High Stability Anomaly Detection in Random Environments

We propose anomaly detection to refine input data for predictive machine learning systems. When training, if there are outliers such as spike noises mixed in the input data, the quality of the trained model is deteriorated. The removing such outliers would be expected the service quality of machine learning systems improves such as autonomous vehicles and ship navigation. Conventionally, anomaly detection methods generally require the support of domain experts, and they could not treat with unstable random environments well. We propose a new anomaly detection method, which is highly stable and is capable of treating with random environments without experts. The proposed methods focus on a pairwise correlation between two input time-series, change rates of them are calculated and summarized on a quadrant chart for further analysis. The experiment using an open time-series dataset shows that the proposed methods successfully detect anomalies, and the detected data points are easily illustrated in a human-interpretable way.

The Uncertainty of Counterfactuals in Deep Learning

Counterfactuals have become a useful tool for explainable Artificial Intelligence (XAI). Counterfactuals provide various perturbations to a data instance to yield an alternate classification from a machine learning model. Several algorithms have been designed to generate counterfactuals using deep neural networks; however, despite their growing use in many mission-critical fields, there has been no investigation to date as to the epistemic uncertainty of generated counterfactuals. This could result in the use of risk-prone explanations in these fields. In this work, we use several data sets to compare the epistemic uncertainty of original instances to that of counterfactuals generated from those instances. As part of our analysis, we also measure the extent to which counterfactuals can be considered anomalies in those data sets. We find that counterfactual uncertainty is higher in three of the four datasets tested. Moreover, our experiments suggest a possible connection between reconstruction error using a deep autoencoder and the difference in epistemic uncertainty between training data and counterfactuals generated from that training data for a deep neural network.

Population-Based Novelty Searches Can Converge

Novelty search is a powerful tool for finding sets of complex objects in complicated, open-ended spaces. Recent empirical analysis on a simplified version of novelty search makes it clear that novelty search happens at the level of the archive space, not the individual point space. The sparseness measure and archive update criterion create a process that is driven by a clear pair of objectives: spread out to cover the space, while trying to remain as efficiently packed as possible driving these simplified variants to converge to an

Analysis of comparative performance of deep learning models for sentiment analysis

Sentiment analysis of text can be performed using machine learning and natural language processing methods. However, there is no single tool or method that is effective in all cases. The objective of this research project is to determine the effectiveness of neural network-based architecture to perform sentiment analysis of customer comments and reviews, such as the ones on Amazon site. A typical sentiment analysis process involves text preparation (of acquired content), sentiment detection, sentiment classification and analysis of results. In this research, the objective is to a) identify the best approach for text preparation in a given application (text filtering approach to remove errors in data), and, most importantly, b) what is the best machine learning (feed forward neural nets, convolutional neural nets, Long Short-Term Memory networks) approach that provides best classification accuracy. In this research, a set of three thousand two hundred reviews of food related products were used to train and experiment with a neural network-based sentiment analysis system. The neural network implementation of six different models provided close to one-hundred percent accuracy of test data, and a decent test accuracy in mid-80%. The results of the research would be useful to businesses in evaluating customer preferences for products or services.

TSAL: Two steps Adversarial learning based domain adaptation

Since labeled data availability differs greatly across domains, Domain Adaptation focuses on learning in new and unfamiliar domains by reducing distribution divergence. Recent research suggests that the adversarial learning approach could be a promising way to achieve the domain adaptation objective. Adversarial learning is a strategy for learning domain-transferable features in robust deep networks. This paper introduces the TSAL paradigm, a two-step adversarial learning framework. It addresses the real-world problem of text classification, where source domain(s) has labeled data but target domain (s) has only unlabeled data. TSAL utilizes joint adversarial learning with class information and domain alignment deep network architecture to learn both domain-invariant and domain-specific features extractors. It consists of two training steps that are similar to the paradigm, in which pre-trained model weights are used as initialization for training with new data. TSAL’s two training phases, however, are based on the same data, not different data, as is the case with fine-tuning. Furthermore, TSAL only uses the learned domain-invariant feature extractor from the first training as an initialization for its peer in subsequent training. By doubling the training, TSAL can emphasize the leverage of the small unlabeled target domain and learn effectively what to share between various domains. A detailed analysis of many benchmark datasets reveals that our model consistently outperforms the prior art across a wide range of dataset distributions.

An Order-Independent Algorithm for Learning Chain Graphs

LWF chain graphs combine directed acyclic graphs and undirected graphs. We propose a PC-like algorithm, called PC4LWF, that finds the structure of chain graphs under the faithfulness assumption to resolve the problem of scalability of the proposed algorithm by Studeny (1997). We prove that PC4LWF is order dependent, in the sense that the output can depend on the order in which the variables are given. This order dependence can be very pronounced in high-dimensional settings. We propose two modifications of the PC4LWF algorithm that remove part or all of this order dependence. Simulation results with different sample sizes, network sizes, and p-values demonstrate the competitive performance of the PC4LWF algorithms in comparison with the LCD algorithm proposed by Ma et al. (2008) in low-dimensional settings and improved performance (with regard to error measures) in high-dimensional settings.

Research Challenges in Designing Differentially Private Text Generation Mechanisms

Accurately learning from user data while ensuring quantifiable privacy guarantees provides an opportunity to build better ML models while maintaining user trust. Recent literature has demonstrated the applicability of a generalized form of Differential Privacy to provide guarantees over text queries. Such mechanisms add privacy preserving noise to vectorial representations of text in high dimension and return a text based projection of the noisy vectors. However, these mechanisms are sub-optimal in their trade-off between privacy and utility. In this proposal paper, we describe some challenges in balancing this trade-off. At a high level, we provide two proposals: (1) a framework called LAC which defers some of the noise to a privacy amplification step and (2), an additional suite of three different techniques for calibrating the noise based on the local region around a word. Our objective in this paper is not to evaluate a single solution but to further the conversation on these challenges and chart pathways for building better mechanisms.