Recent years have witnessed the breakthrough success of deep convolutional neural networks (DCNNs) in image classification and other vision applications. Although freeing users from the troublesome handcrafted feature extraction by providing a uniform feature extraction-classification framework, DCNNs still require a handcrafted design of their architectures. In this paper, we propose the genetic DCNN designer, an autonomous learning algorithm can generate a DCNN architecture automatically based on the data available for a specific image classification problem. We first partition a DCNN into multiple stacked meta convolutional blocks and fully connected blocks, each containing the operations of convolution, pooling, fully connection, batch normalization, activation and drop out, and thus convert the architecture into an integer vector. Then, we use refined evolutionary operations, including selection, mutation and crossover to evolve a population of DCNN architectures. Our results on the MNIST, Fashion-MNIST, EMNISTDigit, EMNIST-Letter, CIFAR10 and CIFAR100 datasets suggest that the proposed genetic DCNN designer is able to produce automatically DCNN architectures, whose performance is comparable to, if not better than, that of stateof- the-art DCNN models
Convolutional neural networks (CNN) have been shown to provide a good solution for classification problems that utilize data obtained from vibrational spectroscopy. Moreover, CNNs are capable of identification from noisy spectra without the need for additional preprocessing. However, their application in practical spectroscopy is limited due to two shortcomings. The effectiveness of the classification using CNNs drops rapidly when only a small number of spectra per substance are available for training (which is a typical situation in real applications). Additionally, to accommodate new, previously unseen substance classes, the network must be retrained which is computationally intensive. Here we address these issues by reformulating a multi-class classification problem with a large number of classes, but a small number of samples per class, to a binary classification problem with sufficient data available for representation learning. Namely, we define the learning task as identifying pairs of inputs as belonging to the same or different classes. We achieve this using a Siamese convolutional neural network. A novel sampling strategy is proposed to address the imbalance problem in training the Siamese Network. The trained network can effectively classify samples of unseen substance classes using just a single reference sample (termed as one-shot learning in the machine learning community). Our results demonstrate better accuracy than other practical systems to date, while allowing effortless updates of the system's database with novel substance classes.
Interpretable deep learning is a fundamental building block towards safer AI, especially when the deployment possibilities of deep learning-based computer-aided medical diagnostic systems are so eminent. However, without a computational formulation of black-box interpretation, general interpretability research rely heavily on subjective bias. Clear decision structure of the medical diagnostics lets us approximate the decision process of a radiologist as a model - removed from subjective bias. We define the process of interpretation as a finite communication between a known model and a black-box model to optimally map the black box's decision process in the known model. Consequently, we define interpretability as maximal information gain over the initial uncertainty about the black-box's decision within finite communication. We relax this definition based on the observation that diagnostic interpretation is typically achieved by a process of minimal querying. We derive an algorithm to calculate diagnostic interpretability. The usual question of accuracy-interpretability tradeoff, i.e. whether a black-box model's prediction accuracy is dependent on its ability to be interpreted by a known source model, does not arise in this theory. With multiple example simulation experiments of various complexity levels, we demonstrate the working of such a theoretical model in synthetic supervised classification scenarios.
Natural language explanations of deep neural network decisions provide an intuitive way for a AI agent to articulate a reasoning process. Current textual explanations learn to discuss class discriminative features in an image. However, it is also helpful to understand which attributes might change a classification decision if present in an image (e.g., "This is not a Scarlet Tanager because it does not have black wings.") We call such textual explanations counterfactual explanations, and propose an intuitive method to generate counterfactual explanations by inspecting which evidence in an input is missing, but might contribute to a different classification decision if present in the image. To demonstrate our method we consider a fine-grained image classification task in which we take as input an image and a counterfactual class and output text which explains why the image does not belong to a counterfactual class. We then analyze our generated counterfactual explanations both qualitatively and quantitatively using proposed automatic metrics.
Unsupervised transfer of object recognition models from synthetic to real data is an important problem with many potential applications. The challenge is how to "adapt" a model trained on simulated images so that it performs well on real-world data without any additional supervision. Unfortunately, current benchmarks for this problem are limited in size and task diversity. In this paper, we present a new large-scale benchmark called Syn2Real, which consists of a synthetic domain rendered from 3D object models and two real-image domains containing the same object categories. We define three related tasks on this benchmark: closed-set object classification, open-set object classification, and object detection. Our evaluation of multiple state-of-the-art methods reveals a large gap in adaptation performance between the easier closed-set classification task and the more difficult open-set and detection tasks. We conclude that developing adaptation methods that work well across all three tasks presents a significant future challenge for syn2real domain transfer.
A picture is worth a thousand words. Albeit a clich\'e, for the fashion industry, an image of a clothing piece allows one to perceive its category (e.g., dress), sub-category (e.g., day dress) and properties (e.g., white colour with floral patterns). The seasonal nature of the fashion industry creates a highly dynamic and creative domain with evermore data, making it unpractical to manually describe a large set of images (of products). In this paper, we explore the concept of visual recognition for fashion images through an end-to-end architecture embedding the hierarchical nature of the annotations directly into the model. Towards that goal, and inspired by the work of [7], we have modified and adapted the original architecture proposal. Namely, we have removed the message passing layer symmetry to cope with Farfetch category tree, added extra layers for hierarchy level specificity, and moved the message passing layer into an enriched latent space. We compare the proposed unified architecture against state-of-the-art models and demonstrate the performance advantage of our model for structured multi-level categorization on a dataset of about 350k fashion product images.
An unsupervised learning classification model is described. It achieves classification error probability competitive with that of popular supervised learning classifiers such as SVM or kNN. The model is based on the incremental execution of small step shift and rotation operations upon selected discriminative hyperplanes at the arrival of input samples. When applied, in conjunction with a selected feature extractor, to a subset of the ImageNet dataset benchmark, it yields 6.2 % Top 3 probability of error; this exceeds by merely about 2 % the result achieved by (supervised) k-Nearest Neighbor, both using same feature extractor. This result may also be contrasted with popular unsupervised learning schemes such as k-Means which is shown to be practically useless on same dataset.
We show how the problem of estimating conditional Kendall's tau can be rewritten as a classification task. Conditional Kendall's tau is a conditional dependence parameter that is a characteristic of a given pair of random variables. The goal is to predict whether the pair is concordant (value of $1$) or discordant (value of $-1$) conditionally on some covariates. We prove the consistency and the asymptotic normality of a family of penalized approximate maximum likelihood estimators, including the equivalent of the logit and probit regressions in our framework. Then, we detail specific algorithms adapting usual machine learning techniques, including nearest neighbors, decision trees, random forests and neural networks, to the setting of the estimation of conditional Kendall's tau. Finite sample properties of these estimators and their sensitivities to each component of the data-generating process are assessed in a simulation study. Finally, we apply all these estimators to a dataset of European stock indices.
Deep learning (DL) techniques have shown unprecedented success when applied to images, waveforms, and text. Generally, when the sample size ($N$) is much bigger than the number of features ($d$), DL often outperforms other machine learning (ML) techniques, often through the use of Convolutional Neural Networks (CNNs). However, in many bioinformatics fields (including metagenomics), we encounter the opposite situation where $d$ is significantly greater than $N$. In these situations, applying DL techniques would lead to severe overfitting. Here we aim to improve classification of various diseases with metagenomic data through the use of CNNs. For this we proposed to represent metagenomic data as images. The proposed Met2Img approach relies on taxonomic and t-SNE embeddings to transform abundance data into "synthetic images". We applied our approach to twelve benchmark data sets including more than 1400 metagenomic samples. Our results show significant improvements over the state-of-the-art algorithms (Random Forest (RF), Support Vector Machine (SVM)). We observe that the integration of phylogenetic information alongside abundance data improves classification. The proposed approach is not only important in classification setting but also allows to visualize complex metagenomic data. The Met2Img is implemented in Python.
There is a growing body of literature showing that deep neural networks are vulnerable to adversarial input modification. Recently this work has been extended from image classification to malware classification over boolean features. In this paper we present several new methods for training restricted networks in this specific domain that are highly effective at preventing adversarial perturbations. We start with a fully adversarially resistant neural network that has hard non-negative weight restrictions and is equivalent to learning a monotonic boolean function and then attempt to relax the constraints to improve classifier accuracy.
It has been shown that for automated PAP-smear image classification, nucleus features can be very informative. Therefore, the primary step for automated screening can be cell-nuclei detection followed by segmentation of nuclei in the resulting single cell PAP-smear images. We propose a patch based approach using CNN for segmentation of nuclei in single cell images. We then pose the question of ion of segmentation for classification using representation learning with CNN, and whether low-level CNN features may be useful for classification. We suggest a CNN-based feature level analysis and a transfer learning based approach for classification using both segmented as well full single cell images. We also propose a decision-tree based approach for classification. Experimental results demonstrate the effectiveness of the proposed algorithms individually (with low-level CNN features), and simultaneously proving the sufficiency of cell-nuclei detection (rather than accurate segmentation) for classification. Thus, we propose a system for analysis of multi-cell PAP-smear images consisting of a simple nuclei detection algorithm followed by classification using transfer learning.
Current approaches to speech emotion recognition focus on speech features that can capture the emotional content of a speech signal. Mel Frequency Cepstral Coefficients (MFCCs) are one of the most commonly used representations for audio speech recognition and classification. This paper proposes Gammatone Frequency Cepstral Coefficients (GFCCs) as a potentially better representation of speech signals for emotion recognition. The effectiveness of MFCC and GFCC representations are compared and evaluated over emotion and intensity classification tasks with fully connected and recurrent neural network architectures. The results provide evidence that GFCCs outperform MFCCs in speech emotion recognition.
There has been growing interest in developing accurate models that can also be explained to humans. Unfortunately, if there exist multiple distinct but accurate models for some dataset, current machine learning methods are unlikely to find them: standard techniques will likely recover a complex model that combines them. In this work, we introduce a way to identify a maximal set of distinct but accurate models for a dataset. We demonstrate empirically that, in situations where the data supports multiple accurate classifiers, we tend to recover simpler, more interpretable classifiers rather than more complex ones.
Many people are currently suffering from heart diseases that can lead to untimely death. The most common heart abnormality is arrhythmia, which is simply irregular beating of the heart. A prediction system for the early intervention and prevention of heart diseases, including cardiovascular diseases (CDVs) and arrhythmia, is important. This paper introduces the classification of electrocardiogram (ECG) heartbeats into normal or abnormal. The approach is based on the combination of swarm optimization algorithms with a modified PannTompkins algorithm (MPTA) and support vector machines (SVMs). The MPTA was implemented to remove ECG noise, followed by the application of the extended features extraction algorithm (EFEA) for ECG feature extraction. Then, elephant herding optimization (EHO) was used to find a subset of ECG features from a larger feature pool that provided better classification performance than that achieved using the whole set. Finally, SVMs were used for classification. The results show that the EHOSVM approach achieved good classification results in terms of five statistical indices: accuracy, 93.31%; sensitivity, 45.49%; precision, 46.45%; F-measure, 45.48%; and specificity, 45.48%. Furthermore, the results demonstrate a clear improvement in accuracy compared to that of other methods when applied to the MITBIH arrhythmia database.
There are many applications where it important to perform well on a set of examples as opposed to individual examples. For example in image or video classification the question is does an object appear somewhere in the image or video while there are several candidates of the object per image or video. In this context, it is not important what is the performance per candidate. Instead the performance per group is the ultimate objective. For such problems one popular approach assumes weak supervision where labels exist for the entire group and then multiple instance learning is utilized. Another approach is to optimize per candidate, assuming each candidate is labeled, in the belief that this will achieve good performance per group. We will show that better results can be achieved if we offer a new methodology which synthesizes the aforementioned approaches and directly optimizes for the final optimization objective while consisting of a convex optimization problem which solves the global optimization problem. The benefit of grouping examples is demonstrated on an image classification task for detecting polyps in images from capsule endoscopy of the colon. The algorithm was designed to efficiently handle hundreds of millions of examples. Furthermore, modifications to the penalty function of the standard SVM algorithm, have proven to significantly improve performance in our test case.
Recognition of surgical gesture is crucial for surgical skill assessment and efficient surgery training. Prior works on this task are based on either variant graphical models such as HMMs and CRFs, or deep learning models such as Recurrent Neural Networks and Temporal Convolutional Networks. Most of the current approaches usually suffer from over-segmentation and therefore low segment-level edit scores. In contrast, we present an essentially different methodology by modeling the task as a sequential decision-making process. An intelligent agent is trained using reinforcement learning with hierarchical features from a deep model. Temporal consistency is integrated into our action design and reward mechanism to reduce over-segmentation errors. Experiments on JIGSAWS dataset demonstrate that the proposed method performs better than state-of-the-art methods in terms of the edit score and on par in frame-wise accuracy. Our code will be released later.
Classification techniques for images of handwritten characters are susceptible to noise. Quadtrees can be an efficient representation for learning from sparse features. In this paper, we improve the effectiveness of probabilistic quadtrees by using a pixel level classifier to extract the character pixels and remove noise from handwritten character images. The pixel level denoiser (a deep belief network) uses the map responses obtained from a pretrained CNN as features for reconstructing the characters eliminating noise. We experimentally demonstrate the effectiveness of our approach by reconstructing and classifying a noisy version of handwritten Bangla Numeral and Basic Character datasets.
In this paper we determine how multi-layer ensembling improves performance on multilingual intent classification. We develop a novel multi-layer ensembling approach that ensembles both different model initializations and different model architectures. We also introduce a new banking domain dataset and compare results against the standard ATIS dataset and the Chinese SMP2017 dataset to determine ensembling performance in multilingual and multi-domain contexts. We run ensemble experiments across all three datasets, and conclude that ensembling provides significant performance increases, and that multi-layer ensembling is a no-risk way to improve performance on intent classification. We also find that a diverse ensemble of simple models can reach perform comparable to much more sophisticated state-of-the-art models. Our best F 1 scores on ATIS, Banking, and SMP are 97.54%, 91.79%, and 93.55% respectively, which compare well with the state-of-the-art on ATIS and best submission to the SMP2017 competition. The total ensembling performance increases we achieve are 0.23%, 1.96%, and 4.04% F 1 respectively.
To test the effectiveness of process discovery algorithms, a Process Discovery Contest (PDC) has been set up. This PDC uses a classification approach to measure this effectiveness: The better the discovered model can classify whether or not a new trace conforms to the event log, the better the discovery algorithm is supposed to be. Unfortunately, even the state-of-the-art fully-automated discovery algorithms score poorly on this classification. Even the best of these algorithms, the Inductive Miner, scored only 147 correct classified traces out of 200 traces on the PDC of 2017. This paper introduces the rule-based log skeleton model, which is closely related to the Declare constraint model, together with a way to classify traces using this model. This classification using log skeletons is shown to score better on the PDC of 2017 than state-of-the-art discovery algorithms: 194 out of 200. As a result, one can argue that the fully-automated algorithm to construct (or: discover) a log skeleton from an event log outperforms existing state-of-the-art fully-automated discovery algorithms.
The work in this paper is driven by the question if spatio-temporal correlations are enough for 3D convolutional neural networks (CNN)? Most of the traditional 3D networks use local spatio-temporal features. We introduce a new block that models correlations between channels of a 3D CNN with respect to temporal and spatial features. This new block can be added as a residual unit to different parts of 3D CNNs. We name our novel block 'Spatio-Temporal Channel Correlation' (STC). By embedding this block to the current state-of-the-art architectures such as ResNext and ResNet, we improved the performance by 2-3\% on Kinetics dataset. Our experiments show that adding STC blocks to current state-of-the-art architectures outperforms the state-of-the-art methods on the HMDB51, UCF101 and Kinetics datasets. The other issue in training 3D CNNs is about training them from scratch with a huge labeled dataset to get a reasonable performance. So the knowledge learned in 2D CNNs is completely ignored. Another contribution in this work is a simple and effective technique to transfer knowledge from a pre-trained 2D CNN to a randomly initialized 3D CNN for a stable weight initialization. This allows us to significantly reduce the number of training samples for 3D CNNs. Thus, by fine-tuning this network, we beat the performance of generic and recent methods in 3D CNNs, which were trained on large video datasets, e.g. Sports-1M, and fine-tuned on the target datasets, e.g. HMDB51/UCF101.