Many efforts of research are devoted to semantic role labeling (SRL) which is crucial for natural language understanding. Supervised approaches have achieved impressing performances when large-scale corpora are available for resource-rich languages such as English. While for the low-resource languages with no annotated SRL dataset, it is still challenging to obtain competitive performances. Cross-lingual SRL is one promising way to address the problem, which has achieved great advances with the help of model transferring and annotation projection. In this paper, we propose a novel alternative based on corpus translation, constructing high-quality training datasets for the target languages from the source gold-standard SRL annotations. Experimental results on Universal Proposition Bank show that the translation-based method is highly effective, and the automatic pseudo datasets can improve the target-language SRL performances significantly.

Histopathology is a reflection of the molecular changes and provides prognostic phenotypes representing the disease progression. In this study, we introduced feature scores generated from hematoxylin and eosin histology images based on deep learning (DL) models developed for prostate pathology. We demonstrated that these feature scores were significantly prognostic for time to event endpoints (biochemical recurrence and cancer-specific survival) and had simultaneously molecular biologic associations to relevant genomic alterations and molecular subtypes using already trained DL models that were not previously exposed to the datasets of the current study. Further, we discussed the potential of such feature scores to improve the current tumor grading system and the challenges that are associated with tumor heterogeneity and the development of prognostic models from histology images. Our findings uncover the potential of feature scores from histology images as digital biomarkers in precision medicine and as an expanding utility for digital pathology.

The central importance of large scale eigenvalue problems in scientific computation necessitates the development of massively parallel algorithms for their solution. Recent advances in dense numerical linear algebra have enabled the routine treatment of eigenvalue problems with dimensions on the order of hundreds of thousands on the world's largest supercomputers. In cases where dense treatments are not feasible, Krylov subspace methods offer an attractive alternative due to the fact that they do not require storage of the problem matrices. However, demonstration of scalability of either of these classes of eigenvalue algorithms on computing architectures capable of expressing massive parallelism is non-trivial due to communication requirements and serial bottlenecks, respectively. In this work, we introduce the SISLICE method: a parallel shift-invert algorithm for the solution of the symmetric self-consistent field (SCF) eigenvalue problem. The SISLICE method drastically reduces the communication requirement of current parallel shift-invert eigenvalue algorithms through various shift selection and migration techniques based on density of states estimation and k-means clustering, respectively. This work demonstrates the robustness and parallel performance of the SISLICE method on a representative set of SCF eigenvalue problems and outlines research directions which will be explored in future work.

The standard training algorithm in neural machine translation (NMT) suffers from exposure bias, and alternative algorithms have been proposed to mitigate this. However, the practical impact of exposure bias is under debate. In this paper, we link exposure bias to another well-known problem in NMT, namely the tendency to generate hallucinations under domain shift. In experiments on three datasets with multiple test domains, we show that exposure bias is partially to blame for hallucinations, and that training with Minimum Risk Training, which avoids exposure bias, can mitigate this. Our analysis explains why exposure bias is more problematic under domain shift, and also links exposure bias to the beam search problem, i.e. performance deterioration with increasing beam size. Our results provide a new justification for methods that reduce exposure bias: even if they do not increase performance on in-domain test sets, they can increase model robustness to domain shift.

Sentence level quality estimation (QE) for machine translation (MT) attempts to predict the translation edit rate (TER) cost of post-editing work required to correct MT output. We describe our view on sentence-level QE as dictated by several practical setups encountered in the industry. We find consumers of MT output---whether human or algorithmic ones---to be primarily interested in a binary quality metric: is the translated sentence adequate as-is or does it need post-editing? Motivated by this we propose a quality classification (QC) view on sentence-level QE whereby we focus on maximizing recall at precision above a given threshold. We demonstrate that, while classical QE regression models fare poorly on this task, they can be re-purposed by replacing the output regression layer with a binary classification one, achieving 50-60\% recall at 90\% precision. For a high-quality MT system producing 75-80\% correct translations, this promises a significant reduction in post-editing work indeed.

In encoder-decoder neural models, multiple encoders are in general used to represent the contextual information in addition to the individual sentence. In this paper, we investigate multi-encoder approaches in documentlevel neural machine translation (NMT). Surprisingly, we find that the context encoder does not only encode the surrounding sentences but also behaves as a noise generator. This makes us rethink the real benefits of multi-encoder in context-aware translation - some of the improvements come from robust training. We compare several methods that introduce noise and/or well-tuned dropout setup into the training of these encoders. Experimental results show that noisy training plays an important role in multi-encoder-based NMT, especially when the training data is small. Also, we establish a new state-of-the-art on IWSLT Fr-En task by careful use of noise generation and dropout methods.

Thin structures, such as wire-frame sculptures, fences, cables, power lines, and tree branches, are common in the real world.

It is extremely challenging to acquire their 3D digital models using traditional image-based or depth-based reconstruction methods because thin structures often lack distinct point features and have severe self-occlusion.

We propose the first approach that simultaneously estimates camera motion and reconstructs the geometry of complex 3D thin structures in high quality from a color video captured by a handheld camera.

Specifically, we present a new curve-based approach to estimate accurate camera poses by establishing correspondences between featureless thin objects in the foreground in consecutive video frames, without requiring visual texture in the background scene to lock on.

Enabled by this effective curve-based camera pose estimation strategy, we develop an iterative optimization method with tailored measures on geometry, topology as well as self-occlusion handling for reconstructing 3D thin structures.

Extensive validations on a variety of thin structures show that our method achieves accurate camera pose estimation and faithful reconstruction of 3D thin structures with complex shape and topology at a level that has not been attained by other existing reconstruction methods.

Neural machine translation (NMT) needs large parallel corpora for state-of-the-art translation quality. Low-resource NMT is typically addressed by transfer learning which leverages large monolingual or parallel corpora for pre-training. Monolingual pre-training approaches such as MASS (MAsked Sequence to Sequence) are extremely effective in boosting NMT quality for languages with small parallel corpora. However, they do not account for linguistic information obtained using syntactic analyzers which is known to be invaluable for several Natural Language Processing (NLP) tasks. To this end, we propose JASS, Japanese-specific Sequence to Sequence, as a novel pre-training alternative to MASS for NMT involving Japanese as the source or target language. JASS is joint BMASS (Bunsetsu MASS) and BRSS (Bunsetsu Reordering Sequence to Sequence) pre-training which focuses on Japanese linguistic units called bunsetsus. In our experiments on ASPEC Japanese--English and News Commentary Japanese--Russian translation we show that JASS can give results that are competitive with if not better than those given by MASS. Furthermore, we show for the first time that joint MASS and JASS pre-training gives results that significantly surpass the individual methods indicating their complementary nature. We will release our code, pre-trained models and bunsetsu annotated data as resources for researchers to use in their own NLP tasks.

Unsupervised machine translation (MT) has recently achieved impressive results with monolingual corpora only. However, it is still challenging to associate source-target sentences in the latent space. As people speak different languages biologically share similar visual systems, the potential of achieving better alignment through visual content is promising yet under-explored in unsupervised multimodal MT (MMT). In this paper, we investigate how to utilize visual content for disambiguation and promoting latent space alignment in unsupervised MMT. Our model employs multimodal back-translation and features pseudo visual pivoting in which we learn a shared multilingual visual-semantic embedding space and incorporate visually-pivoted captioning as additional weak supervision. The experimental results on the widely used Multi30K dataset show that the proposed model significantly improves over the state-of-the-art methods and generalizes well when the images are not available at the testing time.

By learning from a MERS outbreak in 2015, South Korea was prepared and acted swiftly to ramp up testing when the new coronavirus appeared there. Meanwhile, the U.S., plagued by delay and dysfunction, wasted its advantage. By Stephen Engelberg, Lisa Song and Lydia DePillis ProPublica…

[IMAGE-1] Watching Rian Johnson's Knives Out, I was reminded of my middle school English teacher Mrs. Soderbergh, who loved Agatha Christie books almost as much as she loved diagramming sentences. There was a week when she brought in a box stacked high with her own Christie paperbacks, set it down in front of the classroom and had each of us pick a book based solely on the plot summary on the back.…

Earlier this morning, Sleater-Kinney announced on Twitter that they're hitting the road on a co-headlining tour with Wilco this summer. Great news!…

March is obviously going to be the busiest month for any purveyor of traditional Irish music, and with St. Patrick's Day right around the corner, Spokane's Floating Crowbar has multiple gigs crowding the week's calendar.…

By David Yaffe-Bellany and Michael Corkery The New York Times Company Hundreds of Wendy’s restaurants have run out of hamburgers.…

The puppet's felt hair bounces as she stage-whispers to the other puppets, almost conspiratorially, about their plans.…

Mosey on up to the bar at Lil Sumthin' Saloon for a sip of Southern hospitality by way of Texas, and a samplin' of some old-fashioned country vibes.…

Weeks after being raped at a wedding — an experience already wrapped in feelings of self-blame and fear — the 18-year-old Eastern Washington University student realized something else was wrong.…

At around 5:30 pm on Thursday, there were two camps of people set up in Coeur d'Alene Park in Browne's Addition.…

Various embodiments provide lossless compression of an enumeration space for genetic founder lines. In one embodiment, an input comprising a set of genetic founder lines and a maximum number of generations G is obtained. A set of genetic crossing templates of a height h is generated. A determination is made if at least a first genetic crossing template in the set of genetic crossing templates is redundant with respect to a second genetic crossing template in the set of genetic crossing templates. Based on the at least first genetic crossing template being redundant is redundant with respect to the second genetic crossing template, the at least first genetic crossing template is removed from the set of genetic crossing templates. This process of removing the at least first genetic crossing template from the set of genetic crossing templates the redundant creates an updated set of genetic crossing templates.

Stone is formed from 5 to 60% by weight of polymerised, low-viscosity, transparent or low-colour-resin, 20 to 90% by weight of spherical alumina trihidrate Al2O3.3H2O containing less regular particles containing, advantageously 0 to 100% by weight of a transparent or translucent substitute of alumina trihydrate, and/or with 0 to 20% or pre-prepared particulate, filled resin of a chosen colour, and/or mineral particles and less than 2% by weight of luminophor. These individual components are mixed intensely whilst extracting included gaseous parts. Extraction is carried out whilst mixing, and/or after mixing, and/or before mixing. The mixture is initiated by introducing a starter and intensely mixing it into the mixture. The mixture is poured into a mould or onto a moving endless belt. The cured synthetic stone is removed from the mould or the hardened composite is taken off the the belt. Synthetic stone can be used in products as a light carrier.

A method begins by a processing module obtaining common storage name information regarding data that is stored in storage units of a distributed storage network (DSN) as a set of data slices. Each data slice of the set of data slices has a unique storage name, where each of the unique storage names for the set of data slices has common naming information regarding the data. The method continues where the processing module interprets the common storage name information to determine whether a difference exists between the common naming information of a data slice of the set of data slices and the common naming information of other data slices of the set of data slices. When the difference exists, the method continues where the processing module indicates a potential storage error of the data slice and implements a storage error process regarding the potential storage error of the data slice.

A continuous process for the thermal treatment of a refinery sludge, comprising the following operations: a. drying of the refinery sludge, possibly mixed with pet-coke, at a temperature ranging from 110 to 120° C.; b. gasification of the dried sludge, at a temperature ranging from 750 to 950° C., for a time of 30 to 60 minutes, in the presence of a gas containing oxygen and water vapour, with the associated production of synthesis gas (CO+H2) and a solid residue; c. combustion of the synthesis gas at a temperature ranging from 850 to 1,200° C. and recycling of the combustion products for the drying and gasification phases; and d. inertization of the solid residue, at a temperature ranging from 1,300 to 1,500° C., by vitrification with plasma torches.

Described is a radiation-crosslinkable hotmelt adhesive comprising at least one radiation-crosslinkable poly(meth)acrylate formed to an extent of at least 60% by weight of C1 to C10 alkyl(meth)acrylates and at least one oligo(meth)acrylate which comprises nonacrylic C C double bonds and has a K value of less than or equal to 20. The hotmelt adhesive comprises a photoinitiator which may be present in the form of an additive not attached to the poly(meth)acrylate and/or not attached to the oligo(meth)acrylate, may be incorporated by copolymerization into the poly(meth)acrylate, and/or may be attached to the oligo(meth)acrylate. The hotmelt adhesive can be used for producing adhesive tapes.

A crosslinkable curing super-branched polyester and the cured product and the preparation method thereof are disclosed. The super-branched polyester has high refractive index and comprises a compound represented by the following structural formula (I). In the formula (I), HBP is the backbone of the super-branched polyester; both a and b are positive integers; the sum of a and b is less than or equal to n; n is more than or equal to 10 and less than 80. In the super-branched polyester, A is represented by formula (II) and N is represented by formula (III), wherein R is methyl or hydrogen atom; the mole ratio of N relative to the total mole of A and N is more than 30 mol %, and the ratio of the total mole of A and N relative to the product of the total mole of HBP backbone and n is more than 0.5 and less than or equal to 1.

The present disclosure provides a method of providing an adhesive composition comprising the steps of combining crosslinkable composition including: a) a (meth)acryloyl monomer mixture with the b) photocrosslinking agent mixture, and irradiating with UVC radiation to polymerize and crosslink the composition.

A method and apparatus for method of continuously producing 1,1,1,2,3-pentafluoropropane with high yield is provided. The method includes (a) bringing a CoF3-containing cobalt fluoride in a reactor into contact with 3,3,3-trifluoropropene to produce a CoF2-containing cobalt fluoride and 1,1,1,2,3-pentafluoropropane, (b) transferring the CoF2-containing cobalt fluoride in the reactor to a regenerator and bringing the transferred CoF2-containing cobalt fluoride into contact with fluorine gas to regenerate a CoF3-containing cobalt fluoride, and (c) transferring the CoF3-containing cobalt fluoride in the regenerator to the reactor and employing the transferred CoF3-containing cobalt fluoride in Operation (a). Accordingly, the 1,1,1,2,3-pentafluoropropane can be continuously produced with high yield from the 3,3,3-trifluoropropene using a cobalt fluoride (CoF2/CoF3) as a fluid catalyst, thereby improving the reaction stability and readily adjusting the optimum conversion rate and selectivity.

The disclosure provides a method for crosslinking a colloid, including: (a) providing a colloid solution; (b) adding a crosslinking agent and solid particles to the colloid solution, wherein the amount of solid particles added is enough to convert the colloid solution into a solid mixture, and wherein a crosslinking reaction proceeds in the solid mixture; and (c) removing the solid particles from the solid mixture.

Methods and systems for optimizing generation of natively executable code from non-native binary code are disclosed. One method includes receiving a source file including binary code configured for execution according to a non-native instruction set architecture. The method also includes translating one or more code blocks included in the executable binary code to source code, and applying an optimizing algorithm to instructions in the one or more code blocks. The optimizing algorithm is selected to reduce a number of memory address translations performed when translating the source code to native executable binary code, thereby resulting in one or more optimized code blocks. The method further includes compiling the source code to generate an output file comprising natively executable binary code including the one or more optimized code blocks.

A system and method may generate unit tests for source code concurrently with API documentation. The system may receive a source code file including several comments sections. Each comments section may include a description of a source code unit such as a class, method, member variable, etc. The description may also correspond to input and output parameters the source code unit. The system and method may parsing the source code file to determine a source code function type corresponding to the unit description and copy the unit description to a unit test stub corresponding to the function type. A developer or another module may then complete the unit test stub to transform each stub into a complete unit test corresponding to the source code unit. Additionally, the system and method may execute the unit test and generate a test result indication for each unit test.

A single software project in an integrated development environment (IDE) may be built for multiple target environments in a single build episode. Multiple different output artifacts may be generated by the build process for each of the target environments. The output artifacts are then deployed to the target environments, which may be homogeneous or heterogeneous environments. The same source project may be used to generate multiple output artifacts for the same target environment.

A method is provided for providing consistent logical code across specific programming languages. The method incorporates preprocessor macros in a source computer program code to generate a program control flow. The preprocessor macros can be used to describe program control flow in the source programming language for execution in the source computer program code. The preprocessor macros can also be used to generate control flow objects representing the control flow, which converts the source computer program code into a general language representation. The general language representation when executed is used to output computer programming code in specific programming languages representing the same logical code as that of the source computer program code.

In one embodiment of the invention, a method is disclosed for modifying a pre-existing application program for multi-processing and/or distributed parallel processing. The method includes searching an application program for a computational loop; analyzing the computational loop to determine independence of the computational transactions of the computational loop; and replacing the computational loop with master code and slave code to provide master-slave execution of the computational loop in response to analyzing the computational loop to determine independence of the computational transactions of the computational loop. Multiple instances of the modified application program are executed to provide multi-processing and/or distributed parallel processing.

A cleansing composition for cosmetic or household use may include an ampholytic polymer; a crosslinked cationic polymer; a surfactant component selected from the group consisting of anionic surfactants, amphoteric surfactants, cationic surfactants, nonionic surfactants, and zwitterionic surfactants; and an aqueous and/or organic carrier.

Low bit rate audio coding such as BWE algorithm often encounters conflict goal of achieving high time resolution and high frequency resolution at the same time. In order to achieve best possible quality, input signal can be first classified into fast signal and slow signal. This invention focuses on classifying signal into fast signal and slow signal, based on at least one of the following parameters or a combination of the following parameters: spectral sharpness, temporal sharpness, pitch correlation (pitch gain), and/or spectral envelope variation. This classification information can help to choose different BWE algorithms, different coding algorithms, and different postprocessing algorithms respectively for fast signal and slow signal.

A messaging response system is disclosed wherein a service providing system provides services to users via messaging communications. In accordance with an exemplary embodiment of the present invention, multiple respondents servicing users through messaging communications may appear to simultaneously use a common “screen name” identifier.

A system for translating user motion into multiple object responses of an on-screen object based on user interaction of an application executing on a computing device is provided. User motion data is received from a capture device from one or more users. The user motion data corresponds to user interaction with an on-screen object presented in the application. The on-screen object corresponds to an object other than an on-screen representation of a user that is displayed by the computing device. The user motion data is automatically translated into multiple object responses of the on-screen object. The multiple object responses of the on-screen object are simultaneously displayed to the users.

Various systems and methods are described for configuring a data storage system. In one embodiment, a plurality of actual capacities of a plurality of storage devices of the data storage system are identified and divided into a plurality of capacity slices. The plurality of capacity slices are combined into a plurality of chunks of capacity slices, each having a combination of characteristics of the underlying physical storage devices. The chunks of capacity slices are then mapped to a plurality of logical storage devices. A group of the plurality of logical storage devices is then organized into a redundant array of logical storage devices.

A sequence of storage devices of a data store may include one or more stripesets for storing data stripes of different lengths and of different types. Each data stripe may be stored in a prefix or other portion of a stripeset. Each data stripe may be identified by an array of addresses that identify each page of the data stripe on each included storage device. When a first storage device of a stripeset becomes full, the stripeset may be shifted by removing the full storage device from the stripeset, and adding a next storage device of the data store to the stripeset. A class variable may be associated with storage devices of a stripeset to identify the type of data that the stripeset can store. The class variable may be increased (or otherwise modified) when a computer stores data of a different class in the stripeset.

In a process for removing aromatic carboxylic acid from a slurry thereof in solvent, the slurry is split into sub streams and each of said sub streams is supplied to a respective rotary pressure filter such that the sub stream pass through the filters in parallel. Gas is passed through the rotary pressure filters in series in an open-loop arrangement.

Arrangements for restarting data transmission on a serial low-power inter-chip media bus (SLIMbus) are presented. A clock signal may be provided in an active mode to a component communicatively coupled with the SLIMbus. Immediately prior to the clock signal in the active mode being provided, the clock signal may have been in a paused mode. While the clock signal was in the paused mode at least until the clock signal is provided in the active mode, the data line may have been inactive (e.g., a toggle on the data line may not have been present). Frame synchronization data for a frame may be transmitted. The frame synchronization data for the frame, as received by the component, may not match expected frame synchronization data. Payload data may be transmitted as part of the frame to the component, wherein the payload data is expected to be read properly by the component.

A method, apparatus, article of manufacture, and computer program product display multiple slices of a three-dimensional body in a computer drawing application. A first view of a 3D body is displayed. One or more slice sketches are defined on the first view. A single slice view of the three dimensional body is displayed. The single slice view simultaneously depicts all of the slices with a zero-depth representation of each slice.

Embodiments include a fiducial deployment system and method for use thereof. A fiducial may include one or more protuberances configured to engage one or more slots in a needle of the system. The needle may be configured to deliver a plurality of fiducials to a target location in serial fashion, one at a time. In certain embodiments, echogenic placement of fiducials may present certain advantages.

A method for continuously preparing a carboxylic acid ester is disclosed. In the method of the present invention, a vertical reactor is filled with a solid catalyst, a carboxylic acid and an alcohol are introduced into a lower part of the vertical reactor, esterification is performed to form an esterized mixture, the esterized mixture is output from an upper part of the vertical reactor, and distillation is performed to isolate the carboxylic acid ester. The method of the present invention is simple, easily controlled and environmental friendly, and has significantly high conversion rate and selectivity.

Solid-state deformation processing of crosslinked high molecular weight polymers such as UHMWPE, for example by extrusion below the melt transition, produces materials with a combination of high tensile strength and high oxidative stability. The materials are especially suitable for use as bearing components in artificial hip and other implants. Treated bulk materials are anisotropic, with enhanced strength oriented along the axial direction. The material is oxidatively stable even after four weeks of accelerated aging in a pressure vessel containing five atmospheres of oxygen (ASTM F2003). Because of its oxidative stability, the deformation processed material is a suitable candidate for air-permeable packaging and gas sterilization, which has thus far been reserved for remelted crosslinked UHMWPE.

The present invention relates to a composition comprising at least one propylene-based polymer comprising less than 0.1 wt. % diene-derived units based on the weight of the propylene-based polymer, an antioxidant, and a co-agent. The composition can be at least partially crosslinked by electron beam irradiation in a dose of less than 200 kGy, and may be further formed into articles including fibers, yarns, films, and nonwovens, among others. The propylene-based polymer of the present invention may be a polymer blend formed by forming a reactor blend from of two or more polymers produced in two or more reactors.

A sliding Capo for a musical instrument of the lute family which rides in a track in the neck of the instrument. The Capo includes a rigid cross piece holding elongated flanges which ride in the track. A cylindrical elastomeric roller runs in bearings carried by the cross piece, and adjustment means are provided for adjusting the height of the roller above the fingerboard. In the case of curved fingerboards, the shaft on which the roller runs is flexible and springy, so that it can bend and the cylindrical roller can press all strings to the fingerboard.

A vehicle powertrain has an engine, a driving shaft having a passage defined therein, a driven shaft, a pump, a hydraulic fluid reservoir, and a CVT. A driving pulley of the CVT includes a fixed sheave and a movable sheave disposed on the driving shaft for rotation therewith, a spring biasing a movable sheave away from the fixed sheave, and a CVT chamber fluidly communicating with the passage of the driving shaft. The pump supplies hydraulic fluid from the reservoir to the passage in the driving shaft. The hydraulic fluid flows from the passage to the CVT chamber to create a hydraulic pressure in the CVT chamber. The hydraulic pressure in the CVT chamber biases the movable sheave toward the fixed sheave.

A continuously variable transmission for a vehicle comprises a primary pulley, a secondary pulley, and a belt entrained around both pulleys in an opposite manner between the fixed and movable sheaves thereof. One pulley carries a selectively deployable stop that may be placed into position to mechanically limit the range of motion between the fixed and movable sheaves. This establishes an actual maximum transmission ratio that is larger than the nominal minimum transmission ratio but smaller than the nominal maximum transmission ratio achievable by the transmission during normal operation thereof. When such a transmission is used on a utility vehicle attached to a substance dispensing applicator operatively driven by the vehicle's engine, the application rate stays substantially constant despite changes in engine speed when the user keeps the transmission upshifted to the actual maximum transmission ratio set by the position of the deployed stop.

A belt-type continuously variable transmission for a vehicle includes: a primary pulley including a first fixed sheave fixed to an input shaft, and a first movable sheave relatively non-rotatable and relatively movable in axial direction to the input shaft due to spline-fitting of a female spline formed on an inner circumferential portion thereof to a male spline formed on the input shaft; a secondary pulley including a second fixed sheave fixed to an output shaft in parallel with the input shaft, and a second movable sheave relatively non-rotatable and relatively movable in the axial direction to the output shaft due to spline-fitting of a female spline formed on an inner circumferential portion thereof to a male spline formed on the output shaft; and a transmission belt, the belt-type continuously variable transmission for a vehicle disposed with at least one of a first groove width defining portion disposed on the primary pulley for defining a maximum groove width of the first pulley groove and a second groove width defining portion disposed on the secondary pulley for defining a maximum groove width of the second pulley groove, the first groove width defining portion and the second groove width defining portion being disposed on one shaft of the input shaft and the output shaft, and the other shaft having a uniform shaft diameter in an range of sliding of the first movable sheave or the second movable sheave in the axial direction except a site with the male spline formed, the other shaft being disposed on the opposite side to the one shaft.

A belt-driven continuously variable transmission having a downsized torque cam assembly for creating forward thrust to push a movable sheave toward a fixed sheave. The belt-driven continuously variable transmission is adapted to output a torque while changing a speed ratio continuously by altering an effective diameter position of a driving belt. The belt-driven continuously variable transmission is comprised of: an output gear, which is fitted onto the rotary shaft of the driven pulley to output the torque while being allowed to rotate relatively; a piston, which is fitted onto the rotary shaft of the driven pulley between the movable sheave and the output gear while being allowed to rotate integrally with the movable sheave and to reciprocate in an axial direction of the rotary shaft, and which increases an internal pressure of the hydraulic chamber by moving toward the movable sheave; and a torque cam assembly, which is interposed between the piston and the output gear, and which pushes the piston toward the movable sheave when the torque transmitted through the driven pulley is increased.