Showing 1–27 of 27 results for author: Venkatesan, M

Ru$_{2-x}$Mn$_{1+x}$Al thin films

Authors: K. E. Siewierska, H. Kurt, B. Shortall, A. Jha, N. Teichert, G. Atcheson, M. Venkatesan, J. M. D. Coey, Z. Gercsi, K. Rode

Abstract: The cubic Heusler alloy Ru$_{2-x}$Mn$_{1+x}$Al is grown in thin film form on MgO and MgAl$_2$O$_4$ substrates. It is a highly spin-polarised ferrimagnetic metal, with weak magnetocrystalline anisotropy. Although structurally and chemically similar to $\text{Mn}_2\text{Ru}_x\text{Ga}$, it does not exhibit ferrimagnetic compensation, or large magneto galvanic effects. The differences are attributed… ▽ More The cubic Heusler alloy Ru$_{2-x}$Mn$_{1+x}$Al is grown in thin film form on MgO and MgAl$_2$O$_4$ substrates. It is a highly spin-polarised ferrimagnetic metal, with weak magnetocrystalline anisotropy. Although structurally and chemically similar to $\text{Mn}_2\text{Ru}_x\text{Ga}$, it does not exhibit ferrimagnetic compensation, or large magneto galvanic effects. The differences are attributed to a combination of atomic order and the hybridisation with the group 13 element Al or Ga. The spin polarisation is around 50 to 60 %. There is a gap in the density of states just above the Fermi level in fully ordered compounds. △ Less

Submitted 30 January, 2023; v1 submitted 24 January, 2023; originally announced January 2023.

Comments: 8 pages , 7 figures, 2 tables

Evaporation of water and urea solution in a magnetic field; the role of nuclear isomers

Authors: Sruthy Poulose, M. Venkatesan, Matthias Mobius, J. M. D. Coey

Abstract: Hypothesis. Ortho and para water are the two nuclear isomers where the hydrogen protons align to give a total nuclear spin of 0 or 1.The equilibrium ratio of 3:1 is established slowly in freshly evaporated water vapour and the isomers then behave distinct gasses, with their own partial pressures. Magnetic-field-induced ortho to para transformations are expected to alter the evaporation rate. Exp… ▽ More Hypothesis. Ortho and para water are the two nuclear isomers where the hydrogen protons align to give a total nuclear spin of 0 or 1.The equilibrium ratio of 3:1 is established slowly in freshly evaporated water vapour and the isomers then behave distinct gasses, with their own partial pressures. Magnetic-field-induced ortho to para transformations are expected to alter the evaporation rate. Experiment. Evaporation from beakers of deionized water and a 6 M solution of urea is monitored simultaneously for periods from 1 to 60 hours with and without a 500 mT magnetic field, while logging the ambient temperature and humidity. The balances with the two beakers are shielded in the same Perspex container. Many runs have been conducted over a two-year period. Findings. The evaporation rate of water is found to increase by 12 % of in the field but that of water in urea solution decreases by 28 6 %. Two effects are at play. One is dephasing of the Larmor precession of adjacent protons on a water molecule in a field gradient, which tends to equalize the isomer populations. The other is Lorentz stress on the moving charge dipole, which can increase the proportion of the ortho isomer. From analysis of the time and field dependence of the evaporation, we infer an ortho fraction of 39 % in fresh vapour from water and 60 % in fresh vapour from urea. △ Less

Submitted 28 October, 2022; originally announced October 2022.

Comments: 27 pages (9 Figures, 1 Table)

Evaluating the Performance of StyleGAN2-ADA on Medical Images

Authors: McKell Woodland, John Wood, Brian M. Anderson, Suprateek Kundu, Ethan Lin, Eugene Koay, Bruno Odisio, Caroline Chung, Hyunseon Christine Kang, Aradhana M. Venkatesan, Sireesha Yedururi, Brian De, Yuan-Mao Lin, Ankit B. Patel, Kristy K. Brock

Abstract: Although generative adversarial networks (GANs) have shown promise in medical imaging, they have four main limitations that impeded their utility: computational cost, data requirements, reliable evaluation measures, and training complexity. Our work investigates each of these obstacles in a novel application of StyleGAN2-ADA to high-resolution medical imaging datasets. Our dataset is comprised of… ▽ More Although generative adversarial networks (GANs) have shown promise in medical imaging, they have four main limitations that impeded their utility: computational cost, data requirements, reliable evaluation measures, and training complexity. Our work investigates each of these obstacles in a novel application of StyleGAN2-ADA to high-resolution medical imaging datasets. Our dataset is comprised of liver-containing axial slices from non-contrast and contrast-enhanced computed tomography (CT) scans. Additionally, we utilized four public datasets composed of various imaging modalities. We trained a StyleGAN2 network with transfer learning (from the Flickr-Faces-HQ dataset) and data augmentation (horizontal flip** and adaptive discriminator augmentation). The network's generative quality was measured quantitatively with the Fréchet Inception Distance (FID) and qualitatively with a visual Turing test given to seven radiologists and radiation oncologists. The StyleGAN2-ADA network achieved a FID of 5.22 ($\pm$ 0.17) on our liver CT dataset. It also set new record FIDs of 10.78, 3.52, 21.17, and 5.39 on the publicly available SLIVER07, ChestX-ray14, ACDC, and Medical Segmentation Decathlon (brain tumors) datasets. In the visual Turing test, the clinicians rated generated images as real 42% of the time, approaching random guessing. Our computational ablation study revealed that transfer learning and data augmentation stabilize training and improve the perceptual quality of the generated images. We observed the FID to be consistent with human perceptual evaluation of medical images. Finally, our work found that StyleGAN2-ADA consistently produces high-quality results without hyperparameter searches or retraining. △ Less

Submitted 7 October, 2022; originally announced October 2022.

Comments: This preprint has not undergone post-submission improvements or corrections. The Version of Record of this contribution is published in LNCS, volume 13570, and is available online at https://doi.org/10.1007/978-3-031-16980-9_14

Journal ref: Lecture Notes in Computer Science 13570 (2022)

Magnetic reversal and pinning in a perpendicular zero moment half-metal

Authors: Niclas Teichert, Gwenael Atcheson, Katarzyna Siewierska, Marta Norah Sanz-Ortiz, M. Venkatesan, Karsten Rode, Solveig Felton, Plamen Stamenov, JMD Coey

Abstract: Compensated ferrimagnets are promising materials for fast spintronic applications based on domain wall motion as they combine the favourable properties of ferromagnets and antiferromagnets. They inherit from antiferromagnets immunity to external fields, fast spin dynamics and rapid domain wall motion. From ferromagnets they inherit straightforward ways to read out the magnetic state, especially in… ▽ More Compensated ferrimagnets are promising materials for fast spintronic applications based on domain wall motion as they combine the favourable properties of ferromagnets and antiferromagnets. They inherit from antiferromagnets immunity to external fields, fast spin dynamics and rapid domain wall motion. From ferromagnets they inherit straightforward ways to read out the magnetic state, especially in compensated half metals, where electrons flow in only one spin channel. Here, we investigate domain structure in compensated half-metallic Mn2Ru0.5Ga films and assess their potential in domain wall motion-based spin-electronic devices. Our focus is on understanding and reducing domain wall pinning in unpatterned epitaxial thin films. Two modes of magnetic reversal, driven by nucleation or domain wall motion, are identified for different thin film deposition temperatures $(T_{dep})$. The magnetic aftereffect is analysed to extract activation volumes $(V^*)$, activation energies $(E_A)$, and their variation $(ΔE_A)$. The latter is decisive for the magnetic reversal regime, where domain wall motion dominated reversal (weak pinning) is found for $ΔE_A<0.2$ eV and nucleation dominated reversal (strong pinning) for $ΔE_A>0.5$ eV. A minimum $ΔE_A=28$ meV is found for $T_{dep}=290°$C. Prominent pinning sites are visualized by analysing virgin domain patterns after thermal demagnetization. In the sample investigated they have spacings of order 300 nm, which gives an upper limit of the track-width of spin-torque domain-wall motion-based devices. △ Less

Submitted 17 December, 2020; originally announced December 2020.

Journal ref: Phys. Rev. Materials 5, 034408 (2021)

Deep Fence Estimation using Stereo Guidance and Adversarial Learning

Authors: Paritosh Mittal, Shankar M Venkatesan, Viswanath Veera, Aloknath De

Abstract: People capture memorable images of events and exhibits that are often occluded by a wire mesh loosely termed as fence. Recent works in removing fence have limited performance due to the difficulty in initial fence segmentation. This work aims to accurately segment fence using a novel fence guidance mask (FM) generated from stereo image pair. This binary guidance mask contains deterministic cues ab… ▽ More People capture memorable images of events and exhibits that are often occluded by a wire mesh loosely termed as fence. Recent works in removing fence have limited performance due to the difficulty in initial fence segmentation. This work aims to accurately segment fence using a novel fence guidance mask (FM) generated from stereo image pair. This binary guidance mask contains deterministic cues about the structure of fence and is given as additional input to the deep fence estimation model. We also introduce a directional connectivity loss (DCL), which is used alongside adversarial loss to precisely detect thin wires. Experimental results obtained on real world scenarios demonstrate the superiority of proposed method over state-of-the-art techniques. △ Less

Submitted 3 July, 2020; originally announced July 2020.

Comments: It was previously submitted to IEEE ICIP 2020. A previous version was also submitted to BMVC 2019

ACM Class: I.4

Helium Ion Microscopy for Reduced Spin Orbit Torque Switching Currents

Authors: Peter Dunne, Ciaran Fowley, Gregor Hlawacek, **u Kurian, Gwenaël Atcheson, Silviu Colis, Niclas Teichert, Bohdan Kundys, M. Venkatesan, Jürgen Lindner, Alina Maria Deac, Thomas M. Hermans, J. M. D. Coey, Bernard Doudin

Abstract: Spin orbit torque driven switching is a favourable way to manipulate nanoscale magnetic objects for both memory and wireless communication devices. The critical current required to switch from one magnetic state to another depends on the geometry and the intrinsic properties of the materials used, which are difficult to control locally. Here we demonstrate how focused helium ion beam irradiation c… ▽ More Spin orbit torque driven switching is a favourable way to manipulate nanoscale magnetic objects for both memory and wireless communication devices. The critical current required to switch from one magnetic state to another depends on the geometry and the intrinsic properties of the materials used, which are difficult to control locally. Here we demonstrate how focused helium ion beam irradiation can modulate the local magnetic anisotropy of a Co thin film at the microscopic scale. Real-time in-situ characterisation using the anomalous Hall effect showed up to an order of magnitude reduction of the magnetic anisotropy under irradiation, and using this, multi-level switching is demonstrated. The result is that spin-switching current densities, down to 800 kA cm$^{-2}$, can be achieved on predetermined areas of the film, without the need for lithography. The ability to vary critical currents spatially has implications not only for storage elements, but also neuromorphic and probabilistic computing. △ Less

Submitted 14 September, 2020; v1 submitted 15 May, 2020; originally announced May 2020.

Comments: Main text: 22 pages, 3 figures, 2 tables, 1 TOC graphic. Included SI: 2 pages, 2 figures

Audio-Visual Decision Fusion for WFST-based and seq2seq Models

Authors: Rohith Aralikatti, Sharad Roy, Abhinav Thanda, Dilip Kumar Margam, Pujitha Appan Kandala, Tanay Sharma, Shankar M Venkatesan

Abstract: Under noisy conditions, speech recognition systems suffer from high Word Error Rates (WER). In such cases, information from the visual modality comprising the speaker lip movements can help improve the performance. In this work, we propose novel methods to fuse information from audio and visual modalities at inference time. This enables us to train the acoustic and visual models independently. Fir… ▽ More Under noisy conditions, speech recognition systems suffer from high Word Error Rates (WER). In such cases, information from the visual modality comprising the speaker lip movements can help improve the performance. In this work, we propose novel methods to fuse information from audio and visual modalities at inference time. This enables us to train the acoustic and visual models independently. First, we train separate RNN-HMM based acoustic and visual models. A common WFST generated by taking a special union of the HMM components is used for decoding using a modified Viterbi algorithm. Second, we train separate seq2seq acoustic and visual models. The decoding step is performed simultaneously for both modalities using shallow fusion while maintaining a common hypothesis beam. We also present results for a novel seq2seq fusion without the weighing parameter. We present results at varying SNR and show that our methods give significant improvements over acoustic-only WER. △ Less

Submitted 29 January, 2020; originally announced January 2020.

Comments: Submitted for review to ICASSP 2020 on October 21st, 2019

LipReading with 3D-2D-CNN BLSTM-HMM and word-CTC models

Authors: Dilip Kumar Margam, Rohith Aralikatti, Tanay Sharma, Abhinav Thanda, Pujitha A K, Sharad Roy, Shankar M Venkatesan

Abstract: In recent years, deep learning based machine lipreading has gained prominence. To this end, several architectures such as LipNet, LCANet and others have been proposed which perform extremely well compared to traditional lipreading DNN-HMM hybrid systems trained on DCT features. In this work, we propose a simpler architecture of 3D-2D-CNN-BLSTM network with a bottleneck layer. We also present analy… ▽ More In recent years, deep learning based machine lipreading has gained prominence. To this end, several architectures such as LipNet, LCANet and others have been proposed which perform extremely well compared to traditional lipreading DNN-HMM hybrid systems trained on DCT features. In this work, we propose a simpler architecture of 3D-2D-CNN-BLSTM network with a bottleneck layer. We also present analysis of two different approaches for lipreading on this architecture. In the first approach, 3D-2D-CNN-BLSTM network is trained with CTC loss on characters (ch-CTC). Then BLSTM-HMM model is trained on bottleneck lip features (extracted from 3D-2D-CNN-BLSTM ch-CTC network) in a traditional ASR training pipeline. In the second approach, same 3D-2D-CNN-BLSTM network is trained with CTC loss on word labels (w-CTC). The first approach shows that bottleneck features perform better compared to DCT features. Using the second approach on Grid corpus' seen speaker test set, we report $1.3\%$ WER - a $55\%$ improvement relative to LCANet. On unseen speaker test set we report $8.6\%$ WER which is $24.5\%$ improvement relative to LipNet. We also verify the method on a second dataset of $81$ speakers which we collected. Finally, we also discuss the effect of feature duplication on BLSTM-HMM model performance. △ Less

Submitted 25 June, 2019; originally announced June 2019.

Comments: Submitted to Interspeech 2019

Study of the Effect of Annealing on the Properties of Mn2RuxGa Thin Films

Authors: K. E. Siewierska, G. Atcheson, K. Borisov, M. Venkatesan, K. Rode, J. M. D. Coey

Abstract: The effect of vacuum annealing thin films of the compensated ferrimagnetic half-metal Mn2RuxGa at temperatures from 250 to 400 degree Celsius is investigated. The 39.3 nm films deposited on (100) MgO substrates exhibit perpendicular magnetic anisotropy due to a small 1% tetragonal elongation induced by substrate strain. The main change on annealing is a modification in the compensation temperature… ▽ More The effect of vacuum annealing thin films of the compensated ferrimagnetic half-metal Mn2RuxGa at temperatures from 250 to 400 degree Celsius is investigated. The 39.3 nm films deposited on (100) MgO substrates exhibit perpendicular magnetic anisotropy due to a small 1% tetragonal elongation induced by substrate strain. The main change on annealing is a modification in the compensation temperature, which first increases from 50 K for the as-deposited film to 185 K after annealing at 250 degree Celsius, and then falls to 140 K after annealing at 400 degree Celsius. There are minor changes in the atomic order, coercivity, resistivity and anomalous Hall effect (AHE), but the net magnetization measured by SQUID magnetometry with the field applied in-plane or perpendicular-to-the-plane changes more significantly. It saturates at 20 to 30 kA/m at room temperature, and a small soft component is seen in the perpendicular SQUID loops which is absent in the square AHE hysteresis loops. This is explained by the half-metallic nature of the compound; the AHE probes only the 4c Mn sublattice that provides the spin-polarized electrons at the Fermi level, whereas the SQUID measures the sum of the oppositely-aligned 4c and 4a sublattice magnetisations. △ Less

Submitted 3 April, 2019; originally announced April 2019.

Journal ref: IEEE Transactions on Magnetics, vol. 53, no. 11, pp. 1-5, Nov. 2017, Art no. 2600505

Magnetic Dead Layers in La$_{0.7}$Sr$_{0.3}$MnO$_3$ Revisited

Authors: S. B. Porter, M. Venkatesan, P. Dunne, B. Doudin, K. Rode, J. M. D. Coey

Abstract: The magnetic dead layers in films a few nanometers thick are investigated for La$_{0.7}$Sr$_{0.3}$MnO$_3$ on (001)-oriented SrTiO$_3$ (STO), LaAlO$_3$ (LAO) and (LaAlO$_3$)$_{0.3}$(Sr$_2$TaAlO$_6$)$_{0.7}$ (LSAT) substrates. An anomalous moment found to persist above the Curie temperature of the La$_{0.7}$Sr$_{0.3}$MnO$_3$ films is not attributed to the films, but to oxygen vacancies at or near th… ▽ More The magnetic dead layers in films a few nanometers thick are investigated for La$_{0.7}$Sr$_{0.3}$MnO$_3$ on (001)-oriented SrTiO$_3$ (STO), LaAlO$_3$ (LAO) and (LaAlO$_3$)$_{0.3}$(Sr$_2$TaAlO$_6$)$_{0.7}$ (LSAT) substrates. An anomalous moment found to persist above the Curie temperature of the La$_{0.7}$Sr$_{0.3}$MnO$_3$ films is not attributed to the films, but to oxygen vacancies at or near the surface of the substrate. The contribution to the moment from the substrate is as high as 20 $μ$B/nm$^2$ in the case of STO or LSAT. The effect is increased by adding an STO cap layer. Taking this d-zero magnetism into account, extrapolated magnetic dead layer thicknesses of 0.8 nm, 1.5 nm and 3.0 nm are found for the manganite films grown on LSAT, STO and LAO substrates, respectively. An STO cap layer eliminates the LSMO dead layer. △ Less

Submitted 3 April, 2019; originally announced April 2019.

Journal ref: IEEE Transactions on Magnetics ( Volume: 53 , Issue: 11 , Nov. 2017 )

Global SNR Estimation of Speech Signals using Entropy and Uncertainty Estimates from Dropout Networks

Authors: Rohith Aralikatti, Dilip Margam, Tanay Sharma, Thanda Abhinav, Shankar M Venkatesan

Abstract: This paper demonstrates two novel methods to estimate the global SNR of speech signals. In both methods, Deep Neural Network-Hidden Markov Model (DNN-HMM) acoustic model used in speech recognition systems is leveraged for the additional task of SNR estimation. In the first method, the entropy of the DNN-HMM output is computed. Recent work on bayesian deep learning has shown that a DNN-HMM trained… ▽ More This paper demonstrates two novel methods to estimate the global SNR of speech signals. In both methods, Deep Neural Network-Hidden Markov Model (DNN-HMM) acoustic model used in speech recognition systems is leveraged for the additional task of SNR estimation. In the first method, the entropy of the DNN-HMM output is computed. Recent work on bayesian deep learning has shown that a DNN-HMM trained with dropout can be used to estimate model uncertainty by approximating it as a deep Gaussian process. In the second method, this approximation is used to obtain model uncertainty estimates. Noise specific regressors are used to predict the SNR from the entropy and model uncertainty. The DNN-HMM is trained on GRID corpus and tested on different noise profiles from the DEMAND noise database at SNR levels ranging from -10 dB to 30 dB. △ Less

Submitted 12 April, 2018; originally announced April 2018.

d-zero Magnetism in Nanoporous Amorphous Alumina Membranes

Authors: Amir Sajad Esmaeily, M. Venkatesan, S. Sen, J. M. D. Coey

Abstract: Nanoporous alumina membranes produced by mild or hard anodisation in oxalic acid at potentials ranging from 5 - 140 V have a controllable pore surface area of up to 200 times the membrane area. They exhibit a saturating magnetic response that is temperature-independent and almost anhysteretic below room temperature. The magnetism cannot be explained by the ~1 ppm of transition-metal impurities pre… ▽ More Nanoporous alumina membranes produced by mild or hard anodisation in oxalic acid at potentials ranging from 5 - 140 V have a controllable pore surface area of up to 200 times the membrane area. They exhibit a saturating magnetic response that is temperature-independent and almost anhysteretic below room temperature. The magnetism cannot be explained by the ~1 ppm of transition-metal impurities present in the membranes. The magnetic moment increases with the area of the open nanopores, reaching values of 0.6 Bohr magnetons per square nanometer for mild anodisation and 8 muB/nm2 for hard anodisation where the growth rate is faster. Crystallization of the membrane or chemical treatment with salicylic acid reduces the effect by up to 90 %. The magnetism is therefore associated with the open pore surfaces of the amorphous Al2O3, and it is independent of the orientation of the applied field with respect to the membrane surface. Possible explanations in terms of electrons associated with oxygen vacancies (F or F+ centres) are discussed and it is concluded that the effect is due to giant orbital paramagnetism rather than collective ferromagnetic spin order. △ Less

Submitted 4 October, 2017; v1 submitted 3 October, 2017; originally announced October 2017.

Journal ref: Phys. Rev. Materials 2, 054405 (2018)

Magnetization and Anisotropy of Cobalt Ferrite Thin Films

Authors: F. Eskandari, S. B. Porter, M. Venkatesan, P. Kameli, K. Rode, J. M. D. Coey

Abstract: The magnetization of thin films of cobalt ferrite frequently falls far below the bulk value of 455 kAm-1, which corresponds to an inverse cation distribution in the spinel structure with a significant orbital moment of about 0.6 muB that is associated with the octahedrally-coordinated Co2+ ions. The orbital moment is responsible for the magnetostriction and magnetocrystalline anisotropy, and its s… ▽ More The magnetization of thin films of cobalt ferrite frequently falls far below the bulk value of 455 kAm-1, which corresponds to an inverse cation distribution in the spinel structure with a significant orbital moment of about 0.6 muB that is associated with the octahedrally-coordinated Co2+ ions. The orbital moment is responsible for the magnetostriction and magnetocrystalline anisotropy, and its sensitivity to imposed strain. We have systematically investigated the structure and magnetism of films produced by pulsed-laser deposition on different substrates (TiO2, MgO, MgAl2O4, SrTiO3, LSAT, LaAlO3) and as a function of temperature (500-700 C) and oxygen pressure (10-4 - 10 Pa). Magnetization at room-temperature ranges from 60 to 440 kAm-1, and uniaxial substrate-induced anisotropy ranges from +220 kJm-3 for films on deposited on MgO (100) to -2100 kJm-3 for films deposited on MgAl2O4 (100), where the room-temperature anisotropy field reaches 14 T. No rearrangement of high-spin Fe3+ and Co2+ cations on tetrahedral and octahedral sites can reduce the magnetization below the bulk value, but a switch from Fe3+ and Co2+ to Fe2+ and low-spin Co3+ on octahedral sites will reduce the low-temperature magnetization to 120 kAm-1, and a consequent reduction of Curie temperature can bring the room-temperature value to near zero. Possible reasons for the appearance of low-spin cobalt in the thin films are discussed. Keywords; Cobalt ferrite, thin films, pulsed-laser deposition, low-spin Co3+, strain engineering of magnetization. △ Less

Submitted 28 September, 2017; originally announced September 2017.

Journal ref: Phys. Rev. Materials 1, 074413 (2017)

Multi-task Learning Of Deep Neural Networks For Audio Visual Automatic Speech Recognition

Authors: Abhinav Thanda, Shankar M Venkatesan

Abstract: Multi-task learning (MTL) involves the simultaneous training of two or more related tasks over shared representations. In this work, we apply MTL to audio-visual automatic speech recognition(AV-ASR). Our primary task is to learn a map** between audio-visual fused features and frame labels obtained from acoustic GMM/HMM model. This is combined with an auxiliary task which maps visual features to… ▽ More Multi-task learning (MTL) involves the simultaneous training of two or more related tasks over shared representations. In this work, we apply MTL to audio-visual automatic speech recognition(AV-ASR). Our primary task is to learn a map** between audio-visual fused features and frame labels obtained from acoustic GMM/HMM model. This is combined with an auxiliary task which maps visual features to frame labels obtained from a separate visual GMM/HMM model. The MTL model is tested at various levels of babble noise and the results are compared with a base-line hybrid DNN-HMM AV-ASR model. Our results indicate that MTL is especially useful at higher level of noise. Compared to base-line, upto 7\% relative improvement in WER is reported at -3 SNR dB △ Less

Submitted 10 January, 2017; originally announced January 2017.

Audio Visual Speech Recognition using Deep Recurrent Neural Networks

Authors: Abhinav Thanda, Shankar M Venkatesan

Abstract: In this work, we propose a training algorithm for an audio-visual automatic speech recognition (AV-ASR) system using deep recurrent neural network (RNN).First, we train a deep RNN acoustic model with a Connectionist Temporal Classification (CTC) objective function. The frame labels obtained from the acoustic model are then used to perform a non-linear dimensionality reduction of the visual feature… ▽ More In this work, we propose a training algorithm for an audio-visual automatic speech recognition (AV-ASR) system using deep recurrent neural network (RNN).First, we train a deep RNN acoustic model with a Connectionist Temporal Classification (CTC) objective function. The frame labels obtained from the acoustic model are then used to perform a non-linear dimensionality reduction of the visual features using a deep bottleneck network. Audio and visual features are fused and used to train a fusion RNN. The use of bottleneck features for visual modality helps the model to converge properly during training. Our system is evaluated on GRID corpus. Our results show that presence of visual modality gives significant improvement in character error rate (CER) at various levels of noise even when the model is trained without noisy data. We also provide a comparison of two fusion methods: feature fusion and decision fusion. △ Less

Submitted 9 November, 2016; originally announced November 2016.

Polynomials and Second Order Linear Recurrences

Authors: Soumyabrata Pal, Shankar M. Venkatesan

Abstract: One of the most interesting results of the last century was the proof completed by Matijasevich that computably enumerable sets are precisely the diophantine sets [MRDP Theorem, 9], thus settling, based on previously developed machinery, Hilbert's question whether there exists a general algorithm for checking the solvability in integers of any diophantine equation. In this paper we describe techni… ▽ More One of the most interesting results of the last century was the proof completed by Matijasevich that computably enumerable sets are precisely the diophantine sets [MRDP Theorem, 9], thus settling, based on previously developed machinery, Hilbert's question whether there exists a general algorithm for checking the solvability in integers of any diophantine equation. In this paper we describe techniques to prove the nonexistence of polynomials in two variables for some simple generalizations of the Fibonacci sequence (explicit diophantine representation of Fibonacci numbers were known from Jones' polynomial whose positive values have the same range as that of Fibonacci numbers), and we believe similar techniques exist for the primes. In this paper we mainly show the following results: (1) using one of the many techniques known for solving the Pell's equation, namely the solution in an extended number system, we prove the existence and explicitly find the polynomials for the recurrences of the form $e(n)=ae(n-1)+e(n-2)$ with starting values of 0 and 1 in particular, and for any arbitrary starting values, in the process defining a concept of fundamental starting numbers, (2) we prove a few identities that seem to be quite interesting and useful, (3) we use these identities in a novel way to generate systems of equations of certain rank deficiency using which we disprove for the first time the existence of any polynomial in 2 variables for the generalized recurrence of the form $e(n)=ae(n-1)+be(n-2)$ △ Less

Submitted 9 September, 2016; v1 submitted 23 August, 2016; originally announced August 2016.

Tight lower bounds for connected queen domination problems on the chessboard

Authors: Sneha S. Venkatesan, S. M. Venkatesan

Abstract: 1. We first show a lower bound of 2N/3-1 for the connected minimum queen domination (or cover) problem on the NXN chessboard - the upper bound is only 2 higher at most and is easy to show. 2. We then define the k-colored connected minimum queen domination, and extend the above proof to show a lower bound of (2 N - k - 2)/3, where the parameter k can be increased to get decreasing lower bounds LB(N… ▽ More 1. We first show a lower bound of 2N/3-1 for the connected minimum queen domination (or cover) problem on the NXN chessboard - the upper bound is only 2 higher at most and is easy to show. 2. We then define the k-colored connected minimum queen domination, and extend the above proof to show a lower bound of (2 N - k - 2)/3, where the parameter k can be increased to get decreasing lower bounds LB(N, k) until one reaches the simple domination lower bound of floor N/2. 3. We also discuss extensions of the connected domination problem and additional directions. △ Less

Submitted 8 August, 2016; originally announced August 2016.

Comments: 9 pages, 1 Figure

Surface magnetism of strontium titanate

Authors: J. M. D. Coey, M Venkatesan, P Stamenov

Abstract: SrTiO3 plays a central role in oxide electronics. It is the substrate of choice for functional oxide heterostructures based on perovskite-structure thin-film stacks, and its surface or interface with a polar oxide such as LaAlO3 can become a two-dimensional conductor because of electronic reconstruction or the presence of oxygen defects. Inconsistent reports of magnetic order in SrTiO3 abound in t… ▽ More SrTiO3 plays a central role in oxide electronics. It is the substrate of choice for functional oxide heterostructures based on perovskite-structure thin-film stacks, and its surface or interface with a polar oxide such as LaAlO3 can become a two-dimensional conductor because of electronic reconstruction or the presence of oxygen defects. Inconsistent reports of magnetic order in SrTiO3 abound in the literature. Here we report a systematic experimental study aimed at establishing how and when SrTiO3 can develop a magnetic moment at room temperature. Polished 100, 110 or 111 crystal slices from four different suppliers are characterized before and after vacuum annealing at 750 °C, both in single-crystal and powdered form. Impurity content is analysed at the surface and in the bulk. Besides the underlying intrinsic diamagnetism of SrTiO3, magnetic signals are of three types-a Curie law susceptibility due to dilute magnetic impurities at the ppm level, a hysteretic, temperature-dependent ferromagnetic impurity contribution, and a practically-anhysteretic, defect-related temperature-independent component that saturates in about 200 mT. The latter component is intrinsic. It is often the largest, reaching 10 Bohr magnetons per nm2 of surface area or more and dominating the magnetic response in low fields at room temperature. It is associated with defects near the surface, and can be destroyed by treatment with Tiron (C6H4Na2O8S2), an electron donor molecule that forms a strong complex with titanium at the surface. The origin of this unusual high-temperature ferromagnetic-like response is discussed. △ Less

Submitted 30 June, 2016; originally announced June 2016.

Collective magnetic response of CeO2 nanoparticles

Authors: J. M. D. Coey, Karl Ackland, M. Venkatesan, S. Sen

Abstract: The magnetism of nanoparticles and thin films of wide-bandgap oxides that include no magnetic cations is an unsolved puzzle. Progress has been hampered both by the irreproducibility of much of the experimental data, and the lack of any generally-accepted theoretical explanation. The characteristic signature is a virtually anhysteretic, temperature-independent magnetization curve which saturates in… ▽ More The magnetism of nanoparticles and thin films of wide-bandgap oxides that include no magnetic cations is an unsolved puzzle. Progress has been hampered both by the irreproducibility of much of the experimental data, and the lack of any generally-accepted theoretical explanation. The characteristic signature is a virtually anhysteretic, temperature-independent magnetization curve which saturates in an applied field that is several orders of magnitude greater than the magnetization. It appears as if a tiny volume fraction, < 0.1%, of the samples is magnetic and that the energy scale of the problem is unusually high for spin magnetism. Here we investigate the effect of dispersing 4 nm CeO2 nanoparticles with powders of gamma-Al2O3, sugar or latex microspheres. The saturation magnetization, Ms ~ 60 A/m for compact samples, is maximized by 1 wt% lanthanum do**. Dispersing the CeO2 nanopowder reduces its magnetic moment by up to an order of magnitude. There is a characteristic length scale of order 100 nm for the magnetism to appear in CeO2 nanoparticle clusters. The phenomenon is explained in terms of a giant orbital paramagnetism that appears in coherent mesoscopic domains due to resonant interaction with zero-point fluctuations of the vacuum electromagnetic field. The theory explains the observed temperature-independent magnetization curve and its do** and dispersion dependence, based on a length scale of 300 nm that corresponds to the wavelength of a maximum in the UV absorption spectrum of the magnetic CeO2 nanoparticles. The coherent domains occupy roughly ten percent of the sample volume. △ Less

Submitted 19 January, 2016; originally announced January 2016.

Prime Power Divisibility,Periodicity and Other Properties of Some Second Order Recurrences

Authors: Soumyabrata Pal, Shankar M. Venkatesan

Abstract: Wall published a paper in 1960 on the Fibonacci sequence where he derived many results concerning the period and prime power divisibility modulo m. His periodicity results have been generalized to second order linear recurrences. Here we study the sequences generated by such recurrences, with starting values of {0,1}: among other things, we derive new prime power divisibility results, derive the p… ▽ More Wall published a paper in 1960 on the Fibonacci sequence where he derived many results concerning the period and prime power divisibility modulo m. His periodicity results have been generalized to second order linear recurrences. Here we study the sequences generated by such recurrences, with starting values of {0,1}: among other things, we derive new prime power divisibility results, derive the period by new methods, establish new identities, show derivations involving powers of matrices generated by these general recurrences, etc. △ Less

Submitted 30 September, 2015; v1 submitted 12 September, 2015; originally announced September 2015.

Comments: 11 pages

Event Centric Modeling Approach in Colocation Pattern Snalysis from Spatial Data

Authors: M. Venkatesan, Arunkumar Thangavelu, P. Prabhavathy

Abstract: Spatial co-location patterns are the subsets of Boolean spatial features whose instances are often located in close geographic proximity. Co-location rules can be identified by spatial statistics or data mining approaches. In data mining method, Association rule-based approaches can be used which are further divided into transaction-based approaches and distance-based approaches. Transaction-based… ▽ More Spatial co-location patterns are the subsets of Boolean spatial features whose instances are often located in close geographic proximity. Co-location rules can be identified by spatial statistics or data mining approaches. In data mining method, Association rule-based approaches can be used which are further divided into transaction-based approaches and distance-based approaches. Transaction-based approaches focus on defining transactions over space so that an Apriori algorithm can be used. The natural notion of transactions is absent in spatial data sets which are embedded in continuous geographic space. A new distance -based approach is developed to mine co-location patterns from spatial data by using the concept of proximity neighborhood. A new interest measure, a participation index, is used for spatial co-location patterns as it possesses an anti-monotone property. An algorithm to discover co-location patterns are designed which generates candidate locations and their table instances. Finally the co-location rules are generated to identify the patterns. △ Less

Submitted 6 September, 2011; originally announced September 2011.

Comments: 9 pages

High spin polarization in epitaxial films of ferrimagnetic Mn3Ga

Authors: Huseyin Kurt, Karsten Rode, Munuswamy Venkatesan, Plamen Stanislavov Stamenov, John Michael David Coey

Abstract: Ferrimagnetic Mn3Ga exhibits a unique combination of low saturation magnetization (Ms = 0.11 MA m-1) and high perpendicular anisotropy with a uniaxial anisotropy constant of Ku = 0.89 MJ m-3. Epitaxial c-axis films exhibit spin polarization as high as 58%, measured using point contact Andreev reflection. These epitaxial films will be able to support thermally stable sub-10 nm bits for spin transfe… ▽ More Ferrimagnetic Mn3Ga exhibits a unique combination of low saturation magnetization (Ms = 0.11 MA m-1) and high perpendicular anisotropy with a uniaxial anisotropy constant of Ku = 0.89 MJ m-3. Epitaxial c-axis films exhibit spin polarization as high as 58%, measured using point contact Andreev reflection. These epitaxial films will be able to support thermally stable sub-10 nm bits for spin transfer torque memories. △ Less

Submitted 23 October, 2010; originally announced October 2010.

Comments: 13 pages, 3 figures, 1 table

Ferromagnetism in defect-ridden oxides and related materials

Authors: J. M. D. Coey, P. Stamenov, R. D. Gunning, M. Venkatesan, K. Paul

Abstract: The existence of high-temperature ferromagnetism in thin films and nanoparticles of oxides containing small quantities of magnetic dopants remains controversial. Some regard these materials as dilute magnetic semiconductors, while others think they are ferromagnetic only because the magnetic dopants form secondary ferromagnetic impurity phases such as cobalt metal or magnetite. There are also repo… ▽ More The existence of high-temperature ferromagnetism in thin films and nanoparticles of oxides containing small quantities of magnetic dopants remains controversial. Some regard these materials as dilute magnetic semiconductors, while others think they are ferromagnetic only because the magnetic dopants form secondary ferromagnetic impurity phases such as cobalt metal or magnetite. There are also reports in d0 systems and other defective oxides that contain no magnetic ions. Here, we investigate TiO2 (rutile) containing 1 - 5% of iron cations and find that the room-temperature ferromagnetism of films prepared by pulsed-laser deposition is not due to magnetic ordering of the iron. The films are neither dilute magnetic semiconductors nor hosts to an iron-based ferromagnetic impurity phase. A new model is developed for defect-related ferromagnetism which involves a spin-split defect band populated by charge transfer from a proximate charge reservoir in the present case a mixture Fe2+ and Fe3+ ions in the oxide lattice. The phase diagram for the model shows how inhomogeneous Stoner ferromagnetism depends on the total number of electrons Ntot, the Stoner exchange integral I and the defect bandwidth W; the band occupancy is governed by the d-d Coulomb interaction U. There are regions of ferromagnetic metal, half-metal and insulator as well as nonmagnetic metal and insulator. A characteristic feature of the high-temperature Stoner magnetism is an an anhysteretic magnetization curve which is practically temperature independent below room temperature. This is related to a wandering ferromagnetic axis which is determined by local dipole fields. The magnetization is limited by the defect concentration, not by the 3d do**. Only 1-2 % of the volume of the films is magnetically ordered. △ Less

Submitted 29 March, 2010; originally announced March 2010.

Comments: 22 pages, 6 figures

Journal ref: New Journal of Physics, Vol.12 (2010)

Reversible Image Authentication with Tamper Localization Based on Integer Wavelet Transform

Authors: P. Meenakshi Devi, M. Venkatesan, K. Duraiswamy

Abstract: In this paper, a new reversible image authentication technique with tamper localization based on watermarking in integer wavelet transform is proposed. If the image authenticity is verified, then the distortion due to embedding the watermark can be completely removed from the watermarked image. If the image is tampered, then the tampering positions can also be localized. Two layers of watermarki… ▽ More In this paper, a new reversible image authentication technique with tamper localization based on watermarking in integer wavelet transform is proposed. If the image authenticity is verified, then the distortion due to embedding the watermark can be completely removed from the watermarked image. If the image is tampered, then the tampering positions can also be localized. Two layers of watermarking are used. The first layer embedded in spatial domain verifies authenticity and the second layer embedded in transform domain provides reversibility. This technique utilizes selective LSB embedding and histogram characteristics of the difference images of the wavelet coefficients and modifies pixel values slightly to embed the watermark. Experimental results demonstrate that the proposed scheme can detect any modifications of the watermarked image. △ Less

Submitted 3 December, 2009; originally announced December 2009.

Comments: 8 pages IEEE format, International Journal of Computer Science and Information Security, IJCSIS November 2009, ISSN 1947 5500, http://sites.google.com/site/ijcsis/

Report number: ISSN 1947 5500

Journal ref: International Journal of Computer Science and Information Security, IJCSIS, Vol. 6, No. 2, pp. 067-074, November 2009, USA

Ferromagnetism in substituted zinc oxide

Authors: M. Venkatesan, C. B. Fitzgerald, J. G. Lunney, J. M. D. Coey

Abstract: Room-temperature ferromagnetism is observed in (110) oriented ZnO films containing 5 at % of Sc, Ti, V, Fe, Co or Ni, but not Cr, Mn or Cu ions. There are large moments, 1.9 and 0.5 muB/atom for Co- and Ti-substituted oxides, respectively. Sc-substituted ZnO shows also a moment of 0.3 muB/Sc. Magnetization is very anisotropic, with variations of up to a factor three depending on the orientation… ▽ More Room-temperature ferromagnetism is observed in (110) oriented ZnO films containing 5 at % of Sc, Ti, V, Fe, Co or Ni, but not Cr, Mn or Cu ions. There are large moments, 1.9 and 0.5 muB/atom for Co- and Ti-substituted oxides, respectively. Sc-substituted ZnO shows also a moment of 0.3 muB/Sc. Magnetization is very anisotropic, with variations of up to a factor three depending on the orientation of the applied field relative to the R-cut sapphire substrates. Results are interpreted in terms of a spin-split donor impurity band model, which can account for ferromagnetism in insulating or conducting high-k oxides with concentrations of magnetic ions that lie far below the percolation threshold. The variation of the ferromagnetism with oxygen pressure used during film growth is evidence of a link between ferromagnetism and defect concentration. △ Less

Submitted 29 June, 2004; originally announced June 2004.

Comments: 15 pages, 4 figures

Ferromagnetism in Fe-doped SnO2 thin films

Authors: J. M. D. Coey, A. P. Douvalis, C. B. Fitzgerald, M. Venkatesan

Abstract: Thin films grown by pulsed-laser deposition from targets of Sn0.95Fe0.05O2 are transparent ferromagnets with Curie temperature and spontaneous magnetization of 610 K and 2.2 Am2kg-1, respectively. The 57Fe Mossbauer spectra show the iron is all high-spin Fe3+ but the films are magnetically inhomogeneous on an atomic scale, with only 23 % of the iron ordering magnetically. The net ferromagnetic m… ▽ More Thin films grown by pulsed-laser deposition from targets of Sn0.95Fe0.05O2 are transparent ferromagnets with Curie temperature and spontaneous magnetization of 610 K and 2.2 Am2kg-1, respectively. The 57Fe Mossbauer spectra show the iron is all high-spin Fe3+ but the films are magnetically inhomogeneous on an atomic scale, with only 23 % of the iron ordering magnetically. The net ferromagnetic moment per ordered iron ion, 1.8 Bohr magnetons, is greater than for any simple iron oxide. Ferromagnetic coupling of ferric ions via an electron trapped in a bridging oxygen vacancy (F center) is proposed to explain the high Curie temperature △ Less

Submitted 16 January, 2004; originally announced January 2004.

Comments: 13 pages, 4 figures, to be published in Applied Physics Letters (Feb. 2004)

Ferromagnetism of a graphite nodule from the Canyon Diablo meteorite

Authors: J. M. D. Coey, M. Venkatesan, C. B. Fitzgerald, A. P. Douvalis, I. S. Sanders

Abstract: There have recently been various reports of weak ferromagnetism in graphite (1,2) and synthetic carbon materials (3) such as rhombohedral C60 (4), as well as a theoretical prediction of a ferromagnetic instability in graphene sheets (5). With very small ferromagnetic signals, it is difficult to be certain that the origin is intrinsic, rather than due to minute concentrations of iron-rich impurit… ▽ More There have recently been various reports of weak ferromagnetism in graphite (1,2) and synthetic carbon materials (3) such as rhombohedral C60 (4), as well as a theoretical prediction of a ferromagnetic instability in graphene sheets (5). With very small ferromagnetic signals, it is difficult to be certain that the origin is intrinsic, rather than due to minute concentrations of iron-rich impurities. Here we take a different experimental approach to study ferromagnetism in graphitic materials, by making use of meteoritic graphite, which is strongly ferromagnetic at room temperature. We examined ten samples of extraterrestrial graphite from a nodule in the Canyon Diablo meteorite. Graphite is the major phase in every sample but there are minor amounts of magnetite, kamacite, akaganeite, and other phases. By analysing the phase composition of a series of samples, we find that these iron-rich minerals can only account for about two-thirds of the observed magnetization. The remainder is somehow associated with graphite, corresponding to an average magnetization of 23 Am2kg-1, or 0.05 Bohr magnetons per carbon atom. The magnetic ordering temperature is near 570 K. We suggest that the ferromagnetism is a magnetic proximity effect induced at the interface with magnetite or kamacite inclusions. △ Less

Submitted 14 November, 2002; originally announced November 2002.

Comments: 15 pages, 4 figures, 1 table

Journal ref: Nature 420, 156 - 159 (2002); doi:10.1038/nature01100