PyTorchKR Papers

최신 논문

총 9,642편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2002.05922

Realizing a Low-Power Head-Mounted Phase-Only Holographic Display by Light-Weight Compression

Head-mounted holographic displays (HMHD) are projected to be the first commercial realization of holographic video display systems. HMHDs use liquid crystal on silicon (LCoS) spatial light modulators (SLM), which are best suited to display phase-only holograms (POH). The performance/watt requirement of a monochrome, 60 fps Full HD, 2-eye, POH HMHD system is about 10 TFLOPS/W, which is orders of magnitude higher than that is achievable by commercially available mobile processors. To mitigate this compute power constraint, display-ready POHs shall be generated on a nearby server and sent to the HMHD in compressed form over a wireless link. This paper discusses design of a feasible HMHD-based augmented reality system, focusing on compression requirements and per-pixel rate-distortion trade-off for transmission of display-ready POH from the server to HMHD. Since the decoder in the HMHD needs to operate on low power, only coding methods that have low-power decoder implementation are considered. Effects of 2D phase unwrapping and flat quantization on compression performance are also reported. We next propose a versatile PCM-POH codec with progressive quantization that can adapt to SLM-dynamic-range and available bitrate, and features per-pixel rate-distortion control to achieve acceptable POH quality at target rates of 60-200 Mbit/s that can be reliably achieved by current wireless technologies. Our results demonstrate feasibility of realizing a low-power, quality-ensured, multi-user, interactive HMHD augmented reality system with commercially available components using the proposed adaptive compression of display-ready POH with light-weight decoding.

arXiv:2001.01833

Line-based compressive sensing for low-power visual applications

In this paper, a Line based Compressive Sensing (LCS) scheme is discussed and proposed for low power visual applications, in which image acquisition is performed in a line-by-line manner at the encoder side using same measurement operator. Such approach reduces the computational burden and makes the implementation process easier (at encoder) plus provides better and more efficient initial reconstruction (at decoder) than other CS techniques. The reconstruction algorithm is based on the combination of the conventional augmented Lagrangian method with variable splitting and alternating direction method and is referred as TV-AL3. The simulation results show that the proposed line based CS scheme not only improves the quality of image by 1dB to ~3dB at various subrates, when compared to the conventional CS schemes but also reduces the computational complexity at the encoder side.

arXiv:1911.12656

Optimal Multivariate Tuning with Neuron-Level and Population-Level Energy Constraints

Optimality principles have been useful in explaining many aspects of biological systems. In the context of neural encoding in sensory areas, optimality is naturally formulated in a Bayesian setting, as neural tuning which minimizes mean decoding error. Many works optimize Fisher information, which approximates the Minimum Mean Square Error (MMSE) of the optimal decoder for long encoding time, but may be misleading for short encoding times. We study MMSE-optimal neural encoding of a multivariate stimulus by uniform populations of spiking neurons, under firing rate constraints for each neuron as well as for the entire population. We show that the population-level constraint is essential for the formulation of a well-posed problem having finite optimal tuning widths, and optimal tuning aligns with the principal components of the prior distribution. Numerical evaluation of the two-dimensional case shows that encoding only the dimension with higher variance is optimal for short encoding times. We also compare direct MMSE optimization to optimization of several proxies to MMSE, namely Fisher information, Maximum Likelihood estimation error, and the Bayesian Cram\'er-Rao bound. We find that optimization of these measures yield qualitatively misleading results regarding MMSE-optimal tuning and its dependence on encoding time and energy constraints.

arXiv:1911.05560

Enhanced Voice Post Processing Using Voice Decoder Guidance Indicators

Voice enhancement and voice coding are imperative and important functions in a voice-communication system. However, both functions are commonly treated independently, even though both utilize similar features of the underlying signals. Our proposal is to leverage information from one function to the benefit of the other. Specifically, our proposed changes are focused on changes to the voice enhancement at the downlink side and utilizing information of the voice decoding. Preliminary results show that such an approach results in improved quality. Additionally, suggestions are provided on future extensions of the proposed concept.

arXiv:1910.14650

Generative network complex (GNC) for drug discovery

It remains a challenging task to generate a vast variety of novel compounds with desirable pharmacological properties. In this work, a generative network complex (GNC) is proposed as a new platform for designing novel compounds, predicting their physical and chemical properties, and selecting potential drug candidates that fulfill various druggable criteria such as binding affinity, solubility, partition coefficient, etc. We combine a SMILES string generator, which consists of an encoder, a drug-property controlled or regulated latent space, and a decoder, with verification deep neural networks, a target-specific three-dimensional (3D) pose generator, and mathematical deep learning networks to generate new compounds, predict their drug properties, construct 3D poses associated with target proteins, and reevaluate druggability, respectively. New compounds were generated in the latent space by either randomized output, controlled output, or optimized output. In our demonstration, 2.08 million and 2.8 million novel compounds are generated respectively for Cathepsin S and BACE targets. These new compounds are very different from the seeds and cover a larger chemical space. For potentially active compounds, their 3D poses are generated using a state-of-the-art method. The resulting 3D complexes are further evaluated for druggability by a championing deep learning algorithm based on algebraic topology, differential geometry, and algebraic graph theories. Performed on supercomputers, the whole process took less than one week. Therefore, our GNC is an efficient new paradigm for discovering new drug candidates.

arXiv:1910.08721공식 코드 · Papers with Code

Learning to Reconstruct Crack Profiles for Eddy Current Nondestructive Testing

Eddy current testing (ECT) is one of the most popular Nondestructive Testing (NDT) techniques, especially for conductive materials. Reconstructing the crack profile from measured EC signals is one of the main goals of ECT. This task is highly challenging, as the EC signals are nonlinear responses resulted from the presence of cracks, and reconstructing the crack profile requires establishing the forward model of the nonlinear electromagnetic dynamics and solving its inverse problem, which is an ill-posed numerical optimization problem. Instead of solving the inverse problem numerically, we propose to directly learn the inverse mapping from EC signals to crack profiles with a deep encoder-decoder convolutional neural network named EddyNet. EddyNet is trained on a set of randomly generated crack profiles and the corresponding simulated EC responses generated from a realistic forward model. On the held-out test data, EddyNet achieved a mean absolute error of 0.198 between predicted profiles and ground truth ones. Qualitatively, the geometries of predicted profiles are visually similar to the ground truth profiles. Our method greatly reduces the usual reliance on domain experts, and the reconstruction is extremely fast both on GPUs and on CPUs. The source code of EddyNet is released on https://github.com/askerlee/EddyNet.

arXiv:1909.04223

MAJoRCom: A Dual-Function Radar Communication System Using Index Modulation

Dual-function radar communication (DFRC) systems implement both sensing and communication using the same hardware. Such schemes are often more efficient in terms of size, power, and cost, over using distinct radar and communication systems. Since these functionalities share resources such as spectrum, power, and antennas, DFRC methods typically entail some degradation in both radar and communication performance. In this work we propose a DFRC scheme based on the carrier agile phased array radar (CAESAR), which combines frequency and spatial agility. The proposed DFRC system, referred to as multi-carrier agile joint radar communication (MAJoRCom), exploits the inherent spatial and spectral randomness of CAESAR to convey digital messages in the form of index modulation. The resulting communication scheme naturally coexists with the radar functionality, and thus does not come at the cost of reduced radar performance. We analyze the performance of MAJoRCom, quantifying its achievable bit rate. In addition, we develop a low complexity decoder and a codebook design approach, which simplify the recovery of the communicated bits. Our numerical results demonstrate that MAJoRCom is capable of achieving a bit rate which is comparable to utilizing independent communication modules without affecting the radar performance, and that our proposed low-complexity decoder allows the receiver to reliably recover the transmitted symbols with an affordable computational burden.

arXiv:1908.07045

Salient Speech Representations Based on Cloned Networks

We define salient features as features that are shared by signals that are defined as being equivalent by a system designer. The definition allows the designer to contribute qualitative information. We aim to find salient features that are useful as conditioning for generative networks. We extract salient features by jointly training a set of clones of an encoder network. Each network clone receives as input a different signal from a set of equivalent signals. The objective function encourages the network clones to map their input into a set of features that is identical across the clones. It additionally encourages feature independence and, optionally, reconstruction of a desired target signal by a decoder. As an application, we train a system that extracts a time-sequence of feature vectors of speech and uses it as a conditioning of a WaveNet generative system, facilitating both coding and enhancement.

arXiv:1908.04672

Estimating & Mitigating the Impact of Acoustic Environments on Machine-to-Machine Signalling

The advance of technology for transmitting Data-over-Sound in various IoT and telecommunication applications has led to the concept of machine-to-machine over-the-air acoustic signalling. Reverberation can have a detrimental effect on such machine-to-machine signals while decoding. Various methods have been studied to combat the effects of reverberation in speech and audio signals, but it is not clear how well they generalise to other sound types. We look at extending these models to facilitate machine-to-machine acoustic signalling. This research investigates dereverberation techniques to shortlist a single-channel reverberation suppression method through a pilot test. In order to apply the chosen dereverberation method a novel method of estimating acoustic parameters governing reverberation is proposed. The performance of the final algorithm is evaluated on quality metrics as well as the performance of a real machine-to-machine decoder. We demonstrate a dramatic reduction in error rate for both audible and ultrasonic signals.

arXiv:1907.12613

Autoencoder for Interconnect's Bandwidth Relaxation in Large Scale MIMO-OFDM Processing

Deep learning is playing an instrumental role in the design of the next generation of communication systems. In this letter, we address the massive MIMO interconnect's bandwidth constraint relaxation using autoencoders. The autoencoder is trained to learn the received signal structure so that a low dimension latent variable is transferred as opposed to the original high dimension signal. For an efficient implementation, the approach suggests to separately deploy the autoencoder components, namely the encoder and the decoder, in massive MIMO radio head and central processing units respectively. The simulation results show that one can relax the interconnect's bandwidth by a factor of up to 16 with non-substantial performance degradation of the centralized and the decentralized processing. Fortunately, such a loss can be compensated by running few extra iterations in the detection process which renders the autoencoder-iterative detection an interconnect's bandwidth and computational complexity design trade-off.

arXiv:1905.04628공식 코드 · Papers with Code

Improving Opus Low Bit Rate Quality with Neural Speech Synthesis

The voice mode of the Opus audio coder can compress wideband speech at bit rates ranging from 6 kb/s to 40 kb/s. However, Opus is at its core a waveform matching coder, and as the rate drops below 10 kb/s, quality degrades quickly. As the rate reduces even further, parametric coders tend to perform better than waveform coders. In this paper we propose a backward-compatible way of improving low bit rate Opus quality by re-synthesizing speech from the decoded parameters. We compare two different neural generative models, WaveNet and LPCNet. WaveNet is a powerful, high-complexity, and high-latency architecture that is not feasible for a practical system, yet provides a best known achievable quality with generative models. LPCNet is a low-complexity, low-latency RNN-based generative model, and practically implementable on mobile phones. We apply these systems with parameters from Opus coded at 6 kb/s as conditioning features for the generative models. A listening test shows that for the same 6 kb/s Opus bit stream, synthesized speech using LPCNet clearly outperforms the output of the standard Opus decoder. This opens up ways to improve the decoding quality of existing speech and audio waveform coders without breaking compatibility.

arXiv:1904.06599

TDMR Detection System with Local Area Influence Probabilistic a Priori Detector

We propose a three-track detection system for two dimensional magnetic recording (TDMR) in which a local area influence probabilistic (LAIP) detector works with a trellis-based Bahl-Cocke-Jelinek-Raviv (BCJR) detector to remove intersymbol interference (ISI) and intertrack interference (ITI) among coded data bits as well as media noise due to magnetic grain-bit interactions. Two minimum mean-squared error (MMSE) linear equalizers with different response targets are employed before the LAIP and BCJR detectors. The LAIP detector considers local grain-bit interactions and passes coded bit log-likelihood ratios (LLRs) to the channel decoder, whose output LLRs serve as a priori information to the BCJR detector, which is followed by a second channel decoding pass. Simulation results under 1-shot decoding on a grain-flipping-probability (GFP) media model show that the proposed LAIP/BCJR detection system achieves density gains of 6.8% for center-track detection and 1.2% for three-track detection compared to a standard BCJR/1D-PDNP. The proposed system's BCJR detector bit error rates (BERs) are lower than those of a recently proposed two-track BCJR/2D-PDNP system by factors of (0.55, 0.08) for tracks 1 and 2 respectively.

arXiv:1901.06537

Deep-Learning-based Millimeter-Wave Massive MIMO for Hybrid Precoding

Millimeter wave (mmWave) massive multiple-input multiple-output (MIMO) has been regarded to be an emerging solution for the next generation of communications, in which hybrid analog and digital precoding is an important method for reducing the hardware complexity and energy consumption associated with mixed signal components. However, the fundamental limitations of the existing hybrid precoding schemes is that they have high computational complexity and fail to fully exploit the spatial information. To overcome these limitations, this paper proposes, a deep-learning-enabled mmWave massive MIMO framework for effective hybrid precoding, in which each selection of the precoders for obtaining the optimized decoder is regarded as a mapping relation in the deep neural network (DNN). Specifically, the hybrid precoder is selected through training based on the DNN for optimizing precoding process of the mmWave massive MIMO. Additionally, we present extensive simulation results to validate the excellent performance of the proposed scheme. The results exhibit that the DNN-based approach is capable ofminimizing the bit error ratio (BER) and enhancing spectrum efficiency of the mmWave massive MIMO, which achieves better performance in hybrid precoding compared with conventional schemes while substantially reducing the required computational complexity.

arXiv:1812.11455

High-performance Decoder for Convolutional Code with Deep Neural Network

The use of deep neural network for decoding error control code will encounter two problems, namely, the high-precision requirements of the error control code and the complexity of the neural network due to the long code. In this paper, a deep neural network decoder is proposed to solve the decoding problem of long code by using the nature of convolutional code window decoding. A deep neural network decoder is utilized as a weak classifier, and an integrated decoder is proposed to improve the decoding performance greatly. The Viterbi decoder is improved by approximately 2 db at a bit error rate of level 5. Both decoder methods proposed in this paper can be decoded in parallel and are suitable for high-bit-rate applications. This study reveals that the accuracy of neural networks can reach level 8 more.

arXiv:1811.12063

Performance Analysis of Deep Learning based on Recurrent Neural Networks for Channel Coding

Channel Coding has been one of the central disciplines driving the success stories of current generation LTE systems and beyond. In particular, turbo codes are mostly used for cellular and other applications where a reliable data transfer is required for latency-constrained communication in the presence of data-corrupting noise. However, the decoding algorithm for turbo codes is computationally intensive and thereby limiting its applicability in hand-held devices. In this paper, we study the feasibility of using Deep Learning (DL) architectures based on Recurrent Neural Networks (RNNs) for encoding and decoding of turbo codes. In this regard, we simulate and use data from various stages of the transmission chain (turbo encoder output, Additive White Gaussian Noise (AWGN) channel output, demodulator output) to train our proposed RNN architecture and compare its performance to the conventional turbo encoder/decoder algorithms. Simulation results show, that the proposed RNN model outperforms the decoding performance of a conventional turbo decoder at low Signal to Noise Ratio (SNR) regions

arXiv:1811.08857

Decoding Staircase Codes with Marked Bits

Staircase codes (SCCs) are typically decoded using iterative bounded-distance decoding (BDD) and hard decisions. In this paper, a novel decoding algorithm is proposed, which partially uses soft information from the channel. The proposed algorithm is based on marking certain number of highly reliable and highly unreliable bits. These marked bits are used to improve the miscorrection-detection capability of the SCC decoder and the error-correcting capability of BDD. For SCCs with $2$-error-correcting BCH component codes, our algorithm improves upon standard SCC decoding by up to $0.30$~dB at a bit-error rate of $10^{-7}$. The proposed algorithm is shown to achieve almost half of the gain achievable by an idealized decoder with this structure.

arXiv:1811.07505

Low Complexity Iterative Detection for a Large-scale Distributed MIMO Prototyping System

In this paper, we study the low-complexity iterative soft-input soft-output (SISO) detection algorithm in a large-scale distributed multiple-input multiple-output (MIMO) system. The uplink interference suppression matrix is designed to decompose the received multi-user signal into independent single-user receptions. An improved minimum-mean-square-error iterative soft decision interference cancellation (MMSE-ISDIC) based on eigenvalue decomposition (EVD-MMSE-ISDIC) is proposed to perform low-complexity detection of the decomposed signals. Furthermore, two iteration schemes are given to improve receiving performance, which are iterative detection and decoding (IDD) scheme and iterative detection (ID) scheme. While IDD utilizes the external information generated by the decoder for iterative detection, the output information of the detector is directly exploited with ID. In particular, a large-scale distributed MIMO prototyping system is introduced and a 32$ \times $32 (4 user equipments (UEs) and 4 remote antenna units (RAUs), each equipped with 8 antennas) experimental tests at 3.5 GHz was performed. The experimental results show that the proposed iterative receiver greatly outperforms the linear MMSE receiver, since it reduces the average number of error blocks of the system significantly.

arXiv:1811.02656

Quaternion Convolutional Neural Networks for Heterogeneous Image Processing

Convolutional neural networks (CNN) have recently achieved state-of-the-art results in various applications. In the case of image recognition, an ideal model has to learn independently of the training data, both local dependencies between the three components (R,G,B) of a pixel, and the global relations describing edges or shapes, making it efficient with small or heterogeneous datasets. Quaternion-valued convolutional neural networks (QCNN) solved this problematic by introducing multidimensional algebra to CNN. This paper proposes to explore the fundamental reason of the success of QCNN over CNN, by investigating the impact of the Hamilton product on a color image reconstruction task performed from a gray-scale only training. By learning independently both internal and external relations and with less parameters than real valued convolutional encoder-decoder (CAE), quaternion convolutional encoder-decoders (QCAE) perfectly reconstructed unseen color images while CAE produced worst and gray-scale versions.

arXiv:1811.00879

CHIRRUP: a practical algorithm for unsourced multiple access

Unsourced multiple access abstracts grantless simultaneous communication of a large number of devices (messages) each of which transmits (is transmitted) infrequently. It provides a model for machine-to-machine communication in the Internet of Things (IoT), including the special case of radio-frequency identification (RFID), as well as neighbor discovery in ad hoc wireless networks. This paper presents a fast algorithm for unsourced multiple access that scales to $2^{100}$ devices (arbitrary $100$ bit messages). The primary building block is multiuser detection of binary chirps which are simply codewords in the second order Reed Muller code. The chirp detection algorithm originally presented by Howard et al. is enhanced and integrated into a peeling decoder designed for a patching and slotting framework. In terms of both energy per bit and number of transmitted messages, the proposed algorithm is within a factor of $2$ of state of the art approaches. A significant advantage of our algorithm is its computational efficiency. We prove that the worst-case complexity of the basic chirp reconstruction algorithm is $\mathcal{O}[nK(\log_2 n + K)]$, where $n$ is the codeword length and $K$ is the number of active users, and we report computing times for our algorithm. Our performance and computing time results represent a benchmark against which other practical algorithms can be measured.

arXiv:1810.00660

Attention-based Encoder-Decoder Networks for Spelling and Grammatical Error Correction

Automatic spelling and grammatical correction systems are one of the most widely used tools within natural language applications. In this thesis, we assume the task of error correction as a type of monolingual machine translation where the source sentence is potentially erroneous and the target sentence should be the corrected form of the input. Our main focus in this project is building neural network models for the task of error correction. In particular, we investigate sequence-to-sequence and attention-based models which have recently shown a higher performance than the state-of-the-art of many language processing problems. We demonstrate that neural machine translation models can be successfully applied to the task of error correction. While the experiments of this research are performed on an Arabic corpus, our methods in this thesis can be easily applied to any language.