Although Large Language Models (LLMs) represent a revolution in the way we interact with computers, allowing the construction of complex questions and the ability to reason over a sequence of statements, their use is restricted due to the need for dedicated hardware for execution. In this study, we evaluate the performance of LLMs based on the 7 and 13 billion LLaMA models, subjected to a quantization process and run on home hardware. The models considered were Alpaca, Koala, and Vicuna. To evaluate the effectiveness of these models, we developed a database containing 1,006 questions from the ENEM (Brazilian National Secondary School Exam). Our analysis revealed that the best performing models achieved an accuracy of approximately 46% for the original texts of the Portuguese questions and 49% on their English translations. In addition, we evaluated the computational efficiency of the models by measuring the time required for execution. On average, the 7 and 13 billion LLMs took approximately 20 and 50 seconds, respectively, to process the queries on a machine equipped with an AMD Ryzen 5 3600x processor
This paper presents BELT, a novel model and learning framework for the pivotal topic of brain-to-language translation research. The translation from noninvasive brain signals into readable natural language has the potential to promote the application scenario as well as the development of brain-computer interfaces (BCI) as a whole. The critical problem in brain signal decoding or brain-to-language translation is the acquisition of semantically appropriate and discriminative EEG representation from a dataset of limited scale and quality. The proposed BELT method is a generic and efficient framework that bootstraps EEG representation learning using off-the-shelf large-scale pretrained language models (LMs). With a large LM's capacity for understanding semantic information and zero-shot generalization, BELT utilizes large LMs trained on Internet-scale datasets to bring significant improvements to the understanding of EEG signals. In particular, the BELT model is composed of a deep conformer encoder and a vector quantization encoder. Semantical EEG representation is achieved by a contrastive learning step that provides natural language supervision. We achieve state-of-the-art results on two featuring brain decoding tasks including the brain-to-language translation and zero-shot sentiment classification. Specifically, our model surpasses the baseline model on both tasks by 5.45% and over 10% and archives a 42.31% BLEU-1 score and 67.32% precision on the main evaluation metrics for translation and zero-shot sentiment classification respectively.
Discrete audio representation, aka audio tokenization, has seen renewed interest driven by its potential to facilitate the application of text language modeling approaches in audio domain. To this end, various compression and representation-learning based tokenization schemes have been proposed. However, there is limited investigation into the performance of compression-based audio tokens compared to well-established mel-spectrogram features across various speaker and speech related tasks. In this paper, we evaluate compression based audio tokens on three tasks: Speaker Verification, Diarization and (Multi-lingual) Speech Recognition. Our findings indicate that (i) the models trained on audio tokens perform competitively, on average within $1\%$ of mel-spectrogram features for all the tasks considered, and do not surpass them yet. (ii) these models exhibit robustness for out-of-domain narrowband data, particularly in speaker tasks. (iii) audio tokens allow for compression to 20x compared to mel-spectrogram features with minimal loss of performance in speech and speaker related tasks, which is crucial for low bit-rate applications, and (iv) the examined Residual Vector Quantization (RVQ) based audio tokenizer exhibits a low-pass frequency response characteristic, offering a plausible explanation for the observed results, and providing insight for future tokenizer designs.
Quantization is a widely used compression method that effectively reduces redundancies in over-parameterized neural networks. However, existing quantization techniques for deep neural networks often lack a comprehensive error analysis due to the presence of non-convex loss functions and nonlinear activations. In this paper, we propose a fast stochastic algorithm for quantizing the weights of fully trained neural networks. Our approach leverages a greedy path-following mechanism in combination with a stochastic quantizer. Its computational complexity scales only linearly with the number of weights in the network, thereby enabling the efficient quantization of large networks. Importantly, we establish, for the first time, full-network error bounds, under an infinite alphabet condition and minimal assumptions on the weights and input data. As an application of this result, we prove that when quantizing a multi-layer network having Gaussian weights, the relative square quantization error exhibits a linear decay as the degree of over-parametrization increases. Furthermore, we demonstrate that it is possible to achieve error bounds equivalent to those obtained in the infinite alphabet case, using on the order of a mere $\log\log N$ bits per weight, where $N$ represents the largest number of neurons in a layer.
The proliferation of edge devices has unlocked unprecedented opportunities for deep learning model deployment in computer vision applications. However, these complex models require considerable power, memory and compute resources that are typically not available on edge platforms. Ultra low-bit quantization presents an attractive solution to this problem by scaling down the model weights and activations from 32-bit to less than 8-bit. We implement highly optimized ultra low-bit convolution operators for ARM-based targets that outperform existing methods by up to 4.34x. Our operator is implemented within Deeplite Runtime (DeepliteRT), an end-to-end solution for the compilation, tuning, and inference of ultra low-bit models on ARM devices. Compiler passes in DeepliteRT automatically convert a fake-quantized model in full precision to a compact ultra low-bit representation, easing the process of quantized model deployment on commodity hardware. We analyze the performance of DeepliteRT on classification and detection models against optimized 32-bit floating-point, 8-bit integer, and 2-bit baselines, achieving significant speedups of up to 2.20x, 2.33x and 2.17x, respectively.
We present Clinical Prediction with Large Language Models (CPLLM), a method that involves fine-tuning a pre-trained Large Language Model (LLM) for clinical disease and readmission prediction. We utilized quantization and fine-tuned the LLM using prompts. For diagnosis prediction, we predict whether patients will be diagnosed with a target disease during their next visit or in the subsequent diagnosis, leveraging their historical diagnosis records. We compared our results to various baselines, including RETAIN, and Med-BERT, the current state-of-the-art model for disease prediction using temporal structured EHR data. In addition, We also evaluated CPLLM for patient hospital readmission prediction and compared our method's performance with benchmark baselines. Our experiments have shown that our proposed method, CPLLM, surpasses all the tested models in terms of PR-AUC and ROC-AUC metrics, showing state-of-the-art results for diagnosis prediction and patient hospital readmission prediction. Such a method can be easily implemented and integrated into the clinical process to help care providers estimate the next steps of patients
The state-of-the-art neural video codecs have outperformed the most sophisticated traditional codecs in terms of RD performance in certain cases. However, utilizing them for practical applications is still challenging for two major reasons. 1) Cross-platform computational errors resulting from floating point operations can lead to inaccurate decoding of the bitstream. 2) The high computational complexity of the encoding and decoding process poses a challenge in achieving real-time performance. In this paper, we propose a real-time cross-platform neural video codec, which is capable of efficiently decoding of 720P video bitstream from other encoding platforms on a consumer-grade GPU. First, to solve the problem of inconsistency of codec caused by the uncertainty of floating point calculations across platforms, we design a calibration transmitting system to guarantee the consistent quantization of entropy parameters between the encoding and decoding stages. The parameters that may have transboundary quantization between encoding and decoding are identified in the encoding stage, and their coordinates will be delivered by auxiliary transmitted bitstream. By doing so, these inconsistent parameters can be processed properly in the decoding stage. Furthermore, to reduce the bitrate of the auxiliary bitstream, we rectify the distribution of entropy parameters using a piecewise Gaussian constraint. Second, to match the computational limitations on the decoding side for real-time video codec, we design a lightweight model. A series of efficiency techniques enable our model to achieve 25 FPS decoding speed on NVIDIA RTX 2080 GPU. Experimental results demonstrate that our model can achieve real-time decoding of 720P videos while encoding on another platform. Furthermore, the real-time model brings up to a maximum of 24.2\% BD-rate improvement from the perspective of PSNR with the anchor H.265.
We study semantic compression for text where meanings contained in the text are conveyed to a source decoder, e.g., for classification. The main motivator to move to such an approach of recovering the meaning without requiring exact reconstruction is the potential resource savings, both in storage and in conveying the information to another node. Towards this end, we propose semantic quantization and compression approaches for text where we utilize sentence embeddings and the semantic distortion metric to preserve the meaning. Our results demonstrate that the proposed semantic approaches result in substantial (orders of magnitude) savings in the required number of bits for message representation at the expense of very modest accuracy loss compared to the semantic agnostic baseline. We compare the results of proposed approaches and observe that resource savings enabled by semantic quantization can be further amplified by semantic clustering. Importantly, we observe the generalizability of the proposed methodology which produces excellent results on many benchmark text classification datasets with a diverse array of contexts.
Inexpensive tags powered by energy harvesting (EH) can realize green (energy-efficient) Internet of Things (IoT) networks. However, tags are vulnerable to energy insecurities, resulting in poor communication ranges, activation distances, and data rates. To overcome these challenges, we explore the use of a reconfigurable intelligent surface (RIS) for EH-based IoT networks. The RIS is deployed to enhance RF power at the tag, improving EH capabilities. We consider linear and non-linear EH models and analyze single-tag and multi-tag scenarios. For single-tag networks, the tag's maximum received power and the reader's signal-to-noise ratio with the optimized RIS phase-shifts are derived. Key metrics, such as received power, harvested power, achievable rate, outage probability, bit error rate, and diversity order, are also evaluated. The impact of RIS phase shift quantization errors is also studied. For the multi-tag case, an algorithm to compute the optimal RIS phase-shifts is developed. Numerical results and simulations demonstrate significant improvements compared to the benchmarks of no-RIS case and random RIS-phase design. For instance, our optimal design with a \num{200}-element RIS increases the activation distance by \qty{270}{\percent} and \qty{55}{\percent} compared to those benchmarks. In summary, RIS deployment improves the energy autonomy of tags while maintaining the basic tag design intact.
Direction of arrival (DOA) estimation employing low-resolution analog-to-digital convertors (ADCs) has emerged as a challenging and intriguing problem, particularly with the rise in popularity of large-scale arrays. The substantial quantization distortion complicates the extraction of signal and noise subspaces from the quantized data. To address this issue, this paper introduces a novel approach that leverages the Transformer model to aid the subspace estimation. In this model, multiple snapshots are processed in parallel, enabling the capture of global correlations that span them. The learned subspace empowers us to construct the MUSIC spectrum and perform gridless DOA estimation using a neural network-based peak finder. Additionally, the acquired subspace encodes the vital information of model order, allowing us to determine the exact number of sources. These integrated components form a unified algorithmic framework referred to as TransMUSIC. Numerical results demonstrate the superiority of the TransMUSIC algorithm, even when dealing with one-bit quantized data. The results highlight the potential of Transformer-based techniques in DOA estimation.
Extreme edge platforms, such as in-vehicle smart devices, require efficient deployment of quantized deep neural networks (DNNs) to enable intelligent applications with limited amounts of energy, memory, and computing resources. However, many edge devices struggle to boost inference throughput of various quantized DNNs due to the varying quantization levels, and these devices lack floating-point (FP) support for on-device learning, which prevents them from improving model accuracy while ensuring data privacy. To tackle the challenges above, we propose a precision-scalable RISC-V DNN processor with on-device learning capability. It facilitates diverse precision levels of fixed-point DNN inference, spanning from 2-bit to 16-bit, and enhances on-device learning through improved support with FP16 operations. Moreover, we employ multiple methods such as FP16 multiplier reuse and multi-precision integer multiplier reuse, along with balanced mapping of FPGA resources, to significantly improve hardware resource utilization. Experimental results on the Xilinx ZCU102 FPGA show that our processor significantly improves inference throughput by 1.6$\sim$14.6$\times$ and energy efficiency by 1.1$\sim$14.6$\times$ across various DNNs, compared to the prior art, XpulpNN. Additionally, our processor achieves a 16.5$\times$ higher FP throughput for on-device learning.
In this paper, we propose methods to build a powerful and efficient Image-to-Speech captioning (Im2Sp) model. To this end, we start with importing the rich knowledge related to image comprehension and language modeling from a large-scale pre-trained vision-language model into Im2Sp. We set the output of the proposed Im2Sp as discretized speech units, i.e., the quantized speech features of a self-supervised speech model. The speech units mainly contain linguistic information while suppressing other characteristics of speech. This allows us to incorporate the language modeling capability of the pre-trained vision-language model into the spoken language modeling of Im2Sp. With the vision-language pre-training strategy, we set new state-of-the-art Im2Sp performances on two widely used benchmark databases, COCO and Flickr8k. Then, we further improve the efficiency of the Im2Sp model. Similar to the speech unit case, we convert the original image into image units, which are derived through vector quantization of the raw image. With these image units, we can drastically reduce the required data storage for saving image data to just 0.8% when compared to the original image data in terms of bits. Demo page: https://ms-dot-k.github.io/Image-to-Speech-Captioning.
The task of preserving privacy while ensuring efficient communication is a fundamental challenge in federated learning. In this work, we tackle this challenge in the trusted aggregator model, and propose a solution that achieves both objectives simultaneously. We show that employing a quantization scheme based on subtractive dithering at the clients can effectively replicate the normal noise addition process at the aggregator. This implies that we can guarantee the same level of differential privacy against other clients while substantially reducing the amount of communication required, as opposed to transmitting full precision gradients and using central noise addition. We also experimentally demonstrate that the accuracy of our proposed approach matches that of the full precision gradient method.
JPEG remains one of the most widespread lossy image coding methods. However, the non-differentiable nature of JPEG restricts the application in deep learning pipelines. Several differentiable approximations of JPEG have recently been proposed to address this issue. This paper conducts a comprehensive review of existing diff. JPEG approaches and identifies critical details that have been missed by previous methods. To this end, we propose a novel diff. JPEG approach, overcoming previous limitations. Our approach is differentiable w.r.t. the input image, the JPEG quality, the quantization tables, and the color conversion parameters. We evaluate the forward and backward performance of our diff. JPEG approach against existing methods. Additionally, extensive ablations are performed to evaluate crucial design choices. Our proposed diff. JPEG resembles the (non-diff.) reference implementation best, significantly surpassing the recent-best diff. approach by $3.47$dB (PSNR) on average. For strong compression rates, we can even improve PSNR by $9.51$dB. Strong adversarial attack results are yielded by our diff. JPEG, demonstrating the effective gradient approximation. Our code is available at https://github.com/necla-ml/Diff-JPEG.
Summary: Raw nanopore signals can be analyzed while they are being generated, a process known as real-time analysis. Real-time analysis of raw signals is essential to utilize the unique features that nanopore sequencing provides, enabling the early stopping of the sequencing of a read or the entire sequencing run based on the analysis. The state-of-the-art mechanism, RawHash, offers the first hash-based efficient and accurate similarity identification between raw signals and a reference genome by quickly matching their hash values. In this work, we introduce RawHash2, which provides major improvements over RawHash, including a more sensitive quantization and chaining implementation, weighted mapping decisions, frequency filters to reduce ambiguous seed hits, minimizers for hash-based sketching, and support for the R10.4 flow cell version and various data formats such as POD5 and SLOW5. Compared to RawHash, RawHash2 provides better F1 accuracy (on average by 10.57% and up to 20.25%) and better throughput (on average by 4.0x and up to 9.9x) than RawHash. Availability and Implementation: RawHash2 is available at https://github.com/CMU-SAFARI/RawHash. We also provide the scripts to fully reproduce our results on our GitHub page.
This work introduces a novel class of channel estimators tailored for coarse quantization systems. The proposed estimators are founded on conditionally Gaussian latent generative models, specifically Gaussian mixture models (GMMs), mixture of factor analyzers (MFAs), and variational autoencoders (VAEs). These models effectively learn the unknown channel distribution inherent in radio propagation scenarios, providing valuable prior information. Conditioning on the latent variable of these generative models yields a locally Gaussian channel distribution, thus enabling the application of the well-known Bussgang decomposition. By exploiting the resulting conditional Bussgang decomposition, we derive parameterized linear minimum mean square error (MMSE) estimators for the considered generative latent variable models. In this context, we explore leveraging model-based structural features to reduce memory and complexity overhead associated with the proposed estimators. Furthermore, we devise necessary training adaptations, enabling direct learning of the generative models from quantized pilot observations without requiring ground-truth channel samples during the training phase. Through extensive simulations, we demonstrate the superiority of our introduced estimators over existing state-of-the-art methods for coarsely quantized systems, as evidenced by significant improvements in mean square error (MSE) and achievable rate metrics.
Modulo sampling and dithered one-bit quantization frameworks have emerged as promising solutions to overcome the limitations of traditional analog-to-digital converters (ADCs) and sensors. Modulo sampling, with its high-resolution approach utilizing modulo ADCs, offers an unlimited dynamic range, while dithered one-bit quantization offers cost-efficiency and reduced power consumption while operating at elevated sampling rates. Our goal is to explore the synergies between these two techniques, leveraging their unique advantages, and to apply them to non-bandlimited signals within spline spaces. One noteworthy application of these signals lies in High Dynamic Range (HDR) imaging. In this paper, we expand upon the Unlimited One-Bit (UNO) sampling framework, initially conceived for bandlimited signals, to encompass non-bandlimited signals found in the context of HDR imaging. We present a novel algorithm rigorously examined for its ability to recover images from one-bit modulo samples. Additionally, we introduce a sufficient condition specifically designed for UNO sampling to perfectly recover non-bandlimited signals within spline spaces. Our numerical results vividly demonstrate the effectiveness of UNO sampling in the realm of HDR imaging.
High data rates require vast bandwidths, that can be found in the sub-THz band, and high sampling frequencies, which are predicted to lead to a problematically high analog-to-digital converter (ADC) power consumption. It was proposed to use 1-bit ADCs to mitigate this problem. Moreover, oscillator phase noise is predicted to be especially high at sub-THz carrier frequencies. For synchronization the phase must be tracked based on 1-bit quantized observations. We study iterative data-aided phase estimation, i.e., the expectation-maximization and the Fisher-scoring algorithm, compared to least-squares (LS) phase estimation. For phase interpolation at the data symbols, we consider the Kalman filter and the Rauch-Tung-Striebel algorithm. Compared to LS estimation, iterative phase noise tracking leads to a significantly lower estimation error variance at high signal-to-noise ratios. However, its benefit for the spectral efficiency using zero-crossing modulation (ZXM) is limited to marginal gains for high faster-than-Nyquist signaling factors, i.e., higher order ZXM modulation.
In this paper, we study unconstrained distributed optimization strongly convex problems, in which the exchange of information in the network is captured by a directed graph topology over digital channels that have limited capacity (and hence information should be quantized). Distributed methods in which nodes use quantized communication yield a solution at the proximity of the optimal solution, hence reaching an error floor that depends on the quantization level used; the finer the quantization the lower the error floor. However, it is not possible to determine in advance the optimal quantization level that ensures specific performance guarantees (such as achieving an error floor below a predefined threshold). Choosing a very small quantization level that would guarantee the desired performance, requires {information} packets of very large size, which is not desirable (could increase the probability of packet losses, increase delays, etc) and often not feasible due to the limited capacity of the channels available. In order to obtain a communication-efficient distributed solution and a sufficiently close proximity to the optimal solution, we propose a quantized distributed optimization algorithm that converges in a finite number of steps and is able to adjust the quantization level accordingly. The proposed solution uses a finite-time distributed optimization protocol to find a solution to the problem for a given quantization level in a finite number of steps and keeps refining the quantization level until the difference in the solution between two successive solutions with different quantization levels is below a certain pre-specified threshold.
Modulo sampling has recently drawn a great deal of attention for cutting-edge applications, due to overcoming the barrier of information loss through sensor saturation and clipping. This is a significant problem, especially when the range of signal amplitudes is unknown or in the near-far case. To overcome this fundamental bottleneck, we propose a one-bit-aided (1bit-aided) modulo sampling scheme for direction-of-arrival (DOA) estimation. On the one hand, one-bit quantization involving a simple comparator offers the advantages of low-cost and low-complexity implementation. On the other hand, one-bit quantization provides an estimate of the normalized covariance matrix of the unquantized measurements via the arcsin law. The estimate of the normalized covariance matrix is used to implement blind integer-forcing (BIF) decoder to unwrap the modulo samples to construct the covariance matrix, and subspace methods can be used to perform the DOA estimation. Our approach named as 1bit-aided-BIF addresses the near-far problem well and overcomes the intrinsic low dynamic range of one-bit quantization. Numerical experiments validate the excellent performance of the proposed algorithm.