Large Language Models (LLMs) like BERT have gained significant prominence due to their remarkable performance in various natural language processing tasks. However, they come with substantial computational and memory costs. Additionally, they are essentially black-box models, challenging to explain and interpret. In this article, we propose Optimus BERT Compression and Explainability (OBCE), a methodology to bring explainability to BERT models using persistent homology, aiming to measure the importance of each neuron by studying the topological characteristics of their outputs. As a result, we can compress BERT significantly by reducing the number of parameters (58.47% of the original parameters for BERT Base, 52.3% for BERT Large). We evaluated our methodology on the standard GLUE Benchmark, comparing the results with state-of-the-art techniques and achieving outstanding results. Consequently, our methodology can "whiten" BERT models by providing explainability to its neurons and reducing the model's size, making it more suitable for deployment on resource-constrained devices.
相關內容
Wordle, a word guessing game rose to global popularity in the January of 2022. The goal of the game is to guess a five-letter English word within six tries. Each try provides the player with hints by means of colour changing tiles which inform whether or not a given character is part of the solution as well as, in cases where it is part of the solution, whether or not it is in the correct placement. Numerous attempts have been made to find the best starting word and best strategy to solve the daily wordle. This study uses character statistics of five-letter words to determine the best three starting words.
Which phonemes convey more speaker traits is a long-standing question, and various perception experiments were conducted with human subjects. For speaker recognition, studies were conducted with the conventional statistical models and the drawn conclusions are more or less consistent with the perception results. However, which phonemes are more important with modern deep neural models is still unexplored, due to the opaqueness of the decision process. This paper conducts a novel study for the attribution of phonemes with two types of deep speaker models that are based on TDNN and CNN respectively, from the perspective of model explanation. Specifically, we conducted the study by two post-explanation methods: LayerCAM and Time Align Occlusion (TAO). Experimental results showed that: (1) At the population level, vowels are more important than consonants, confirming the human perception studies. However, fricatives are among the most unimportant phonemes, which contrasts with previous studies. (2) At the speaker level, a large between-speaker variation is observed regarding phoneme importance, indicating that whether a phoneme is important or not is largely speaker-dependent.
Support Vector Machines (SVMs) are an important tool for performing classification on scattered data, where one usually has to deal with many data points in high-dimensional spaces. We propose solving SVMs in primal form using feature maps based on trigonometric functions or wavelets. In small dimensional settings the Fast Fourier Transform (FFT) and related methods are a powerful tool in order to deal with the considered basis functions. For growing dimensions the classical FFT-based methods become inefficient due to the curse of dimensionality. Therefore, we restrict ourselves to multivariate basis functions, each one of them depends only on a small number of dimensions. This is motivated by the well-known sparsity of effects and recent results regarding the reconstruction of functions from scattered data in terms of truncated analysis of variance (ANOVA) decomposition, which makes the resulting model even interpretable in terms of importance of the features as well as their couplings. The usage of small superposition dimensions has the consequence that the computational effort no longer grows exponentially but only polynomially with respect to the dimension. In order to enforce sparsity regarding the basis coefficients, we use the frequently applied $\ell_2$-norm and, in addition, $\ell_1$-norm regularization. The found classifying function, which is the linear combination of basis functions, and its variance can then be analyzed in terms of the classical ANOVA decomposition of functions. Based on numerical examples we show that we are able to recover the signum of a function that perfectly fits our model assumptions. We obtain better results with $\ell_1$-norm regularization, both in terms of accuracy and clarity of interpretability.
We consider a three-block alternating direction method of multipliers (ADMM) for solving the nonconvex nonseparable optimization problem with linear constraint. Inspired by [1], the third variable is updated twice in each iteration to ensure the global convergence. Based on the powerful Kurdyka-Lojasiewicz property, we prove that the sequence generated by the ADMM converges globally to the critical point of the augmented Lagrangian function. We also point out the convergence of proposed ADMM with swapping the update order of the first and second variables, and with adding a proximal term to the first variable for more general nonseparable problems, respectively. Moreover, we make numerical experiments on three nonconvex problems: multiple measurement vector (MMV), robust PCA (RPCA) and nonnegative matrix completion (NMC). The results show the efficiency and outperformance of proposed ADMM.
In recent years, Large Language Models (LLMs) have achieved significant success in natural language processing (NLP) and various interdisciplinary areas. However, applying LLMs to chemistry is a complex task that requires specialized domain knowledge. This paper provides a thorough exploration of the nuanced methodologies employed in integrating LLMs into the field of chemistry, delving into the complexities and innovations at this interdisciplinary juncture. Specifically, our analysis begins with examining how molecular information is fed into LLMs through various representation and tokenization methods. We then categorize chemical LLMs into three distinct groups based on the domain and modality of their input data, and discuss approaches for integrating these inputs for LLMs. Furthermore, this paper delves into the pretraining objectives with adaptations to chemical LLMs. After that, we explore the diverse applications of LLMs in chemistry, including novel paradigms for their application in chemistry tasks. Finally, we identify promising research directions, including further integration with chemical knowledge, advancements in continual learning, and improvements in model interpretability, paving the way for groundbreaking developments in the field.
Extremely large aperture arrays can enable unprecedented spatial multiplexing in beyond 5G systems due to their extremely narrow beamfocusing capabilities. However, acquiring the spatial correlation matrix to enable efficient channel estimation is a complex task due to the vast number of antenna dimensions. Recently, a new estimation method called the "reduced-subspace least squares (RS-LS) estimator" has been proposed for densely packed arrays. This method relies solely on the geometry of the array to limit the estimation resources. In this paper, we address a gap in the existing literature by deriving the average spectral efficiency for a certain distribution of user equipments (UEs) and a lower bound on it when using the RS-LS estimator. This bound is determined by the channel gain and the statistics of the normalized spatial correlation matrices of potential UEs but, importantly, does not require knowledge of a specific UE's spatial correlation matrix. We establish that there exists a pilot length that maximizes this expression. Additionally, we derive an approximate expression for the optimal pilot length under low signal-to-noise ratio (SNR) conditions. Simulation results validate the tightness of the derived lower bound and the effectiveness of using the optimized pilot length.
Recently pre-trained language representation models such as BERT have shown great success when fine-tuned on downstream tasks including information retrieval (IR). However, pre-training objectives tailored for ad-hoc retrieval have not been well explored. In this paper, we propose Pre-training with Representative wOrds Prediction (PROP) for ad-hoc retrieval. PROP is inspired by the classical statistical language model for IR, specifically the query likelihood model, which assumes that the query is generated as the piece of text representative of the "ideal" document. Based on this idea, we construct the representative words prediction (ROP) task for pre-training. Given an input document, we sample a pair of word sets according to the document language model, where the set with higher likelihood is deemed as more representative of the document. We then pre-train the Transformer model to predict the pairwise preference between the two word sets, jointly with the Masked Language Model (MLM) objective. By further fine-tuning on a variety of representative downstream ad-hoc retrieval tasks, PROP achieves significant improvements over baselines without pre-training or with other pre-training methods. We also show that PROP can achieve exciting performance under both the zero- and low-resource IR settings. The code and pre-trained models are available at //github.com/Albert-Ma/PROP.
Graph Neural Networks (GNNs) have recently become increasingly popular due to their ability to learn complex systems of relations or interactions arising in a broad spectrum of problems ranging from biology and particle physics to social networks and recommendation systems. Despite the plethora of different models for deep learning on graphs, few approaches have been proposed thus far for dealing with graphs that present some sort of dynamic nature (e.g. evolving features or connectivity over time). In this paper, we present Temporal Graph Networks (TGNs), a generic, efficient framework for deep learning on dynamic graphs represented as sequences of timed events. Thanks to a novel combination of memory modules and graph-based operators, TGNs are able to significantly outperform previous approaches being at the same time more computationally efficient. We furthermore show that several previous models for learning on dynamic graphs can be cast as specific instances of our framework. We perform a detailed ablation study of different components of our framework and devise the best configuration that achieves state-of-the-art performance on several transductive and inductive prediction tasks for dynamic graphs.
Although measuring held-out accuracy has been the primary approach to evaluate generalization, it often overestimates the performance of NLP models, while alternative approaches for evaluating models either focus on individual tasks or on specific behaviors. Inspired by principles of behavioral testing in software engineering, we introduce CheckList, a task-agnostic methodology for testing NLP models. CheckList includes a matrix of general linguistic capabilities and test types that facilitate comprehensive test ideation, as well as a software tool to generate a large and diverse number of test cases quickly. We illustrate the utility of CheckList with tests for three tasks, identifying critical failures in both commercial and state-of-art models. In a user study, a team responsible for a commercial sentiment analysis model found new and actionable bugs in an extensively tested model. In another user study, NLP practitioners with CheckList created twice as many tests, and found almost three times as many bugs as users without it.
We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models, BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5% (7.7% point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement).
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191