Modeling High-Dimensional Audio Sequences with Recurrent Neural Networks

Cette thèse étudie des modèles de séquences de haute dimension basés sur des réseaux de neurones récurrents (RNN) et leur application à la musique et à la parole. Bien qu'en principe les RNN puissent représenter les dépendances à long terme et la dynamique temporelle complexe propres aux séquences d'intérêt comme la vidéo, l'audio et la langue naturelle, ceux-ci n'ont pas été utilisés à leur plein potentiel depuis leur introduction par Rumelhart et al. (1986a) en raison de la difficulté de les entraîner efficacement par descente de gradient. Récemment, l'application fructueuse de l'optimisation Hessian-free et d'autres techniques d'entraînement avancées ont entraîné la recrudescence de leur utilisation dans plusieurs systèmes de l'état de l'art. Le travail de cette thèse prend part à ce développement. L'idée centrale consiste à exploiter la flexibilité des RNN pour apprendre une description probabiliste de séquences de symboles, c'est-à-dire une information de haut niveau associée aux signaux observés, qui en retour pourra servir d'à priori pour améliorer la précision de la recherche d'information. Par exemple, en modélisant l'évolution de groupes de notes dans la musique polyphonique, d'accords dans une progression harmonique, de phonèmes dans un énoncé oral ou encore de sources individuelles dans un mélange audio, nous pouvons améliorer significativement les méthodes de transcription polyphonique, de reconnaissance d'accords, de reconnaissance de la parole et de séparation de sources audio respectivement. L'application pratique de nos modèles à ces tâches est détaillée dans les quatre derniers articles présentés dans cette thèse. Dans le premier article, nous remplaçons la couche de sortie d'un RNN par des machines de Boltzmann restreintes conditionnelles pour décrire des distributions de sortie multimodales beaucoup plus riches. Dans le deuxième article, nous évaluons et proposons des méthodes avancées pour entraîner les RNN. Dans les quatre derniers articles, nous examinons différentes façons de combiner nos modèles symboliques à des réseaux profonds et à la factorisation matricielle non-négative, notamment par des produits d'experts, des architectures entrée/sortie et des cadres génératifs généralisant les modèles de Markov cachés. Nous proposons et analysons également des méthodes d'inférence efficaces pour ces modèles, telles la recherche vorace chronologique, la recherche en faisceau à haute dimension, la recherche en faisceau élagué et la descente de gradient. Finalement, nous abordons les questions de l'étiquette biaisée, du maître imposant, du lissage temporel, de la régularisation et du pré-entraînement.

A spectral performance model for the High Resolution Dynamics Limb Sounder (6-18µm)

A spectral performance model, designed to simulate the system spectral throughput for each of the 21 channels in the HIRDLS radiometer, is described. This model uses the measured spectral characteristics of each of the components in the optical train, appropriately corrected for their optical environment, to determine the end-to-end spectral throughput profile for each channel. This profile is then combined with the predicted thermal emission from the atmosphere, arising from the height of interest, to establish an in-band (wanted) to out-of-band (unwanted) radiance ratio. The results from the use of the model demonstrate that the instrument level radiometric requirements for the instrument will be achieved. The optical arrangement and spectral design requirements for filtering in the HIRDLS instrument are described together with a presentation of the performance achieved for the complete set of manufactured filters. Compliance of the predicted passband throughput model to the spectral positioning requi rements of the instrument is also demonstrated.

Ultra High Temperatures #1

Ultra High Temperature #1, initiated by Rebecca Bibby forms the first in an ongoing project which explores the realms of collaboration, performance, writing and publication as artistic vehicle of production, dispersion and progression. With Bibby's text -that re-fictions the futuristic projections of technosexuality in Metropolis (1927)- at its core was launched, printed, compiled and distributed in a live performance by POLLYFIBRE at Eastside Projects in Birmingham. The limited edition printed publication was designed by An Endless Supply whose Risograph stencil printer was used as an instrument in the performed production of the text. As a crude avatar of Rebecca Bibby’s practice, Aikon-II, a mechanically programmed signature machine automatically signed each copy of the text during the performance. POLLYFIBRE's ‘flat-pack’ costumes were on display throughout the duration of the exhibition. POLLYFIBRE is a performance project created by Christine Ellison.

High dosage folic acid supplementation, oral cleft recurrence and fetal growth

Objectives: To evaluate the effects of folic acid supplementation on isolated oral cleft recurrence and fetal growth. Patients and Methods: The study included 2,508 women who were at-risk for oral cleft recurrence and randomized into two folic acid supplementation groups: 0.4 and 4 mg per day before pregnancy and throughout the first trimester. The infant outcome data were based on 234 live births. In addition to oral cleft recurrence, several secondary outcomes were compared between the two folic acid groups. Cleft recurrence rates were also compared to historic recurrence rates. Results: The oral cleft recurrence rates were 2.9% and 2.5% in the 0.4 and 4 mg groups, respectively. The recurrence rates in the two folic acid groups both separately and combined were significantly different from the 6.3% historic recurrence rate post the folic acid fortification program for this population (p = 0.0009 when combining the two folic acid groups). The rate of cleft lip with palate recurrence was 2.9% in the 0.4 mg group and 0.8% in the 4 mg group. There were no elevated fetal growth complications in the 4 mg group compared to the 0.4 mg group. Conclusions: The study is the first double-blinded randomized clinical trial (RCT) to study the effect of high dosage folic acid supplementation on isolated oral cleft recurrence. The recurrence rates were similar between the two folic acid groups. However, the results are suggestive of a decrease in oral cleft recurrence compared to the historic recurrence rate. A RCT is still needed to identify the effect of folic acid on oral cleft recurrence given these suggestive results and the supportive results from previous interventional and observational studies, and the study offers suggestions for such future studies. The results also suggest that high dosage folic acid does not compromise fetal growth. © 2013 by the authors; licensee MDPI, Basel, Switzerland.

Maternal obesity in high-risk pregnancies and postpartum infectious complications

Objective: To analyze the association between maternal obesity and postnatal infectious complications in high-risk pregnancies. Methods: Prospective study from August 2009 through August 2010 with the following inclusion criteria: women up to the 5th postpartum day; age L 18 years; high-risk pregnancy; singleton pregnancy with live fetus at labor onset; delivery at the institution; maternal weight measured on day of delivery. The nutritional status in late pregnancy was assessed by the body mass index (BMI), with the application of the Atalah et al. curve. Patients were graded as underweight, adequate weight, overweight, or obese. Postpartum complications investigated during the hospital stay and 30 days post-discharge were: surgical wound infection and/or secretion, urinary infection, postpartum infection, fever, hospitalization, antibiotic use, and composite morbidity (at least one of the complications mentioned). Results: 374 puerperal women were included, graded according to the final BMI as: underweight (n = 54, 14.4%); adequate weight (n = 126, 33.7%); overweight (n = 105, 28.1%); and obese (n = 89, 23.8%). Maternal obesity was shown to have a significant association with the following postpartum complications: surgical wound infection (16.7%, p = 0.042), urinary infection (9.0%, p = 0.004), antibiotic use (12.3%, p < 0.001), and composite morbidity (25.6%, p = 0.016). By applying the logistic regression model, obesity in late pregnancy was found to be an independent variable regardless of the composite morbidity predicted (OR: 2.09; 95% CI: 1.15-3.80, p = 0.015). Conclusion: Maternal obesity during late pregnancy in high-risk patients is independently associated with postpartum infectious complications, which demonstrates the need for a closer follow-up of maternal weight gain in these pregnancies.

High dosage folic acid supplementation, oral cleft recurrence and fetal growth

Objectives: To evaluate the effects of folic acid supplementation on isolated oral cleft recurrence and fetal growth. Patients and Methods: The study included 2,508 women who were at-risk for oral cleft recurrence and randomized into two folic acid supplementation groups: 0.4 and 4 mg per day before pregnancy and throughout the first trimester. The infant outcome data were based on 234 live births. In addition to oral cleft recurrence, several secondary outcomes were compared between the two folic acid groups. Cleft recurrence rates were also compared to historic recurrence rates. Results: The oral cleft recurrence rates were 2.9% and 2.5% in the 0.4 and 4 mg groups, respectively. The recurrence rates in the two folic acid groups both separately and combined were significantly different from the 6.3% historic recurrence rate post the folic acid fortification program for this population (p = 0.0009 when combining the two folic acid groups). The rate of cleft lip with palate recurrence was 2.9% in the 0.4 mg group and 0.8% in the 4 mg group. There were no elevated fetal growth complications in the 4 mg group compared to the 0.4 mg group. Conclusions: The study is the first double-blinded randomized clinical trial (RCT) to study the effect of high dosage folic acid supplementation on isolated oral cleft recurrence. The recurrence rates were similar between the two folic acid groups. However, the results are suggestive of a decrease in oral cleft recurrence compared to the historic recurrence rate. A RCT is still needed to identify the effect of folic acid on oral cleft recurrence given these suggestive results and the supportive results from previous interventional and observational studies, and the study offers suggestions for such future studies. The results also suggest that high dosage folic acid does not compromise fetal growth

Applications of High Speed Configurable Logic Devices in Modern Particle Physics Experiments

Several activities were conducted during my PhD activity. For the NEMO experiment a collaboration between the INFN/University groups of Catania and Bologna led to the development and production of a mixed signal acquisition board for the Nemo Km3 telescope. The research concerned the feasibility study for a different acquisition technique quite far from that adopted in the NEMO Phase 1 telescope. The DAQ board that we realized exploits the LIRA06 front-end chip for the analog acquisition of anodic an dynodic sources of a PMT (Photo-Multiplier Tube). The low-power analog acquisition allows to sample contemporaneously multiple channels of the PMT at different gain factors in order to increase the signal response linearity over a wider dynamic range. Also the auto triggering and self-event-classification features help to improve the acquisition performance and the knowledge on the neutrino event. A fully functional interface towards the first level data concentrator, the Floor Control Module, has been integrated as well on the board, and a specific firmware has been realized to comply with the present communication protocols. This stage of the project foresees the use of an FPGA, a high speed configurable device, to provide the board with a flexible digital logic control core. After the validation of the whole front-end architecture this feature would be probably integrated in a common mixed-signal ASIC (Application Specific Integrated Circuit). The volatile nature of the configuration memory of the FPGA implied the integration of a flash ISP (In System Programming) memory and a smart architecture for a safe remote reconfiguration of it. All the integrated features of the board have been tested. At the Catania laboratory the behavior of the LIRA chip has been investigated in the digital environment of the DAQ board and we succeeded in driving the acquisition with the FPGA. The PMT pulses generated with an arbitrary waveform generator were correctly triggered and acquired by the analog chip, and successively they were digitized by the on board ADC under the supervision of the FPGA. For the communication towards the data concentrator a test bench has been realized in Bologna where, thanks to a lending of the Roma University and INFN, a full readout chain equivalent to that present in the NEMO phase-1 was installed. These tests showed a good behavior of the digital electronic that was able to receive and to execute command imparted by the PC console and to answer back with a reply. The remotely configurable logic behaved well too and demonstrated, at least in principle, the validity of this technique. A new prototype board is now under development at the Catania laboratory as an evolution of the one described above. This board is going to be deployed within the NEMO Phase-2 tower in one of its floors dedicated to new front-end proposals. This board will integrate a new analog acquisition chip called SAS (Smart Auto-triggering Sampler) introducing thus a new analog front-end but inheriting most of the digital logic present in the current DAQ board discussed in this thesis. For what concern the activity on high-resolution vertex detectors, I worked within the SLIM5 collaboration for the characterization of a MAPS (Monolithic Active Pixel Sensor) device called APSEL-4D. The mentioned chip is a matrix of 4096 active pixel sensors with deep N-well implantations meant for charge collection and to shield the analog electronics from digital noise. The chip integrates the full-custom sensors matrix and the sparsifification/readout logic realized with standard-cells in STM CMOS technology 130 nm. For the chip characterization a test-beam has been set up on the 12 GeV PS (Proton Synchrotron) line facility at CERN of Geneva (CH). The collaboration prepared a silicon strip telescope and a DAQ system (hardware and software) for data acquisition and control of the telescope that allowed to store about 90 million events in 7 equivalent days of live-time of the beam. My activities concerned basically the realization of a firmware interface towards and from the MAPS chip in order to integrate it on the general DAQ system. Thereafter I worked on the DAQ software to implement on it a proper Slow Control interface of the APSEL4D. Several APSEL4D chips with different thinning have been tested during the test beam. Those with 100 and 300 um presented an overall efficiency of about 90% imparting a threshold of 450 electrons. The test-beam allowed to estimate also the resolution of the pixel sensor providing good results consistent with the pitch/sqrt(12) formula. The MAPS intrinsic resolution has been extracted from the width of the residual plot taking into account the multiple scattering effect.

Length scales of dynamic heterogeneities of low and high molecular weight glass formers from multidimensional NMR

Das Ziel der vorliegenden Arbeit ist die Untersuchung der räumlichen und zeitlichen Aspekte der heterogenen Dynamik in Modellglasbildnern. Dabei wird vor allem die langsame alpha-Relaxationsdynamik oberhalb des Glasüberganges Tg untersucht. Die nukleare Magnetresonanz zeigt ihre einmalige Vielseitigkeit bei der Untersuchung molekularer Dynamik, wenn die angewandten Techniken und Experimente durch Simulationen unterstützt werden. Die räumliche Aspekt dynamischer Heterogenitäten wird untersucht durch ein reduziertes vierdimensionales Spindiffusionsexperiment (4D3CP), ein Experiment, das Reorientierungsraten örtlich korreliert. Eine Simulation dieses Experimentes an einem System harter Kugeln liefert wertvolle Informationen über die Auswertemethode des 4D3CP Experiments. Glycerol und o-terphenyl werden durch das 4D3CP Experiment untersucht. Die erhaltenen Resultate werden mit bereits publizierten Daten des polymeren Glasbildners PVAc verglichen. Während PVAc und o-terphenyl eine Längenskale von 3.7 nm bzw. 2.9 nm aufweisen, ist die Längenskale von Glycerol signifikant kleiner bei 1.1 nm. Ein neues Experiment, welches sensitiv auf Translationsbewegung reagiert, wird vorgestellt. Durch Verwendung eines pi-Impulszuges kann eine separate Evolution unter dem Hamiltonian der dipolaren Kopplung und der chemischen Verschiebungsanisotropie erreicht werden.

X-ray properties and evolution of high-redshift AGN, and the gas content of host galaxies

In this thesis, I have investigated the evolution of the high-redshift (z > 3) AGN population by collecting data from some of the major Chandra and XMM-Newton surveys. The final sample (141 sources) is one of the largest selected at z> 3 in the X- rays and it is characterised by a very high redshift completeness (98%). I derived the spectral slopes and obscurations through a spectral anaysis and I assessed the high-z evolution by deriving the luminosity function and the number counts of the sample. The best representation of the AGN evolution is a pure density evolution (PDE) model: the AGN space density is found to decrease by a factor of 10 from z=3 to z=5. I also found that about 50% of AGN are obscured by large column densities (logNH > 23). By comparing these data with those in the Local Universe, I found a positive evolution of the obscured AGN fraction with redshift, especially for luminous (logLx > 44) AGN. I also studied the gas content of z < 1 AGN-hosting galaxies and compared it with that of inactive galaxies. For the first time, I applied to AGN a method to derive the gas mass previously used for inactive galaxies only. AGN are found to live preferentially in gas-rich galaxies. This result on the one hand can help us in understanding the AGN triggering mechanisms, on the other hand explains why AGN are preferentially hosted by star-forming galaxies.

Loss of the high-affinity pentamidine transporter is responsible for high levels of cross-resistance between arsenical and diamidine drugs in African trypanosomes

Treatment of many infectious diseases is under threat from drug resistance. Understanding the mechanisms of resistance is as high a priority as the development of new drugs. We have investigated the basis for cross-resistance between the diamidine and melaminophenyl arsenical classes of drugs in African trypanosomes. We induced high levels of pentamidine resistance in a line without the tbat1 gene that encodes the P2 transporter previously implicated in drug uptake. We isolated independent clones that displayed very considerable cross-resistance with melarsen oxide but not phenylarsine oxide and reduced uptake of [(3)H]pentamidine. In particular, the high-affinity pentamidine transport (HAPT1) activity was absent in the pentamidine-adapted lines, whereas the low affinity pentamidine transport (LAPT1) activity was unchanged. The parental tbat1(-/-) line was sensitive to lysis by melarsen oxide, and this process was inhibited by low concentrations of pentamidine, indicating the involvement of HAPT1. This pentamidine-inhibitable lysis was absent in the adapted line KO-B48. Likewise, uptake of the fluorescent diamidine 4',6-diamidino-2-phenylindole dihydrochloride was much delayed in live KO-B48 cells and insensitive to competition with up to 10 muM pentamidine. No overexpression of the Trypanosoma brucei brucei ATP-binding cassette transporter TbMRPA could be detected in KO-B48. We also show that a laboratory line of Trypanosoma brucei gambiense, adapted to high levels of resistance for the melaminophenyl arsenical drug melarsamine hydrochloride (Cymelarsan), had similarly lost TbAT1 and HAPT1 activity while retaining LAPT1 activity. It seems therefore that selection for resistance to either pentamidine or arsenical drugs can result in a similar phenotype of reduced drug accumulation, explaining the occurrence of cross-resistance.

Pseudovacuoles--immobilized by high-pressure freezing--are associated with blebbing in walker carcinosarcoma cells

By applying high pressure freezing and freeze-substitution, we observed large inclusions of homogeneous appearance in the front of locomoting Walker carcinosarcoma cells that have not been described earlier. Live cell imaging revealed that these inclusions were poor in lipids and nucleic acids but had a high lysine (and hence protein) content. Usually one such structure 2-5 mum in size was present at the front of motile Walker cells, predominantly in the immediate vicinity of newly forming blebs. By correlating the lysine-rich areas in fixed and embedded cells with electron microscopic pictures, inclusions could be assigned to confined, faintly stained cytoplasmic areas that lacked a surrounding membrane; they were therefore called pseudovacuoles. After high-pressure freezing and freeze substitution, pseudovacuoles appeared to be filled with 20 nm large electron-transparent patches surrounded by 12 and 15 nm large particles. The heat shock protein Hsp90 was identified by peptide sequencing as a major fluorescent band on SDS-PAGE of lysine-labelled Walker cell extracts. By immunofluorescence, Hsp90 was found to be enriched in pseudovacuoles. Colocalization of the lysine with a potassium-specific dye in living cells revealed that pseudovacuoles act as K+ stores in the vicinity of forming blebs. We propose that pseudovacuoles might support blebbing by locally regulating the intracellular hydrostatic pressure.

AN ASSESSMENT OF BALLAST SOURCE'S QUALITY AND LOCATIONS WITH RESPECT TO MDOT'S UPGRADE OF THE WOLVERINE RAILROAD LINE FOR HIGH SPEED RAIL USEAGE

The Michigan Department of Transportation is evaluating upgrading their portion of the Wolverine Line between Chicago and Detroit to accommodate high speed rail. This will entail upgrading the track to allow trains to run at speeds in excess of 110 miles per hour (mph). An important component of this upgrade will be to assess the requirement for ballast material for high speed rail. In the event that the existing ballast materials do not meet specifications for higher speed train, additional ballast will be required. The purpose of this study, therefore, is to investigate the current MDOT railroad ballast quality specifications and compare them to both the national and international specifications for use on high speed rail lines. The study found that while MDOT has quality specifications for railroad ballast it does not have any for high speed rail. In addition, the American Railway Engineering and Maintenance-of-Way Association (AREMA), while also having specifications for railroad ballast, does not have specific specifications for high speed rail lines. The AREMA aggregate specifications for ballast include the following tests: (1) LA Abrasion, (2) Percent Moisture Absorption, (3) Flat and Elongated Particles, (4) Sulfate Soundness test. Internationally, some countries do require a highly standard for high speed rail such as the Los Angeles (LA) Abrasion test, which is uses a higher standard performance and the Micro Duval test, which is used to determine the maximum speed that a high speed can operate at. Since there are no existing MDOT ballast specification for high speed rail, it is assumed that aggregate ballast specifications for the Wolverine Line will use the higher international specifications. The Wolverine line, however, is located in southern Michigan is a region of sedimentary rocks which generally do not meet the existing MDOT ballast specifications. The investigation found that there were only 12 quarries in the Michigan that meet the MDOT specification. Of these 12 quarries, six were igneous or metamorphic rock quarries, while six were carbonate quarries. Of the six carbonate quarries four were locate in the Lower Peninsula and two in the Upper Peninsula. Two of the carbonate quarries were located in near proximity to the Wolverine Line, while the remaining quarries were at a significant haulage distance. In either case, the cost of haulage becomes an important consideration. In this regard, four of the quarries were located with lake terminals allowing water transportation to down state ports. The Upper Peninsula also has a significant amount of metal based mining in both igneous and metamorphic rock that generate significant amount of waste rock that could be used as a ballast material. The main drawback, however, is the distance to the Wolverine rail line. One potential source is the Cliffs Natural Resources that operates two large surface mines in the Marquette area with rail and water transportation to both Lake Superior and Lake Michigan. Both mines mine rock with a very high compressive strength far in excess of most ballast materials used in the United States and would make an excellent ballast materials. Discussions with Cliffs, however, indicated that due to environmental concerns that they would most likely not be interested in producing a ballast material. In the United States carbonate aggregates, while used for ballast, many times don't meet the ballast specifications in addition to the problem of particle degradation that can lead to fouling and cementation issues. Thus, many carbonate aggregate quarries in close proximity to railroads are not used. Since Michigan has a significant amount of carbonate quarries, the research also investigated using the dynamic properties of aggregate as a possible additional test for aggregate ballast quality. The dynamic strength of a material can be assessed using a split Hopkinson Pressure Bar (SHPB). The SHPB has been traditionally used to assess the dynamic properties of metal but over the past 20 years it is now being used to assess the dynamic properties of brittle materials such as ceramics and rock. In addition, the wear properties of metals have been related to their dynamic properties. Wear or breakdown of railroad ballast materials is one of the main problems with ballast material due to the dynamic loading generated by trains and which will be significantly higher for high speed rails. Previous research has indicated that the Port Inland quarry along Lake Michigan in the Southern Upper Peninsula has significant dynamic properties that might make it potentially useable as an aggregate for high speed rail. The dynamic strength testing conducted in this research indicate that the Port Inland limestone in fact has a dynamic strength close to igneous rocks and much higher than other carbonate rocks in the Great Lakes region. It is recommended that further research be conducted to investigate the Port Inland limestone as a high speed ballast material.

Infection prevention strategies in a stem cell transplant unit: impact of change of care in isolation practice and routine use of high dose intravenous immunoglobulins on infectious complications and transplant related mortality

OBJECTIVE: Nursing in 'live islands' and routine high dose intravenous immunoglobulins after allogeneic hematopoietic stem cell transplantation were abandoned by many teams in view of limited evidence and high costs. METHODS: This retrospective single-center study examines the impact of change from nursing in 'live islands' to care in single rooms (SR) and from high dose to targeted intravenous immunoglobulins (IVIG) on mortality and infection rate of adult patients receiving an allogeneic stem cell or bone marrow transplantation in two steps and three time cohorts (1993-1997, 1997-2000, 2000-2003). RESULTS: Two hundred forty-eight allogeneic hematopoetic stem cell transplantations were performed in 227 patients. Patient characteristics were comparable in the three cohorts for gender, median age, underlying disease, and disease stage, prophylaxis for graft versus host disease (GvHD) and cytomegalovirus constellation. The incidence of infections (78.4%) and infection rates remained stable (rates/1000 days of neutropenia for sepsis 17.61, for pneumonia 6.76). Cumulative incidence of GvHD and transplant-related mortality did not change over time. CONCLUSIONS: Change from nursing in 'live islands' to SR and reduction of high dose to targeted IVIG did not result in increased infection rates or mortality despite an increase in patient age. These results support the current practice.

Effect of Optimum® High Oil Corn on Performance and Carcass Characteristics of Finishing Yearling Steers

Steers fed Optimum® high oil corn had statistically similar live performance as steers fed isogenetic control corn or the control corn + fat. Numerically steers fed high oil corn gained 3% faster during the 107-day study with similar feed conversion. During the first half of the experiment, steers fed high oil corn did not perform as well as those fed control corn. During the second half of the experiment, steers fed high oil corn gained 21% faster and were 17% more efficient. There were no effects of feeding high oil corn on carcass characteristics, except there were more Choice carcasses from the steers fed high oil corn as compared with control corn (57% vs. 43% Choice).

High frequency repetitive transcranial magnetic stimulation (rTMS) of the left dorsolateral cortex: EEG topography during waking and subsequent sleep.

Repetitive transcranial magnetic stimulation (rTMS) is a novel research tool in neurology and psychiatry. It is currently being evaluated as a conceivable alternative to electroconvulsive therapy for the treatment of mood disorders. Eight healthy young (age range 21-25 years) right-handed men without sleep complaints participated in the study. Two sessions at a 1-week interval, each consisting of an adaptation night (sham stimulation) and an experimental night (rTMS in the left dorsolateral prefrontal cortex or sham stimulation; crossover design), were scheduled. In each subject, 40 trains of 2-s duration of rTMS (inter-train interval 28 s) were applied at a frequency of 20 Hz (i.e. 1600 pulses per session) and at an intensity of 90% of the motor threshold. Stimulations were scheduled 80 min before lights off. The waking EEG was recorded for 10-min intervals approximately 30 min prior to and after the 20-min stimulations, and polysomnographic recordings were obtained during the subsequent sleep episode (23.00-07.00 h). The power spectra of two referential derivations, as well as of bipolar derivations along the antero-posterior axis over the left and right hemispheres, were analyzed. rTMS induced a small reduction of sleep stage 1 (in min and percentage of total sleep time) over the whole night and a small enhancement of sleep stage 4 during the first non-REM sleep episode. Other sleep variables were not affected. rTMS of the left dorsolateral cortex did not alter the topography of EEG power spectra in waking following stimulation, in the all-night sleep EEG, or during the first non-REM sleep episode. Our results indicate that a single session of rTMS using parameters like those used in depression treatment protocols has no detectable side effects with respect to sleep in young healthy males.