value either. Erdos’s probabilistic method gives R(k,k)>2^k/2, which at k=10 is one 24.94th of the truth (Section 4). Fifth, the upper bound is further off still. W(3,3)=27, yet the bound from the classical proof was of Ackermann type (Section 5). Sixth, the separator is existence against quantity. A different cut from Paper 163’s existence against construction (Section 6). Ramsey’s theorem says that if it is large enough, order must appear. But it does not say how large. Only R(3,3)=6 and R(4,4)=18 are settled, and R(5,5) is pinned only to the range [43,48]──even settling R(4,4) took 65 years from the theorem. Counting the work of exhaustion shows why──R(5,5)=43 has 903 edges and 2^903=10^271.8 colourings, 191.8 orders more than the atoms in the observable universe. What is lacking is not the speed of computers but the quantity of matter. The lower bound is no better──the probabilistic method gives R(k,k)>2^k/2, one 24.94th of the truth at k=10, and the ratio widens with k. The upper bound is further off still──W(3,3) is truly 27, yet the classical bound could not be written down. A theorem’s upper bound measures the power of the proof’s tools, not the size of the object. One thing separates them──whether it exists, how large it is, and which one it is, are three different questions. The first may be “yes” while the other two stay open. The single word “proved” points at only one of the three. On the making of this work: The ideas and content of this work stem from the author's own considerations. Assistance from an AI (a large language model) was used for structuring, English translation, and checking the algebra. Any remaining errors or misinterpretations are solely the author's. Feedback and corrections are sincerely appreciated. ----- ラムゼーの定理は「十分大きければ必ず秩序が現れる」と言う。本稿が問うのは、「十分大きい」がどれだけかである──答は、定理は教えないである。新しい数学定理も新しい法則も主張しない。 本稿の射程(射程注記):新しい数学定理も新しい法則も主張しない──ラムゼーの定理、ラムゼー数の既知の値と範囲、ファン・デル・ヴェルデン数、エルデシュの確率論的下限は、いずれも標準的である。組合せ論を作らない──使うのは一つの二項係数と、二つの対数だけである。ラムゼーの定理を証明しない──引くだけである。ラムゼー数を計算しない──R(5,5) の値を求めようとはしていない。求めるのに要る手間だけを数える。範囲の値を主張しない──R(5,5) in [43,48]、R(6,6) in [102,160] はある時点で知られている範囲であり、改善されうる。本稿の主張は範囲が空でないことであって、端の値ではない。総当たりが唯一の道だと言わない──実際の探索は対称性と枝刈りで大幅に減らす。2^903 は素朴な上界であって、必要な計算量ではない。それでも到底届かない、という桁の話である。アッカーマン級という語を厳密に使わない──ガワーズ以前の上界が塔の高さが増える型だったことを指す。本稿は上界の形を主張せず、真の値との乖離だけを言う。既刊との関係:論文163 は「良い符号が在る」と「これが良い符号だ」を分けた──あちらは存在と構成、本稿は存在と定量である。構成できても大きさが分からない場合を扱うので、切り口が違う。論文179 は「構成できない」に別根があると示した──本稿は構成できるのに手が届かない場合である。論文190 は「稀」を対数の目盛りで測った──本稿の 271.8 桁も対数で読む。論文259 は独立性を示す道が二つあると示した──本稿も示し方の種類を問う。加えたのはR(5,5) の総当たりを 10^271.8 通りと計算し、宇宙の原子との差を 191.8 桁と測ったこと、R(3,3)・R(4,4)・R(5,5) で手間が跳ぶ段を並べたこと、確率論的下限が k=10 で 24.94 分の一だと計算したこと、分離子を「存在と定量」に置いたことである。 第一に、確定している値は二つだけである。 R(3,3)=6 と R(4,4)=18 で、R(5,5) は[43,48] の範囲にしか収まっていない(第2節)。 第二に、これが本稿の芯である。 R(5,5)=43 を総当たりで確かめるには 2^903=10^271.8 通りが要り、宇宙の原子より 191.8 桁多い(第3節)。 第三に、一段上がるごとに跳ぶ。 R(3,3) は 32768 通りで手が届き、R(4,4) は 1.14x10^46 通りで計算機がようやく届いた(第3節)。 第四に、下限の証明も値を教えない。エルデシュの確率論的方法は R(k,k)>2^k/2 を与えるが、k=10 で実際の 24.94 分の一である(第4節)。 第五に、上界はもっと外れる。 W(3,3)=27 なのに、古典的証明が与えた上界はアッカーマン級だった(第5節)。 第六に、分離子は「存在と定量」である。論文163 の「存在と構成」とは別の切り口である(第6節)。 ラムゼーの定理は「十分大きければ必ず秩序が現れる」と言う。だが「十分大きい」を教えない。確定している値は R(3,3)=6 と R(4,4)=18 の二つだけで、R(5,5) は[43,48] の範囲にしか収まっていない──R(4,4) の確定でさえ、定理から 65 年かかった。総当たりの手間を数えれば理由が分かる──R(5,5)=43 は辺が 903 本、塗り分けが 2^903=10^271.8 通りで、観測可能な宇宙の原子より 191.8 桁多い。足りないのは計算機の速さではなく、物質の量である。下限の側も同じである──確率論的方法は R(k,k)>2^k/2 を与えるが、k=10 で実際の 24.94 分の一であり、比は k とともに開いていく。上界はもっと外れる──W(3,3) の真の値は 27 なのに、古典的証明の上界は書き下せない大きさだった。定理の上界は、証明の道具の性能を測っているのであって、対象の大きさを測っていない。分けるものは一つ──存在するかと、いくつかと、どれかは、三つの別の問いである。一つ目が「はい」でも、残り二つは開いたままでありうる。「証明された」という一語が、三つのうち一つしか指していない。 作成にあたって:本稿の着想と内容は、著者自身の考察に基づくものです。文章の構成整理や英訳、数式の確認には AI(大規模言語モデル)の助力を得ました。最終的な内容の解釈や誤りがあれば、それらはすべて著者の責に帰します。お気づきの点があれば、ご教示いただければ幸いです。
Yuuki Yamagishi· Zenodo (CERN European Organi...· 0 citations
The doubling strategy──stake twice as much after every loss──can push the win rate arbitrarily close to 1. This paper asks whether the expectation moves with it──the answer is it stays exactly 0. No new mathematical theorem and no new law is claimed. Scope of this paper (scope note): No new mathematical theorem and no new law is claimed──the doubling strategy, martingales, the optional stopping theorem and its three sufficient conditions are all standard. We do not build probability theory──all we use is one geometric series and one expectation. We do not prove the optional stopping theorem──the sufficient conditions are merely quoted, and we do not enter the proof. We do not recommend gambling──this paper only counts how the expectation behaves. As Section 4 shows, in a game with a house edge the expectation is negative. We do not discuss optimal strategy──we do not enter the Kelly criterion, utility functions, or minimising the probability of ruin. We do not discuss real casinos──table limits and the practicalities of capital constraints are not treated. The bankroll 2^k-1 is a mathematical setting. We do not call variance evil──the factor of 1024 in Section 3 is a statement of fact, and we do not claim that large variance is always bad. We do not treat the tails──Paper 190 treated heavy tails and large deviations. This paper does not enter the shape of the distribution and counts only expectation and variance. Relation to earlier papers: Paper 143 showed that minimum entropy production is not a principle but a theorem with a fence──the optional stopping theorem here is likewise a theorem with a fence, of the same shape: remove the fence and the conclusion collapses. Paper 194 showed that the four means are one family with only one fence of equality──here too there is only one fence. Paper 220 counted four things called “independence”──the successive bets here are independent in one of those senses. Paper 190 measured “rare” on a logarithmic scale──the 2^-20 here is likewise read logarithmically. What is added is computing that the expectation is exactly 0 for every k, showing that the product of loss and probability is 1-2^-k and balances the gain exactly, measuring the standard deviation as 1024 times larger, computing that a house edge makes the expected loss 26.07 times larger, and naming the fence as capital rather than probability. First, the win rate can be raised without limit. A bankroll of 2^k-1 survives k consecutive losses, and at k=20 the win rate is 99.9999046% (Section 2). Second, this is the core of the paper. The expectation is nevertheless exactly 0 for every k──1-2^-k-1+2^-k=0 (Section 2). Third, the product of loss and its probability exactly balances the gain.2^-kx(2^k-1)=1-2^-k, which tends to 1 (Section 3). Fourth, what moved is the variance. At k=20 the standard deviation becomes 1024 times larger while the expectation stays 0 (Section 3). Fifth, with a house edge it gets worse. To win the same 1 unit, the expected loss becomes 26.07 times larger (Section 4). Sixth, the separator is capital, not probability. Make the capital unbounded and all three premises of the optional stopping theorem fail (Section 5). A strategy that wins 99.9999% of the time has expectation exactly 0. A bankroll of 2^k-1 survives k consecutive losses, and at k=20 the win rate is 99.9999046%──yet (1-2^-k)-2^-k(2^k-1)=0 holds identically. What balances is the product──lowering the losing probability to 2^-k and raising the loss to 2^k-1 are one and the same operation, and one cannot move only one of them. What moved was the variance alone──doubling multiplied the standard deviation by 1024 and left the expectation at 0. With a house edge it is worse still──chasing the same 1 unit makes the expected loss 26.07 times larger. One thing separates them──not probability, but capital. With unbounded capital all three premises of the optional stopping theorem fail, and a fair game yields a certain profit. The fence stands on the side of capital, not of probability. So the figure “a win rate of 99.9999%” guarantees nothing on its own. On the making of this work: The ideas and content of this work stem from the author's own considerations. Assistance from an AI (a large language model) was used for structuring, English translation, and checking the algebra. Any remaining errors or misinterpretations are solely the author's. Feedback and corrections are sincerely appreciated. ----- 負けたら賭金を倍にする「倍賭け」は、勝率をいくらでも 1 に近づけられる。本稿が問うのは、それで期待値が動くかである──答は、ちょうど 0 のままである。新しい数学定理も新しい法則も主張しない。 本稿の射程(射程注記):新しい数学定理も新しい法則も主張しない──倍賭け戦略、マルチンゲール、任意停止定理、その三つの十分条件は、いずれも標準的である。確率論を作らない──使うのは一つの等比級数と、一つの期待値の計算だけである。任意停止定理を証明しない──十分条件を引くだけであり、証明には立ち入らない。賭博を勧めない──本稿は期待値がどう振る舞うかだけを数える。第4節が示すとおり、胴元の取り分がある賭けでは期待値は負である。最適戦略を論じない──ケリー基準にも、効用関数にも、破産確率の最小化にも立ち入らない。現実の賭場を論じない──賭金上限も、資金制約の実際も扱わない。資金 2^k-1 は数学的な設定である。分散を悪だと言わない──第3節の 1024 倍は事実の記述であって、分散が大きいことが常に悪いとは主張しない。裾の分布を扱わない──論文190 が重い裾と大偏差を扱った。本稿は分布の形に入らず、期待値と分散だけを数える。既刊との関係:論文143 は最小エントロピー生成が原理ではなく柵つきの定理だと示した──本稿の任意停止定理も柵つきの定理であり、柵が外れると結論が崩れるという同じ形である。論文194 は四つの平均が一つの族であり、等号の柵が一つしかないと示した──本稿も柵が一つだけである。論文220 は「独立」が四つあると数えた──本稿の各回の賭けはそのうち一つの意味で独立である。論文190 は「稀」を対数の目盛りで測った──本稿の 2^-20 も対数で読む。加えたのは期待値が k によらず厳密に 0 になると計算したこと、負ける確率と損失の積が 1-2^-k で利得とちょうど釣り合うと示したこと、標準偏差が 1024 倍になると測ったこと、胴元の取り分があると期待損失が 26.07 倍になると計算したこと、柵を「確率ではなく資金」と名指したことである。 第一に、勝率はいくらでも上げられる。資金 2^k-1 で k 連敗まで耐えられ、k=20 なら勝率は 99.9999046% である(第2節)。 第二に、これが本稿の芯である。それでも期待値は k によらずちょうど 0──1-2^-k-1+2^-k=0 である(第2節)。 第三に、負ける確率と損失の積が、利得とちょうど釣り合う。2^-kx(2^k-1)=1-2^-k であり、k->infinity で 1 に近づく(第3節)。 第四に、動いたのは分散である。 k=20 で標準偏差が 1024 倍になり、期待値は 0 のままである(第3節)。 第五に、胴元の取り分があると悪化する。同じ 1 単位を勝つのに、期待損失が 26.07 倍になる(第4節)。 第六に、分離子は確率ではなく資金である。資金を無限にすると任意停止定理の前提が三つとも破れる(第5節)。 勝率 99.9999% の戦略の期待値は、ちょうど 0 である。資金 2^k-1 で k 連敗まで耐えられ、k=20 なら勝率は 99.9999046%──それでも (1-2^-k)-2^-k(2^k-1)=0 が恒等的に成り立つ。釣り合っているのは積である──負ける確率を 2^-k に下げる操作と、損失を 2^k-1 に上げる操作が、同じ一つの操作であり、片方だけを動かすことはできない。動いたのは分散だけである──倍賭けは標準偏差を 1024 倍にし、期待値は 0 のまま置いた。胴元の取り分があれば、さらに悪い──同じ 1 単位を狙うのに、期待損失が 26.07 倍になる。分けるものは一つ──確率ではなく、資金。資金が無限なら任意停止定理の前提が三つとも破れ、公平な賭けから確実な利益が出てしまう。柵は確率の側ではなく、資金の側に立っている。だから「勝率 99.9999%」という数字は、それだけでは何も保証しない。 作成にあたって:本稿の着想と内容は、著者自身の考察に基づくものです。文章の構成整理や英訳、数式の確認には AI(大規模言語モデル)の助力を得ました。最終的な内容の解釈や誤りがあれば、それらはすべて著者の責に帰します。お気づきの点があれば、ご教示いただければ幸いです。
Yuuki Yamagishi· Zenodo (CERN European Organi...· 0 citations
Abstract Communication barriers experienced by deaf and hard-of-hearing individuals remain a significant challenge, particularly in situations where conventional communication infrastructure is unavailable or fixed-camera systems have limited coverage. This paper presents SIGN-AIR, a UAV-assisted framework for real-time sign language recognition and bidirectional translation. Unlike conventional approaches based on fixed cameras, wearable sensors, or depth cameras, the proposed framework integrates a drone-mounted stabilised RGB camera with onboard edge computing to enable flexible gesture recognition in indoor and outdoor environments. The system combines MediaPipe-based hand landmark extraction with a hybrid CNN–BiLSTM architecture to model both the spatial and temporal characteristics of American Sign Language gestures. Experimental evaluation on a combined dataset comprising the ASL Alphabet and WLASL achieved an overall recognition accuracy of 98.63%, together with high precision, high recall, and a median inference latency of less than 2 s. These results demonstrate the technical feasibility of the proposed UAV-assisted recognition framework under the evaluated experimental conditions and indicate its potential to support communication in scenarios where fixed-camera systems may be affected by occlusions, limited fields of view, or environmental variability. Furthermore, the proposed bidirectional translation pipeline has the potential to facilitate both sign-to-speech and speech-to-sign communication for prospective applications in accessibility, education, public service interactions, and emergency response. Although the current study demonstrates promising technical performance, comprehensive real-world field validation under representative operational conditions remains an important direction for future work.
Hafsa Waheed, Ghulam E Mustafa Abro, Sufyan Ali Memon et al.· Scientific Reports· 0 citations
Ramsey’s theorem says that if it is large enough, order must appear. This paper asks how large “large enough” is──the answer is the theorem does not say. No new mathematical theorem and no new law is claimed. Scope of this paper (scope note): No new mathematical theorem and no new law is claimed──Ramsey’s theorem, the known values and ranges of Ramsey numbers, van der Waerden numbers, and Erdos’s probabilistic lower bound are all standard. We do not build combinatorics──all we use is one binomial coefficient and two logarithms. We do not prove Ramsey’s theorem──we merely quote it. We do not compute Ramsey numbers──no attempt is made to find R(5,5). We count only the effort it would take. We assert no values for the ranges──R(5,5) in [43,48] and R(6,6) in [102,160] are ranges known at one epoch and may be improved. The claim is that the range is not empty, not the endpoints. We do not say exhaustion is the only route──actual searches cut enormously by symmetry and pruning. 2^903 is a naive upper bound, not the necessary work. The point is the order of magnitude by which it remains out of reach. We do not use “Ackermann type” strictly──it indicates that the pre-Gowers bound was of a tower-of-growing-height kind. This paper asserts no classification of the bound, only its divergence from the true value. Relation to earlier papers: Paper 163 separated “a good code exists” from “here is a good code”──that is existence against construction; this is existence against quantity. It treats the case where one can construct yet cannot reach, so the cut differs. Paper 179 showed that “cannot be constructed” has distinct roots──this paper treats what can be constructed yet is out of reach. Paper 190 measured “rare” on a logarithmic scale──the 271.8 orders here are read logarithmically too. Paper 259 showed there are two roads to independence──this paper likewise asks what kinds of showing there are. What is added is computing the exhaustion for R(5,5) as 10^271.8 and measuring the gap to the atoms of the universe as 191.8 orders, lining up the jumps at R(3,3), R(4,4), R(5,5), computing that the probabilistic lower bound is one 24.94th at k=10, and placing the separator at existence against quantity. First, only two values are settled. R(3,3)=6 and R(4,4)=18; R(5,5) is pinned only to the range [43,48] (Section 2). Second, this is the core of the paper. Checking R(5,5)=43 by exhaustion needs 2^903=10^271.8 colourings, 191.8 orders more than the atoms in the universe (Section 3). Third, each step up jumps. R(3,3) is 32768 colourings, within reach by hand; R(4,4) is 1.14x10^46, which computers barely reached (Section 3). Fourth, the lower-bound proof does not give the value either. Erdos’s probabilistic method gives R(k,k)>2^k/2, which at k=10 is one 24.94th of the truth (Section 4). Fifth, the upper bound is further off still. W(3,3)=27, yet the bound from the classical proof was of Ackermann type (Section 5). Sixth, the separator is existence against quantity. A different cut from Paper 163’s existence against construction (Section 6). Ramsey’s theorem says that if it is large enough, order must appear. But it does not say how large. Only R(3,3)=6 and R(4,4)=18 are settled, and R(5,5) is pinned only to the range [43,48]──even settling R(4,4) took 65 years from the theorem. Counting the work of exhaustion shows why──R(5,5)=43 has 903 edges and 2^903=10^271.8 colourings, 191.8 orders more than the atoms in the observable universe. What is lacking is not the speed of computers but the quantity of matter. The lower bound is no better──the probabilistic method gives R(k,k)>2^k/2, one 24.94th of the truth at k=10, and the ratio widens with k. The upper bound is further off still──W(3,3) is truly 27, yet the classical bound could not be written down. A theorem’s upper bound measures the power of the proof’s tools, not the size of the object. One thing separates them──whether it exists, how large it is, and which one it is, are three different questions. The first may be “yes” while the other two stay open. The single word “proved” points at only one of the three. On the making of this work: The ideas and content of this work stem from the author's own considerations. Assistance from an AI (a large language model) was used for structuring, English translation, and checking the algebra. Any remaining errors or misinterpretations are solely the author's. Feedback and corrections are sincerely appreciated. ----- ラムゼーの定理は「十分大きければ必ず秩序が現れる」と言う。本稿が問うのは、「十分大きい」がどれだけかである──答は、定理は教えないである。新しい数学定理も新しい法則も主張しない。 本稿の射程(射程注記):新しい数学定理も新しい法則も主張しない──ラムゼーの定理、ラムゼー数の既知の値と範囲、ファン・デル・ヴェルデン数、エルデシュの確率論的下限は、いずれも標準的である。組合せ論を作らない──使うのは一つの二項係数と、二つの対数だけである。ラムゼーの定理を証明しない──引くだけである。ラムゼー数を計算しない──R(5,5) の値を求めようとはしていない。求めるのに要る手間だけを数える。範囲の値を主張しない──R(5,5) in [43,48]、R(6,6) in [102,160] はある時点で知られている範囲であり、改善されうる。本稿の主張は範囲が空でないことであって、端の値ではない。総当たりが唯一の道だと言わない──実際の探索は対称性と枝刈りで大幅に減らす。2^903 は素朴な上界であって、必要な計算量ではない。それでも到底届かない、という桁の話である。アッカーマン級という語を厳密に使わない──ガワーズ以前の上界が塔の高さが増える型だったことを指す。本稿は上界の形を主張せず、真の値との乖離だけを言う。既刊との関係:論文163 は「良い符号が在る」と「これが良い符号だ」を分けた──あちらは存在と構成、本稿は存在と定量である。構成できても大きさが分からない場合を扱うので、切り口が違う。論文179 は「構成できない」に別根があると示した──本稿は構成できるのに手が届かない場合である。論文190 は「稀」を対数の目盛りで測った──本稿の 271.8 桁も対数で読む。論文259 は独立性を示す道が二つあると示した──本稿も示し方の種類を問う。加えたのはR(5,5) の総当たりを 10^271.8 通りと計算し、宇宙の原子との差を 191.8 桁と測ったこと、R(3,3)・R(4,4)・R(5,5) で手間が跳ぶ段を並べたこと、確率論的下限が k=10 で 24.94 分の一だと計算したこと、分離子を「存在と定量」に置いたことである。 第一に、確定している値は二つだけである。 R(3,3)=6 と R(4,4)=18 で、R(5,5) は[43,48] の範囲にしか収まっていない(第2節)。 第二に、これが本稿の芯である。 R(5,5)=43 を総当たりで確かめるには 2^903=10^271.8 通りが要り、宇宙の原子より 191.8 桁多い(第3節)。 第三に、一段上がるごとに跳ぶ。 R(3,3) は 32768 通りで手が届き、R(4,4) は 1.14x10^46 通りで計算機がようやく届いた(第3節)。 第四に、下限の証明も値を教えない。エルデシュの確率論的方法は R(k,k)>2^k/2 を与えるが、k=10 で実際の 24.94 分の一である(第4節)。 第五に、上界はもっと外れる。 W(3,3)=27 なのに、古典的証明が与えた上界はアッカーマン級だった(第5節)。 第六に、分離子は「存在と定量」である。論文163 の「存在と構成」とは別の切り口である(第6節)。 ラムゼーの定理は「十分大きければ必ず秩序が現れる」と言う。だが「十分大きい」を教えない。確定している値は R(3,3)=6 と R(4,4)=18 の二つだけで、R(5,5) は[43,48] の範囲にしか収まっていない──R(4,4) の確定でさえ、定理から 65 年かかった。総当たりの手間を数えれば理由が分かる──R(5,5)=43 は辺が 903 本、塗り分けが 2^903=10^271.8 通りで、観測可能な宇宙の原子より 191.8 桁多い。足りないのは計算機の速さではなく、物質の量である。下限の側も同じである──確率論的方法は R(k,k)>2^k/2 を与えるが、k=10 で実際の 24.94 分の一であり、比は k とともに開いていく。上界はもっと外れる──W(3,3) の真の値は 27 なのに、古典的証明の上界は書き下せない大きさだった。定理の上界は、証明の道具の性能を測っているのであって、対象の大きさを測っていない。分けるものは一つ──存在するかと、いくつかと、どれかは、三つの別の問いである。一つ目が「はい」でも、残り二つは開いたままでありうる。「証明された」という一語が、三つのうち一つしか指していない。 作成にあたって:本稿の着想と内容は、著者自身の考察に基づくものです。文章の構成整理や英訳、数式の確認には AI(大規模言語モデル)の助力を得ました。最終的な内容の解釈や誤りがあれば、それらはすべて著者の責に帰します。お気づきの点があれば、ご教示いただければ幸いです。
Yuuki Yamagishi· Zenodo (CERN European Organi...· 0 citations
The doubling strategy──stake twice as much after every loss──can push the win rate arbitrarily close to 1. This paper asks whether the expectation moves with it──the answer is it stays exactly 0. No new mathematical theorem and no new law is claimed. Scope of this paper (scope note): No new mathematical theorem and no new law is claimed──the doubling strategy, martingales, the optional stopping theorem and its three sufficient conditions are all standard. We do not build probability theory──all we use is one geometric series and one expectation. We do not prove the optional stopping theorem──the sufficient conditions are merely quoted, and we do not enter the proof. We do not recommend gambling──this paper only counts how the expectation behaves. As Section 4 shows, in a game with a house edge the expectation is negative. We do not discuss optimal strategy──we do not enter the Kelly criterion, utility functions, or minimising the probability of ruin. We do not discuss real casinos──table limits and the practicalities of capital constraints are not treated. The bankroll 2^k-1 is a mathematical setting. We do not call variance evil──the factor of 1024 in Section 3 is a statement of fact, and we do not claim that large variance is always bad. We do not treat the tails──Paper 190 treated heavy tails and large deviations. This paper does not enter the shape of the distribution and counts only expectation and variance. Relation to earlier papers: Paper 143 showed that minimum entropy production is not a principle but a theorem with a fence──the optional stopping theorem here is likewise a theorem with a fence, of the same shape: remove the fence and the conclusion collapses. Paper 194 showed that the four means are one family with only one fence of equality──here too there is only one fence. Paper 220 counted four things called “independence”──the successive bets here are independent in one of those senses. Paper 190 measured “rare” on a logarithmic scale──the 2^-20 here is likewise read logarithmically. What is added is computing that the expectation is exactly 0 for every k, showing that the product of loss and probability is 1-2^-k and balances the gain exactly, measuring the standard deviation as 1024 times larger, computing that a house edge makes the expected loss 26.07 times larger, and naming the fence as capital rather than probability. First, the win rate can be raised without limit. A bankroll of 2^k-1 survives k consecutive losses, and at k=20 the win rate is 99.9999046% (Section 2). Second, this is the core of the paper. The expectation is nevertheless exactly 0 for every k──1-2^-k-1+2^-k=0 (Section 2). Third, the product of loss and its probability exactly balances the gain.2^-kx(2^k-1)=1-2^-k, which tends to 1 (Section 3). Fourth, what moved is the variance. At k=20 the standard deviation becomes 1024 times larger while the expectation stays 0 (Section 3). Fifth, with a house edge it gets worse. To win the same 1 unit, the expected loss becomes 26.07 times larger (Section 4). Sixth, the separator is capital, not probability. Make the capital unbounded and all three premises of the optional stopping theorem fail (Section 5). A strategy that wins 99.9999% of the time has expectation exactly 0. A bankroll of 2^k-1 survives k consecutive losses, and at k=20 the win rate is 99.9999046%──yet (1-2^-k)-2^-k(2^k-1)=0 holds identically. What balances is the product──lowering the losing probability to 2^-k and raising the loss to 2^k-1 are one and the same operation, and one cannot move only one of them. What moved was the variance alone──doubling multiplied the standard deviation by 1024 and left the expectation at 0. With a house edge it is worse still──chasing the same 1 unit makes the expected loss 26.07 times larger. One thing separates them──not probability, but capital. With unbounded capital all three premises of the optional stopping theorem fail, and a fair game yields a certain profit. The fence stands on the side of capital, not of probability. So the figure “a win rate of 99.9999%” guarantees nothing on its own. On the making of this work: The ideas and content of this work stem from the author's own considerations. Assistance from an AI (a large language model) was used for structuring, English translation, and checking the algebra. Any remaining errors or misinterpretations are solely the author's. Feedback and corrections are sincerely appreciated. ----- 負けたら賭金を倍にする「倍賭け」は、勝率をいくらでも 1 に近づけられる。本稿が問うのは、それで期待値が動くかである──答は、ちょうど 0 のままである。新しい数学定理も新しい法則も主張しない。 本稿の射程(射程注記):新しい数学定理も新しい法則も主張しない──倍賭け戦略、マルチンゲール、任意停止定理、その三つの十分条件は、いずれも標準的である。確率論を作らない──使うのは一つの等比級数と、一つの期待値の計算だけである。任意停止定理を証明しない──十分条件を引くだけであり、証明には立ち入らない。賭博を勧めない──本稿は期待値がどう振る舞うかだけを数える。第4節が示すとおり、胴元の取り分がある賭けでは期待値は負である。最適戦略を論じない──ケリー基準にも、効用関数にも、破産確率の最小化にも立ち入らない。現実の賭場を論じない──賭金上限も、資金制約の実際も扱わない。資金 2^k-1 は数学的な設定である。分散を悪だと言わない──第3節の 1024 倍は事実の記述であって、分散が大きいことが常に悪いとは主張しない。裾の分布を扱わない──論文190 が重い裾と大偏差を扱った。本稿は分布の形に入らず、期待値と分散だけを数える。既刊との関係:論文143 は最小エントロピー生成が原理ではなく柵つきの定理だと示した──本稿の任意停止定理も柵つきの定理であり、柵が外れると結論が崩れるという同じ形である。論文194 は四つの平均が一つの族であり、等号の柵が一つしかないと示した──本稿も柵が一つだけである。論文220 は「独立」が四つあると数えた──本稿の各回の賭けはそのうち一つの意味で独立である。論文190 は「稀」を対数の目盛りで測った──本稿の 2^-20 も対数で読む。加えたのは期待値が k によらず厳密に 0 になると計算したこと、負ける確率と損失の積が 1-2^-k で利得とちょうど釣り合うと示したこと、標準偏差が 1024 倍になると測ったこと、胴元の取り分があると期待損失が 26.07 倍になると計算したこと、柵を「確率ではなく資金」と名指したことである。 第一に、勝率はいくらでも上げられる。資金 2^k-1 で k 連敗まで耐えられ、k=20 なら勝率は 99.9999046% である(第2節)。 第二に、これが本稿の芯である。それでも期待値は k によらずちょうど 0──1-2^-k-1+2^-k=0 である(第2節)。 第三に、負ける確率と損失の積が、利得とちょうど釣り合う。2^-kx(2^k-1)=1-2^-k であり、k->infinity で 1 に近づく(第3節)。 第四に、動いたのは分散である。 k=20 で標準偏差が 1024 倍になり、期待値は 0 のままである(第3節)。 第五に、胴元の取り分があると悪化する。同じ 1 単位を勝つのに、期待損失が 26.07 倍になる(第4節)。 第六に、分離子は確率ではなく資金である。資金を無限にすると任意停止定理の前提が三つとも破れる(第5節)。 勝率 99.9999% の戦略の期待値は、ちょうど 0 である。資金 2^k-1 で k 連敗まで耐えられ、k=20 なら勝率は 99.9999046%──それでも (1-2^-k)-2^-k(2^k-1)=0 が恒等的に成り立つ。釣り合っているのは積である──負ける確率を 2^-k に下げる操作と、損失を 2^k-1 に上げる操作が、同じ一つの操作であり、片方だけを動かすことはできない。動いたのは分散だけである──倍賭けは標準偏差を 1024 倍にし、期待値は 0 のまま置いた。胴元の取り分があれば、さらに悪い──同じ 1 単位を狙うのに、期待損失が 26.07 倍になる。分けるものは一つ──確率ではなく、資金。資金が無限なら任意停止定理の前提が三つとも破れ、公平な賭けから確実な利益が出てしまう。柵は確率の側ではなく、資金の側に立っている。だから「勝率 99.9999%」という数字は、それだけでは何も保証しない。 作成にあたって:本稿の着想と内容は、著者自身の考察に基づくものです。文章の構成整理や英訳、数式の確認には AI(大規模言語モデル)の助力を得ました。最終的な内容の解釈や誤りがあれば、それらはすべて著者の責に帰します。お気づきの点があれば、ご教示いただければ幸いです。
Yuuki Yamagishi· Zenodo (CERN European Organi...· 0 citations
A companion paper (doi:10.5281/zenodo.22168191, "What must a theory of perturbational complexity explain?") consolidated nine constraints and challenged any theory to pass them — including C3: debiased perturbational complexity (R-dim, the reproducible dimensionality of the evoked response) is an inverted-U in network dynamics, maximal near the edge of chaos. Here we answer part of that challenge with eleven sealed protocols (all preregistered publicly before their blind quantities were computed, each run once; the failed protocols are reported with the same prominence as the successes). (1) An exact linear theory. For linear networks, R-dim is computable in closed form from connectivity alone: component amplitudes from the impulse-response SVD, per-component noise floors from the stationary Lyapunov covariance, and a universal averaging formula. Confirmed on 36 virgin networks: rank correlation +0.966, median absolute error 0.33 dimensions with a single frozen constant, per-network growth-with-trials predicted (ρ = +0.59), and no inverted-U anywhere in the stable linear family — the falling branch is not a linear phenomenon. (2) The formula transports to nonlinear networks. Feeding it the measured coherent spectrum (through an estimator validated on a known-truth bench; all bench iterations documented) and the measured trajectory noise reproduces the full U on 48 virgin networks in order (+0.925) and calibration (median error 0.41 dimensions, constant frozen), while both sealed peak-location criteria failed: the U's top is a flat plateau at this resolution, and a dedicated sealed 128-network estimation returned "indeterminate". (3) Time and richness dissociate. The duration of the coherent response obeys a relaxation law on the stable side (T_c ~ 1/|λ|; +0.943 on virgin networks) and is maximal at the edge (37× deep-stable, 6× deep-chaotic) — yet the *maximum of R-dim does not coincide with the maximum of coherent time* (sealed dissociation, P = 0.99): richness needs more than time. (4) What governs the fall is saturation. In a two-axis design sampling networks by measured λ and spectral radius ρ independently, at matched criticality saturation crushes R-dim (Spearman −0.62 to −0.88 across all λ strata; 7/7 sealed criteria on 156 virgin networks), while at matched saturation criticality's effect is bounded (|ρₛ| < 0.19). The falling branch of the U is largely saturation wearing chaos's clothes. A finer "graded law" across three additional activation families died in its own sealed test (0/3); the burial is reported, with the method lesson it taught — and a properly powered family-level replication then confirmed the coarse tendency in all three families (sealed; median per-stratum Spearman −0.50 / −0.61 / −0.41 for erf, hard-clip, and softsign at 30 networks per stratum): saturation hurts reproducible dimensionality whatever the shape of the ceiling. A final sealed protocol then answered *where* saturation takes over: in a parametric family interpolating softsign to hard clip, the saturation-governed regime advances with the tail speed of the activation (stable-strata effect −0.16 at p = 1 vs −0.70 at p = 8; sealed two-point criterion passed at −0.54 against a −0.30 bar). (5) One formula. A final sealed protocol tested the unification all of the above points to: applying the exact linear theory to the gain-shrunk effective system W_eff = diag(⟨φ′⟩)·W — same frozen constant — predicts R-dim across all seven activation families studied (pooled order +0.653 on 280 virgin networks, p = 2 × 10⁻³⁵), and per-unit gain heterogeneity is sealed as the essential carrier (the scalar-gain comparator loses by 0.435). At the order level, a nonlinear network is, for reproducible dimensionality, its gain-shrunk linear self; absolute calibration remains an open refinement. (6) Scope, attacked. A final sealed protocol re-ran both flagship laws in five worlds never touched: sizes ×2 and ×4, biological E/I wiring, and halved/doubled noise. The saturation law survived all five (stratum-median Spearman −0.48 to −0.63): together with the activation-family campaign, it holds across nonlinearity, size, wiring and noise — a law of systems. The effective-gain theory passed fully in two worlds and drew its boundary in the others, degrading with network size (order +0.29 at N = 256): a small-family approximation with a mapped edge. We close with what remains open: the exact shape of the U's top, the quantitative chaotic decay, the analytic form of the tail-crossover law, the level calibration of the effective theory — and why its order degrades with size.
Nicolás Federico Galindez· Zenodo (CERN European Organi...· 0 citations
Archived source release of BioMCP-TS v0.8.0 (npm package biomcp): 41 core tools across 15 registration modules federating 33 verified upstream data hosts, plus optional plugin sets for read-only SQL analytics and in-process WebAssembly compute (Bioconductor DESeq2/edgeR/limma under webR; samtools/bedtools/bcftools under biowasm). Zero-config stdio MCP server for Node >= 22.13, single runtime dependency. Repository: https://github.com/yeyuan98/biomcp-ts (live development continues there; this record archives the exact tag evaluated in the Technical Report).
Ye Yuan· Zenodo (CERN European Organi...· 0 citations
What if pathology foundation models could do more with less? GigaPath-Flash and GigaTIME-Flash cut computational demands while maintaining strong performance, opening the door to larger studies and broader exploration. The post GigaPath-Flash and GigaTIME-Flash: Toward population-scale discovery with efficient pathology foundation models appeared first on Microsoft Research.
MIT News · Artificial Intelligence· news.mit.eduAug 31, 2026
With millions of users across the world, Julia has been used to conduct cutting-edge research and to design new drugs, jet engines, heat pumps, and more.
MIT News · Artificial Intelligence· news.mit.eduAug 27, 2026
A new machine-learning framework aims to improve the success rate of computational protein design while moving away from results that reproduce sequences found in nature.
