Modern vision models typically rely on fine-tuning general-purpose models pre-trained on large, static datasets. These general-purpose models only capture the knowledge within their pre-training datasets, which are tiny, out-of-date snapshots of the Internet -- where billions of images are uploaded each day. We suggest an alternate approach: rather than hoping our static datasets transfer to our desired tasks after large-scale pre-training, we propose dynamically utilizing the Internet to quickly train a small-scale model that does extremely well on the task at hand. Our approach, called Internet Explorer, explores the web in a self-supervised manner to progressively find relevant examples that improve performance on a desired target dataset. It cycles between searching for images on the Internet with text queries, self-supervised training on downloaded images, determining which images were useful, and prioritizing what to search for next. We evaluate Internet Explorer across several datasets and show that it outperforms or matches CLIP oracle performance by using just a single GPU desktop to actively query the Internet for 30--40 hours. Results, visualizations, and videos at https://internet-explorer-ssl.github.io/
Due to the sweeping digitalization of processes, increasingly vast amounts of time series data are being produced. Accurate classification of such time series facilitates decision making in multiple domains. State-of-the-art classification accuracy is often achieved by ensemble learning where results are synthesized from multiple base models. This characteristic implies that ensemble learning needs substantial computing resources, preventing their use in resource-limited environments, such as in edge devices. To extend the applicability of ensemble learning, we propose the LightTS framework that compresses large ensembles into lightweight models while ensuring competitive accuracy. First, we propose adaptive ensemble distillation that assigns adaptive weights to different base models such that their varying classification capabilities contribute purposefully to the training of the lightweight model. Second, we propose means of identifying Pareto optimal settings w.r.t. model accuracy and model size, thus enabling users with a space budget to select the most accurate lightweight model. We report on experiments using 128 real-world time series sets and different types of base models that justify key decisions in the design of LightTS and provide evidence that LightTS is able to outperform competitors.
Deep learning has revolutionized the computer vision and image classification domains. In this context Convolutional Neural Networks (CNNs) based architectures are the most widely applied models. In this article, we introduced two procedures for training Convolutional Neural Networks (CNNs) and Deep Neural Network based on Gradient Boosting (GB), namely GB-CNN and GB-DNN. These models are trained to fit the gradient of the loss function or pseudo-residuals of previous models. At each iteration, the proposed method adds one dense layer to an exact copy of the previous deep NN model. The weights of the dense layers trained on previous iterations are frozen to prevent over-fitting, permitting the model to fit the new dense as well as to fine-tune the convolutional layers (for GB-CNN) while still utilizing the information already learned. Through extensive experimentation on different 2D-image classification and tabular datasets, the presented models show superior performance in terms of classification accuracy with respect to standard CNN and Deep-NN with the same architectures.
Deep learning provides an excellent avenue for optimizing diagnosis and patient monitoring for clinical-based applications, which can critically enhance the response time to the onset of various conditions. For cardiovascular disease, one such condition where the rising number of patients increasingly outweighs the availability of medical resources in different parts of the world, a core challenge is the automated classification of various cardiac abnormalities. Existing deep learning approaches have largely been limited to detecting the existence of an irregularity, as in binary classification, which has been achieved using networks such as CNNs and RNN/LSTMs. The next step is to accurately perform multi-class classification and determine the specific condition(s) from the inherently noisy multi-variate waveform, which is a difficult task that could benefit from (1) a more powerful sequential network, and (2) the integration of clinical notes, which provide valuable semantic and clinical context from human doctors. Recently, Transformers have emerged as the state-of-the-art architecture for forecasting and prediction using time-series data, with their multi-headed attention mechanism, and ability to process whole sequences and learn both long and short-range dependencies. The proposed novel multi-modal Transformer architecture would be able to accurately perform this task while demonstrating the cross-domain effectiveness of Transformers, establishing a method for incorporating multiple data modalities within a Transformer for classification tasks, and laying the groundwork for automating real-time patient condition monitoring in clinical and ER settings.
We present Classy Ensemble, a novel ensemble-generation algorithm for classification tasks, which aggregates models through a weighted combination of per-class accuracy. Tested over 153 machine learning datasets we demonstrate that Classy Ensemble outperforms two other well-known aggregation algorithms -- order-based pruning and clustering-based pruning -- as well as the recently introduced lexigarden ensemble generator. We then present three enhancements: 1) Classy Cluster Ensemble, which combines Classy Ensemble and cluster-based pruning; 2) Deep Learning experiments, showing the merits of Classy Ensemble over four image datasets: Fashion MNIST, CIFAR10, CIFAR100, and ImageNet; and 3) Classy Evolutionary Ensemble, wherein an evolutionary algorithm is used to select the set of models which Classy Ensemble picks from. This latter, combining learning and evolution, resulted in improved performance on the hardest dataset.
Social media classification tasks (e.g., tweet sentiment analysis, tweet stance detection) are challenging because social media posts are typically short, informal, and ambiguous. Thus, training on tweets is challenging and demands large-scale human-annotated labels, which are time-consuming and costly to obtain. In this paper, we find that providing hashtags to social media tweets can help alleviate this issue because hashtags can enrich short and ambiguous tweets in terms of various information, such as topic, sentiment, and stance. This motivates us to propose a novel Hashtag-guided Tweet Classification model (HashTation), which automatically generates meaningful hashtags for the input tweet to provide useful auxiliary signals for tweet classification. To generate high-quality and insightful hashtags, our hashtag generation model retrieves and encodes the post-level and entity-level information across the whole corpus. Experiments show that HashTation achieves significant improvements on seven low-resource tweet classification tasks, in which only a limited amount of training data is provided, showing that automatically enriching tweets with model-generated hashtags could significantly reduce the demand for large-scale human-labeled data. Further analysis demonstrates that HashTation is able to generate high-quality hashtags that are consistent with the tweets and their labels. The code is available at https://github.com/shizhediao/HashTation.
Theoretically-inspired sequential density ratio estimation (SDRE) algorithms are proposed for the early classification of time series. Conventional SDRE algorithms can fail to estimate DRs precisely due to the internal overnormalization problem, which prevents the DR-based sequential algorithm, Sequential Probability Ratio Test (SPRT), from reaching its asymptotic Bayes optimality. Two novel SPRT-based algorithms, B2Bsqrt-TANDEM and TANDEMformer, are designed to avoid the overnormalization problem for precise unsupervised regression of SDRs. The two algorithms statistically significantly reduce DR estimation errors and classification errors on an artificial sequential Gaussian dataset and real datasets (SiW, UCF101, and HMDB51), respectively. The code is available at: https://github.com/Akinori-F-Ebihara/LLR_saturation_problem.
For extreme multi-label classification (XMC), existing classification-based models poorly perform for tail labels and often ignore the semantic relations among labels, like treating "Wikipedia" and "Wiki" as independent and separate labels. In this paper, we cast XMC as a generation task (XLGen), where we benefit from pre-trained text-to-text models. However, generating labels from the extremely large label space is challenging without any constraints or guidance. We, therefore, propose to guide label generation using label cluster information to hierarchically generate lower-level labels. We also find that frequency-based label ordering and using decoding ensemble methods are critical factors for the improvements in XLGen. XLGen with cluster guidance significantly outperforms the classification and generation baselines on tail labels, and also generally improves the overall performance in four popular XMC benchmarks. In human evaluation, we also find XLGen generates unseen but plausible labels. Our code is now available at https://github.com/alexa/xlgen-eacl-2023.
With generative models proliferating at a rapid rate, there is a growing need for general purpose fake image detectors. In this work, we first show that the existing paradigm, which consists of training a deep network for real-vs-fake classification, fails to detect fake images from newer breeds of generative models when trained to detect GAN fake images. Upon analysis, we find that the resulting classifier is asymmetrically tuned to detect patterns that make an image fake. The real class becomes a sink class holding anything that is not fake, including generated images from models not accessible during training. Building upon this discovery, we propose to perform real-vs-fake classification without learning; i.e., using a feature space not explicitly trained to distinguish real from fake images. We use nearest neighbor and linear probing as instantiations of this idea. When given access to the feature space of a large pretrained vision-language model, the very simple baseline of nearest neighbor classification has surprisingly good generalization ability in detecting fake images from a wide variety of generative models; e.g., it improves upon the SoTA by +15.07 mAP and +25.90% acc when tested on unseen diffusion and autoregressive models.
Few-shot text classification systems have impressive capabilities but are infeasible to deploy and use reliably due to their dependence on prompting and billion-parameter language models. SetFit (Tunstall et al., 2022) is a recent, practical approach that fine-tunes a Sentence Transformer under a contrastive learning paradigm and achieves similar results to more unwieldy systems. Inexpensive text classification is important for addressing the problem of domain drift in all classification tasks, and especially in detecting harmful content, which plagues social media platforms. Here, we propose Like a Good Nearest Neighbor (LaGoNN), a modification to SetFit that introduces no learnable parameters but alters input text with information from its nearest neighbor, for example, the label and text, in the training data, making novel data appear similar to an instance on which the model was optimized. LaGoNN is effective at flagging undesirable content and text classification, and improves the performance of SetFit. To demonstrate the value of LaGoNN, we conduct a thorough study of text classification systems in the context of content moderation under four label distributions, and in general and multilingual classification settings.
Weakly-supervised classification of histopathology slides is a computationally intensive task, with a typical whole slide image (WSI) containing billions of pixels to process. We propose Discriminative Region Active Sampling for Multiple Instance Learning (DRAS-MIL), a computationally efficient slide classification method using attention scores to focus sampling on highly discriminative regions. We apply this to the diagnosis of ovarian cancer histological subtypes, which is an essential part of the patient care pathway as different subtypes have different genetic and molecular profiles, treatment options, and patient outcomes. We use a dataset of 714 WSIs acquired from 147 epithelial ovarian cancer patients at Leeds Teaching Hospitals NHS Trust to distinguish the most common subtype, high-grade serous carcinoma, from the other four subtypes (low-grade serous, endometrioid, clear cell, and mucinous carcinomas) combined. We demonstrate that DRAS-MIL can achieve similar classification performance to exhaustive slide analysis, with a 3-fold cross-validated AUC of 0.8679 compared to 0.8781 with standard attention-based MIL classification. Our approach uses at most 18% as much memory as the standard approach, while taking 33% of the time when evaluating on a GPU and only 14% on a CPU alone. Reducing prediction time and memory requirements may benefit clinical deployment and the democratisation of AI, reducing the extent to which computational hardware limits end-user adoption.
DNA sequence classification is a fundamental task in computational biology with vast implications for applications such as disease prevention and drug design. Therefore, fast high-quality sequence classifiers are significantly important. This paper introduces ClaPIM, a scalable DNA sequence classification architecture based on the emerging concept of hybrid in-crossbar and near-crossbar memristive processing-in-memory (PIM). We enable efficient and high-quality classification by uniting the filter and search stages within a single algorithm. Specifically, we propose a custom filtering technique that drastically narrows the search space and a search approach that facilitates approximate string matching through a distance function. ClaPIM is the first PIM architecture for scalable approximate string matching that benefits from the high density of memristive crossbar arrays and the massive computational parallelism of PIM. Compared with Kraken2, a state-of-the-art software classifier, ClaPIM provides significantly higher classification quality (up to 20x improvement in F1 score) and also demonstrates a 1.8x throughput improvement. Compared with EDAM, a recently-proposed SRAM-based accelerator that is restricted to small datasets, we observe both a 30.4x improvement in normalized throughput per area and a 7% increase in classification precision.
Filter pruning has attracted increasing attention in recent years for its capacity in compressing and accelerating convolutional neural networks. Various data-independent criteria, including norm-based and relationship-based ones, were proposed to prune the most unimportant filters. However, these state-of-the-art criteria fail to fully consider the dissimilarity of filters, and thus might lead to performance degradation. In this paper, we first analyze the limitation of relationship-based criteria with examples, and then introduce a new data-independent criterion, Weighted Hybrid Criterion (WHC), to tackle the problems of both norm-based and relationship-based criteria. By taking the magnitude of each filter and the linear dependence between filters into consideration, WHC can robustly recognize the most redundant filters, which can be safely pruned without introducing severe performance degradation to networks. Extensive pruning experiments in a simple one-shot manner demonstrate the effectiveness of the proposed WHC. In particular, WHC can prune ResNet-50 on ImageNet with more than 42% of floating point operations reduced without any performance loss in top-5 accuracy.
Water is a necessary fluid to the human body and automatic checking of its quality and cleanness is an ongoing area of research. One such approach is to present the liquid to various types of signals and make the amount of signal attenuation an indication of the liquid category. In this article, we have utilized the Wi-Fi signal to distinguish clean water from poisoned water via training different machine learning algorithms. The Wi-Fi access points (WAPs) signal is acquired via equivalent smartphone-embedded Wi-Fi chipsets, and then Channel-State-Information CSI measures are extracted and converted into feature vectors to be used as input for machine learning classification algorithms. The measured amplitude and phase of the CSI data are selected as input features into four classifiers k-NN, SVM, LSTM, and Ensemble. The experimental results show that the model is adequate to differentiate poison water from clean water with a classification accuracy of 89% when LSTM is applied, while 92% classification accuracy is achieved when the AdaBoost-Ensemble classifier is applied.
While pre-trained language models (PLMs) have become a de-facto standard promoting the accuracy of text classification tasks, recent studies find that PLMs often predict over-confidently. Although various calibration methods have been proposed, such as ensemble learning and data augmentation, most of the methods have been verified in computer vision benchmarks rather than in PLM-based text classification tasks. In this paper, we present an empirical study on confidence calibration for PLMs, addressing three categories, including confidence penalty losses, data augmentations, and ensemble methods. We find that the ensemble model overfitted to the training set shows sub-par calibration performance and also observe that PLMs trained with confidence penalty loss have a trade-off between calibration and accuracy. Building on these observations, we propose the Calibrated PLM (CALL), a combination of calibration techniques. The CALL complements the drawbacks that may occur when utilizing a calibration method individually and boosts both classification and calibration accuracy. Design choices in CALL's training procedures are extensively studied, and we provide a detailed analysis of how calibration techniques affect the calibration performance of PLMs.
The tremendous growth in smart devices has uplifted several security threats. One of the most prominent threats is malicious software also known as malware. Malware has the capability of corrupting a device and collapsing an entire network. Therefore, its early detection and mitigation are extremely important to avoid catastrophic effects. In this work, we came up with a solution for malware detection using state-of-the-art natural language processing (NLP) techniques. Our main focus is to provide a lightweight yet effective classifier for malware detection which can be used for heterogeneous devices, be it a resource constraint device or a resourceful machine. Our proposed model is tested on the benchmark data set with an accuracy and log loss score of 99.13 percent and 0.04 respectively.
When users can benefit from certain predictive outcomes, they may be prone to act to achieve those outcome, e.g., by strategically modifying their features. The goal in strategic classification is therefore to train predictive models that are robust to such behavior. However, the conventional framework assumes that changing features does not change actual outcomes, which depicts users as "gaming" the system. Here we remove this assumption, and study learning in a causal strategic setting where true outcomes do change. Focusing on accuracy as our primary objective, we show how strategic behavior and causal effects underlie two complementing forms of distribution shift. We characterize these shifts, and propose a learning algorithm that balances between these two forces and over time, and permits end-to-end training. Experiments on synthetic and semi-synthetic data demonstrate the utility of our approach.
For the stable optimization of deep neural networks, regularization methods such as dropout and batch normalization have been used in various tasks. Nevertheless, the correct position to apply dropout has rarely been discussed, and different positions have been employed depending on the practitioners. In this study, we investigate the correct position to apply dropout. We demonstrate that for a residual network with batch normalization, applying dropout at certain positions increases the performance, whereas applying dropout at other positions decreases the performance. Based on theoretical analysis, we provide the following guideline for the correct position to apply dropout: apply one dropout after the last batch normalization but before the last weight layer in the residual branch. We provide detailed theoretical explanations to support this claim and demonstrate them through module tests. In addition, we investigate the correct position of dropout in the head that produces the final prediction. Although the current consensus is to apply dropout after global average pooling, we prove that applying dropout before global average pooling leads to a more stable output. The proposed guidelines are validated through experiments using different datasets and models.
Error-correcting codes (ECC) are used to reduce multiclass classification tasks to multiple binary classification subproblems. In ECC, classes are represented by the rows of a binary matrix, corresponding to codewords in a codebook. Codebooks are commonly either predefined or problem dependent. Given predefined codebooks, codeword-to-class assignments are traditionally overlooked, and codewords are implicitly assigned to classes arbitrarily. Our paper shows that these assignments play a major role in the performance of ECC. Specifically, we examine similarity-preserving assignments, where similar codewords are assigned to similar classes. Addressing a controversy in existing literature, our extensive experiments confirm that similarity-preserving assignments induce easier subproblems and are superior to other assignment policies in terms of their generalization performance. We find that similarity-preserving assignments make predefined codebooks become problem-dependent, without altering other favorable codebook properties. Finally, we show that our findings can improve predefined codebooks dedicated to extreme classification.
Intra-class variations in the open world lead to various challenges in classification tasks. To overcome these challenges, fine-grained classification was introduced, and many approaches were proposed. Some rely on locating and using distinguishable local parts within images to achieve invariance to viewpoint changes, intra-class differences, and local part deformations. Our approach, which is inspired by P2P-Net, offers an end-to-end trainable attention-based parts alignment module, where we replace the graph-matching component used in it with a self-attention mechanism. The attention module is able to learn the optimal arrangement of parts while attending to each other, before contributing to the global loss.