Publications
77 results found
Goodman DFM, Stimberg M, Brette R, 2016, Brian 2.0 simulator
Brian is a simulator for spiking neural networks. It is written in the Python programming language and is available on almost all platforms. We believe that a simulator should not only save the time of processors, but also the time of scientists. Brian is therefore designed to be easy to learn and use, highly flexible and easily extensible.
Rossant C, Kadir SN, Goodman DF, et al., 2016, Spike sorting for large, dense electrode arrays, Nature Neuroscience, Vol: 19, Pages: 634-641, ISSN: 1546-1726
Developments in microfabrication technology have enabled the production of neural electrode arrays with hundreds of closely spaced recording sites, and electrodes with thousands of sites are under development. These probes in principle allow the simultaneous recording of very large numbers of neurons. However, use of this technology requires the development of techniques for decoding the spike times of the recorded neurons from the raw data captured from the probes. Here we present a set of tools to solve this problem, implemented in a suite of practical, user-friendly, open-source software. We validate these methods on data from the cortex, hippocampus and thalamus of rat, mouse, macaque and marmoset, demonstrating error rates as low as 5%.
Kadir SN, Goodman DFM, Harris KD, 2014, High-dimensional cluster analysis with the masked EM algorithm, Neural Computation, Vol: 26, Pages: 2379-2394, ISSN: 0899-7667
Cluster analysis faces two problems in high dimensions: the "curse of dimensionality" that can lead to overfitting and poor generalization performance and the sheer time taken for conventional algorithms to process large amounts of high-dimensional data. We describe a solution to these problems, designed for the application of spike sorting for nextgeneration, high-channel-count neural probes. In this problem, only a small subset of features provides information about the cluster membership of any one data vector, but this informative feature subset is not the same for all data points, rendering classical feature selection ineffective.We introduce a "masked EM" algorithm that allows accurate and time-efficient clustering of up to millions of points in thousands of dimensions. We demonstrate its applicability to synthetic data and to real-world high-channel-count spike sorting data.
Stimberg M, Goodman DF, Benichoux V, et al., 2014, Equation-oriented specification of neural models for simulations, Frontiers in Neuroinformatics, Vol: 8, ISSN: 1662-5196
Simulating biological neuronal networks is a core method of research in computational neuroscience. A full specification of such a network model includes a description of the dynamics and state changes of neurons and synapses, as well as the synaptic connectivity patterns and the initial values of all parameters. A standard approach in neuronal modeling software is to build network models based on a library of pre-defined components and mechanisms; if a model component does not yet exist, it has to be defined in a special-purpose or general low-level language and potentially be compiled and linked with the simulator. Here we propose an alternative approach that allows flexible definition of models by writing textual descriptions based on mathematical notation. We demonstrate that this approach allows the definition of a wide range of models with minimal syntax. Furthermore, such explicit model descriptions allow the generation of executable code for various target languages and devices, since the description is not tied to an implementation. Finally, this approach also has advantages for readability and reproducibility, because the model description is fully explicit, and because it can be automatically parsed and transformed into formatted descriptions. The presented approach has been implemented in the Brian2 simulator.
Goodman DF, Benichoux V, Brette R, 2013, Decoding neural responses to temporal cues for sound localization, eLife, Vol: 2, ISSN: 2050-084X
The activity of sensory neural populations carries information about the environment. This may be extracted from neural activity using different strategies. In the auditory brainstem, a recent theory proposes that sound location in the horizontal plane is decoded from the relative summed activity of two populations in each hemisphere, whereas earlier theories hypothesized that the location was decoded from the identity of the most active cells. We tested the performance of various decoders of neural responses in increasingly complex acoustical situations, including spectrum variations, noise, and sound diffraction. We demonstrate that there is insufficient information in the pooled activity of each hemisphere to estimate sound direction in a reliable way consistent with behavior, whereas robust estimates can be obtained from neural activity by taking into account the heterogeneous tuning of cells. These estimates can still be obtained when only contralateral neural responses are used, consistently with unilateral lesion studies. DOI: http://dx.doi.org/10.7554/eLife.01312.001.
Rossant C, Fontaine B, Goodman DFM, 2013, Playdoh: A lightweight Python library for distributed computing and optimisation, JOURNAL OF COMPUTATIONAL SCIENCE, Vol: 4, Pages: 352-359, ISSN: 1877-7503
- Author Web Link
- Cite
- Citations: 8
Brette R, Goodman DFM, 2012, Simulating spiking neural networks on GPU, NETWORK-COMPUTATION IN NEURAL SYSTEMS, Vol: 23, Pages: 167-182, ISSN: 0954-898X
- Author Web Link
- Cite
- Citations: 43
Fontaine B, Goodman DF, Benichoux V, et al., 2011, Brian hears: online auditory processing using vectorization over channels, Frontiers in Neuroinformatics, Vol: 5, ISSN: 1662-5196
The human cochlea includes about 3000 inner hair cells which filter sounds at frequencies between 20 Hz and 20 kHz. This massively parallel frequency analysis is reflected in models of auditory processing, which are often based on banks of filters. However, existing implementations do not exploit this parallelism. Here we propose algorithms to simulate these models by vectorizing computation over frequency channels, which are implemented in "Brian Hears," a library for the spiking neural network simulator package "Brian." This approach allows us to use high-level programming languages such as Python, because with vectorized operations, the computational cost of interpretation represents a small fraction of the total cost. This makes it possible to define and simulate complex models in a simple way, while all previous implementations were model-specific. In addition, we show that these algorithms can be naturally parallelized using graphics processing units, yielding substantial speed improvements. We demonstrate these algorithms with several state-of-the-art cochlear models, and show that they compare favorably with existing, less flexible, implementations.
Kremer Y, Leger J-F, Goodman D, et al., 2011, Late Emergence of the Vibrissa Direction Selectivity Map in the Rat Barrel Cortex, JOURNAL OF NEUROSCIENCE, Vol: 31, Pages: 10689-10700, ISSN: 0270-6474
- Author Web Link
- Cite
- Citations: 47
Brette R, Goodman DFM, 2011, Vectorized Algorithms for Spiking Neural Network Simulation, NEURAL COMPUTATION, Vol: 23, Pages: 1503-1535, ISSN: 0899-7667
- Author Web Link
- Cite
- Citations: 22
Rossant C, Goodman DF, Fontaine B, et al., 2011, Fitting neuron models to spike trains, Frontiers in Neuroscience, Vol: 5, ISSN: 1662-4548
Computational modeling is increasingly used to understand the function of neural circuits in systems neuroscience. These studies require models of individual neurons with realistic input-output properties. Recently, it was found that spiking models can accurately predict the precisely timed spike trains produced by cortical neurons in response to somatically injected currents, if properly fitted. This requires fitting techniques that are efficient and flexible enough to easily test different candidate models. We present a generic solution, based on the Brian simulator (a neural network simulator in Python), which allows the user to define and fit arbitrary neuron models to electrophysiological recordings. It relies on vectorization and parallel computing techniques to achieve efficiency. We demonstrate its use on neural recordings in the barrel cortex and in the auditory brainstem, and confirm that simple adaptive spiking models can accurately predict the response of cortical neurons. Finally, we show how a complex multicompartmental model can be reduced to a simple effective spiking model.
Goodman DFM, Brette R, 2010, Learning to localise sounds with spiking neural networks
To localise the source of a sound, we use location-specific properties of the signals received at the two ears caused by the asymmetric filtering of the original sound by our head and pinnae, the head-related transfer functions (HRTFs). These HRTFs change throughout an organism's lifetime, during development for example, and so the required neural circuitry cannot be entirely hardwired. Since HRTFs are not directly accessible from perceptual experience, they can only be inferred from filtered sounds. We present a spiking neural network model of sound localisation based on extracting location-specific synchrony patterns, and a simple supervised algorithm to learn the mapping between synchrony patterns and locations from a set of example sounds, with no previous knowledge of HRTFs. After learning, our model was able to accurately localise new sounds in both azimuth and elevation, including the difficult task of distinguishing sounds coming from the front and back.
Fletcher A, Goodman D, 2010, Quasiregular mappings of polynomial type in R<sup>2</sup>, Conformal Geometry and Dynamics, Vol: 14, Pages: 322-336
Complex dynamics deals with the iteration of holomorphic functions. As is well known, the first functions to be studied which gave non-trivial dynamics were quadratic polynomials, which produced beautiful computer generated pictures of Julia sets and the Mandelbrot set. In the same spirit, this article aims to study the dynamics of the simplest non-trivial quasiregular mappings. These are mappings in R2 which are a composition of a quadratic polynomial and an affine stretch. © 2010 American Mathematical Society.
Goodman DF, Brette R, 2010, Spike-timing-based computation in sound localization, PLOS Computational Biology, Vol: 6, ISSN: 1553-734X
Spike timing is precise in the auditory system and it has been argued that it conveys information about auditory stimuli, in particular about the location of a sound source. However, beyond simple time differences, the way in which neurons might extract this information is unclear and the potential computational advantages are unknown. The computational difficulty of this task for an animal is to locate the source of an unexpected sound from two monaural signals that are highly dependent on the unknown source signal. In neuron models consisting of spectro-temporal filtering and spiking nonlinearity, we found that the binaural structure induced by spatialized sounds is mapped to synchrony patterns that depend on source location rather than on source signal. Location-specific synchrony patterns would then result in the activation of location-specific assemblies of postsynaptic neurons. We designed a spiking neuron model which exploited this principle to locate a variety of sound sources in a virtual acoustic environment using measured human head-related transfer functions. The model was able to accurately estimate the location of previously unknown sounds in both azimuth and elevation (including front/back discrimination) in a known acoustic environment. We found that multiple representations of different acoustic environments could coexist as sets of overlapping neural assemblies which could be associated with spatial locations by Hebbian learning. The model demonstrates the computational relevance of relative spike timing to extract spatial information about sources independently of the source signal.
Goodman DFM, 2010, Code Generation: A Strategy for Neural Network Simulators, NEUROINFORMATICS, Vol: 8, Pages: 183-196, ISSN: 1539-2791
- Author Web Link
- Cite
- Citations: 21
Rossant C, Goodman DF, Platkiewicz J, et al., 2010, Automatic fitting of spiking neuron models to electrophysiological recordings, Frontiers in Neuroinformatics, Vol: 4, ISSN: 1662-5196
Spiking models can accurately predict the spike trains produced by cortical neurons in response to somatically injected currents. Since the specific characteristics of the model depend on the neuron, a computational method is required to fit models to electrophysiological recordings. The fitting procedure can be very time consuming both in terms of computer simulations and in terms of code writing. We present algorithms to fit spiking models to electrophysiological data (time-varying input and spike trains) that can run in parallel on graphics processing units (GPUs). The model fitting library is interfaced with Brian, a neural network simulator in Python. If a GPU is present it uses just-in-time compilation to translate model equations into optimized code. Arbitrary models can then be defined at script level and run on the graphics card. This tool can be used to obtain empirically validated spiking models of neurons in various systems. We demonstrate its use on public data from the INCF Quantitative Single-Neuron Modeling 2009 competition by comparing the performance of a number of neuron spiking models.
Goodman DFM, Brette R, 2010, Learning to localise sounds with spiking neural networks
To localise the source of a sound, we use location-specific properties of the signals received at the two ears caused by the asymmetric filtering of the original sound by our head and pinnae, the head-related transfer functions (HRTFs). These HRTFs change throughout an organism's lifetime, during development for example, and so the required neural circuitry cannot be entirely hardwired. Since HRTFs are not directly accessible from perceptual experience, they can only be inferred from filtered sounds. We present a spiking neural network model of sound localisation based on extracting location-specific synchrony patterns, and a simple supervised algorithm to learn the mapping between synchrony patterns and locations from a set of example sounds, with no previous knowledge of HRTFs. After learning, our model was able to accurately localise new sounds in both azimuth and elevation, including the difficult task of distinguishing sounds coming from the front and back.
de Alba Campomanes AG, Rutar T, Crawford JB, et al., 2009, Crystal-storing histiocytosis and crystalline keratopathy caused by monoclonal gammopathy of undetermined significance., Cornea, Vol: 28, Pages: 1081-1084
PURPOSE: The purpose of this study was to report the occurrence of crystalline keratopathy and of orbital infiltrative disease resulting from crystal-storing histiocytosis (CSH) in a patient with monoclonal gammopathy of undetermined significance. METHODS: The authors conducted a review of a medical record and immunohistopathologic studies. RESULTS: A 66-year-old man presented with a 3-year history of unilateral progressive ptosis, proptosis, and external ophthalmoplegia. Magnetic resonance imaging showed orbital fat expansion and extraocular muscle thickening with gadolinium enhancement. The patient also had bilateral crystalline keratopathy and had undergone penetrating keratoplasty in one eye. The urine and serum showed elevated levels of immunoglobulin, but the bone marrow aspirate was normal. The systemic evaluation was consistent with monoclonal gammopathy of undetermined significance. An orbital fat biopsy revealed histiocytes engorged with lambda light chain crystals. The corneal stroma also showed positive immunostaining for lambda light chains. The patient was diagnosed with orbital CSH and with lambda light chain crystalline keratopathy. CONCLUSIONS: CSH is characterized by the accumulation of reactive histiocytes filled with immunoglobulin crystals in various tissues and is frequently associated with systemic hyperglobulinemic states. For unknown reasons, in this patient, a systemic immunologic disorder led to lambda light chain abnormalities with histiocyte infiltration of the orbit and corneal deposition. Although CSH is rare, it should be part of the differential diagnosis of orbital infiltrative disease with or without crystalline keratopathy.
Goodman DF, Brette R, 2009, The brian simulator, Frontiers in Neuroscience, Vol: 3, Pages: 192-197, ISSN: 1662-4548
"Brian" is a simulator for spiking neural networks (http://www.briansimulator.org). The focus is on making the writing of simulation code as quick and easy as possible for the user, and on flexibility: new and non-standard models are no more difficult to define than standard ones. This allows scientists to spend more time on the details of their models, and less on their implementation. Neuron models are defined by writing differential equations in standard mathematical notation, facilitating scientific communication. Brian is written in the Python programming language, and uses vector-based computation to allow for efficient simulations. It is particularly useful for neuroscientific modelling at the systems level, and for teaching computational neuroscience.
Goodman D, Brette R, 2008, Brian: a simulator for spiking neural networks in python, Frontiers in Neuroinformatics, Vol: 2, ISSN: 1662-5196
"Brian" is a new simulator for spiking neural networks, written in Python (http://brian. di.ens.fr). It is an intuitive and highly flexible tool for rapidly developing new models, especially networks of single-compartment neurons. In addition to using standard types of neuron models, users can define models by writing arbitrary differential equations in ordinary mathematical notation. Python scientific libraries can also be used for defining models and analysing data. Vectorisation techniques allow efficient simulations despite the overheads of an interpreted language. Brian will be especially valuable for working on non-standard neuron models not easily covered by existing software, and as an alternative to using Matlab or C for simulations. With its easy and intuitive syntax, Brian is also very well suited for teaching computational neuroscience.
Goodman DFM, Stimberg M, Brette R, 2008, Brian simulator
Brian is a simulator for spiking neural networks available on almost all platforms. The motivation for this project is that a simulator should not only save the time of processors, but also the time of scientists.Brian is easy to learn and use, highly flexible and easily extensible. The Brian package itself and simulations using it are all written in the Python programming language, which is an easy, concise and highly developed language with many advanced features and development tools, excellent documentation and a large community of users providing support and extension packages.
Goodman D, 2006, Spirals in the boundary of slices of quasi-fuchsian space, Conformal Geometry and Dynamics, Vol: 10, Pages: 136-158
We prove that the Bers and Maskit slices of the quasi-Fuchsian space of a once-punctured torus have a dense, uncountable set of points in their boundaries about which the boundary spirals infinitely. © 2006 American Mathematical Society.
Kutzscher EM, Sorenson AL, Goodman DF, 2004, Penetrating keratoplasty performed by residents., Arch Ophthalmol, Vol: 122, Pages: 1333-1336, ISSN: 0003-9950
OBJECTIVE: To report the results of penetrating keratoplasty performed by residents. METHOD: A retrospective medical record review of all patients undergoing penetrating keratoplasty performed by residents at our institution from April 1998 to April 2002. RESULTS: Forty penetrating keratoplasty procedures were performed by 8 residents. The most common indication was keratoconus (17 eyes [43%]), followed by corneal scarring (14 eyes [35%]). Mean preoperative best-corrected visual acuity was 20/250. No intraoperative complications were reported. Mean follow-up time was 15 months. Postoperatively, mean best-corrected visual acuity was 20/40, mean postoperative astigmatism was 3.4 +/- 2.1 diopters, and graft survival was 92.5%. Postoperative complications included elevated intraocular pressure, wound dehiscence, and endophthalmitis. MAIN OUTCOME MEASURES: Best-corrected visual acuity, postoperative astigmatism, graft survival, and intraoperative and postoperative complications. CONCLUSION: Residents can be introduced to penetrating keratoplasty and achieve surgical success with intraoperative and postoperative complication rates similar to those previously published.
Fine I, Wade AR, Brewer AA, et al., 2003, Long-term deprivation affects visual perception and cortex., Nat Neurosci, Vol: 6, Pages: 915-916, ISSN: 1097-6256
Akpek EK, Jun AS, Goodman DF, et al., 2002, Clinical and ultrastructural features of a novel hereditary anterior segment dysgenesis., Ophthalmology, Vol: 109, Pages: 513-519, ISSN: 0161-6420
OBJECTIVE: To describe the clinical, histopathologic, and hereditary features of a novel familial anterior segment dysgenesis. DESIGN: Prospective, observational case series and interventional case report. PARTICIPANTS: Ten individuals from three generations of a single family with iris and corneal abnormalities associated with congenital cataracts. MAIN OUTCOME MEASURES: An ophthalmic evaluation including slit-lamp examination, corneal topography, pachymetry, and specular biomicroscopy of all family members, and histopathologic and ultrastructural evaluation of one excised corneal button. RESULTS: The proband was an 81-year-old man with bilateral aphakia and diffuse corneal haze, and thinning associated with corneal guttae. His pupils were small, mildly eccentric, and difficult to dilate. Pachymeter readings were 335 microm (right eye) and 330 microm (left eye). Topography confirmed advanced steepening of both corneas. Light microscopic and transmission electron microscopic examinations of the corneal button revealed an attenuated endothelium with prominent intracellular random aggregates of small-diameter filaments staining positively for cytokeratin. Descemet's membrane was thickened and had marked posterior nodularity. Various-sized polymorphic vacuoles containing layered electron-dense material were present within and between collagen lamellae and within keratocytes throughout the stroma and Bowman's membrane. Secondary bullous changes of the epithelium with thickening of the basement membrane were also observed. The family pedigree demonstrated an autosomal dominant inheritance pattern. CONCLUSIONS: This constellation of autosomal dominantly inherited corneal endothelial and stromal disorder, with congenital cataracts and iris abnormalities, represents a novel anterior segment disorder. Its etiology may involve an abnormal migration of the secondary mesenchyme.
Levartovsky S, Rosenwasser G, Goodman D, 2001, Bacterial keratitis after [correction of following] laser in situ keratomileusis., Ophthalmology, Vol: 108, Pages: 321-325, ISSN: 0161-6420
PURPOSE: The development of bacterial keratitis after laser in situ keratomileusis (LASIK) has been described in only a few isolated cases. We report the development of bacterial keratitis as a postoperative complication of LASIK in three subjects. DESIGN: A retrospective interventional small case series. PARTICIPANTS: Three patients who underwent LASIK for correction of myopia during July and August 1998 and had bacterial keratitis develop after surgery. METHODS: Bacterial keratitis was encountered in the operated eyes between 1 and 22 days after surgery. Topical antibiotic therapy was administered. In one eye, which had significant opacification and irregularity of the flap was developed, lamellar keratoplasty was performed. MAIN OUTCOME MEASURES: Postoperative inflammation was followed clinically and photographically until it resolved. Visual acuity was measured at intervals throughout the follow-up period. RESULTS: Keratitis resolved within 3 to 16 days of starting antibiotic therapy. The final best-corrected visual acuities were 20/30, 20/15, and 20/25. CONCLUSIONS: Although infectious keratitis after LASIK is an infrequently reported event, it should be discussed preoperatively with patients as a possible complication.
Gottsch JD, Liu SH, Minkovitz JB, et al., 1995, Autoimmunity to a cornea-associated stromal antigen in patients with Mooren's ulcer., Invest Ophthalmol Vis Sci, Vol: 36, Pages: 1541-1547, ISSN: 0146-0404
PURPOSE: To purify and characterize a cornea-associated antigen (CO-Ag) and to determine antibody levels to CO-Ag in patients with Mooren's ulcer. METHOD: Standard ion exchange and gel filtration chromatographies were used to isolate and purify CO-Ag from crude bovine stromal extracts. The serum of a patient with Mooren's ulcer, containing a high level of antibodies directed against CO-Ag, was used to monitor isolation procedures. Using this newly purified CO-Ag, an enzyme-linked immunoabsorbent assay was used to detect the presence of antibodies to CO-Ag in the sera of other patients with Mooren's ulcer. RESULTS: CO-Ag was purified to apparent homogeneity from bovine corneal stromal extracts by a series of ion exchange chromatographies and gel filtration. Polyacrylamide gel electrophoresis showed that CO-Ag was a tetramer with a molecular weight of 30,000 d that may dissociate under denaturing conditions into a monomer of 7000 d. Strong indirect immunofluorescent staining was demonstrated of the stroma by guinea pig anti-CO-Ag antibody. A statistically significant difference in the level of specific antibodies to CO-Ag between patients with Mooren's ulcer and controls was found (P < 0.001). The antibody level was elevated in patients with Mooren's ulcer (mean antibody level, 0.58 +/- 0.13) compared with the controls (mean antibody level, 0.22 +/- 0.04). CONCLUSION: These results suggest that an autoantigen exists in the corneal stroma that reacts with serum antibodies from patients with Mooren's ulcer. The availability of a purified corneal antigen could facilitate the diagnosis and define the pathogenetic mechanisms in Mooren's ulcer.
Olkowski S, Stark WJ, Gottsch JD, et al., 1991, A conservative view of radial keratotomy., Yan Ke Xue Bao, Vol: 7, Pages: 130-139, ISSN: 1000-4432
Talamo JH, Stark WJ, Gottsch JD, et al., 1991, Natural history of corneal astigmatism after cataract surgery., J Cataract Refract Surg, Vol: 17, Pages: 313-318, ISSN: 0886-3350
Little information on the natural course of corneal astigmatism following cataract surgery exists. We report a prospective, computerized analysis of postoperative astigmatism, based on keratometry measurements, of 137 cases of extracapsular cataract extraction with intraocular lens implantation performed by one surgeon. No sutures were cut postoperatively. Surgery induced 1.44 diopters (D) of with-the-rule astigmatism at one month, which declined at a rate of 0.77 D and 0.35 D per month for the next two months, respectively, with a more gradual decline thereafter. The mean surgically induced astigmatism at the last postoperative visit ranged from 0.29 D at six months (minimum follow-up) to 1.23 D at 48 months; both were against-the-rule. Mean follow-up was 28.92 months. These findings may be technique specific and suggest that (1) corneal curvature continues to change slowly even two to four years postoperatively; (2) most patients develop against-the-rule astigmatism, thus more with-the-rule astigmatism is desirable in the early postoperative period; (3) selective suture removal is necessary only when significantly more than 3.00 D of surgically induced with-the-rule astigmatism is present.
Gottsch JD, Gilbert ML, Goodman DF, et al., 1991, Excimer laser ablative treatment of microbial keratitis., Ophthalmology, Vol: 98, Pages: 146-149, ISSN: 0161-6420
The 193-nm excimer laser was used to ablate experimental septate fungal (Fusarium) and an atypical mycobacterial (Mycobacterium fortuitum) keratitis in an animal model. The infections were allowed to proceed for 24 and 72 hours. After incubation, ablation with a 193-nm excimer laser with 5.0-mm treatment zones was performed until all suppurative areas were treated. The corneas were excised, halved, homogenized, and plated. All cultures were negative in the 24-hour group. However, in those corneas in which the infections were allowed to proceed to 72 hours, post-treatment cultures were positive for both organisms. Histopathologic examination confirmed that 24-hour infections had been eradicated and that 72-hour infections had organisms present. Three of the eight eyes in the M. fortuitum group perforated during treatment, even though the treatment depth by computer preselection was only 150 microns. Excimer laser photoablation may be a useful technique to eradicate early, localized microbial infections. However, it is apparent that advanced infections with deep stromal involvement and suppuration cannot be eradicated using this technique. Because corneas may be perforated inadvertently during treatment, excimer laser treatment of infectious keratitis should be approached with caution and used for superficial and well circumscribed lesions.
This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.