Demystifying Deep Neural Networks Through Interpretation: A Survey

Modern deep learning algorithms tend to optimize an objective metric, such as minimize a cross entropy loss on a training dataset, to be able to learn. The problem is that the single metric is an incomplete description of the real world tasks. The single metric cannot explain why the algorithm learn. When an erroneous happens, the lack of interpretability causes a hardness of understanding and fixing the error. Recently, there are works done to tackle the problem of interpretability to provide insights into neural networks behavior and thought process. The works are important to identify potential bias and to ensure algorithm fairness as well as expected performance.

Demysifying Deep Neural Networks Through Interpretation: A Survey

Modern deep learning algorithms tend to optimize an objective metric, such as minimize a cross entropy loss on a training dataset, to be able to learn. The problem is that the single metric is an incomplete description of the real world tasks. The single metric cannot explain why the algorithm learn. When an erroneous happens, the lack of interpretability causes a hardness of understanding and fixing the error. Recently, there are works done to tackle the problem of interpretability to provide insights into neural networks behavior and thought process. The works are important to identify potential bias and to ensure algorithm fairness as well as expected performance.

Few-Shot Keyword Spotting With Prototypical Networks

Recognizing a particular command or a keyword, keyword spotting has been widely used in many voice interfaces such as Amazon's Alexa and Google Home. In order to recognize a set of keywords, most of the recent deep learning based approaches use a neural network trained with a large number of samples to identify certain pre-defined keywords. This restricts the system from recognizing new, user-defined keywords. Therefore, we first formulate this problem as a few-shot keyword spotting and approach it using metric learning. To enable this research, we also synthesize and publish a Few-shot Google Speech Commands dataset. We then propose a solution to the few-shot keyword spotting problem using temporal and dilated convolutions on prototypical networks. Our comparative experimental results demonstrate keyword spotting of new keywords using just a small number of samples.

Deep learning method to remove chemical, kinetic and electric artifacts on ISEs

We suggest a deep learning based sensor signal processing method to remove chemical, kinetic and electrical artifacts from ion selective electrodes' measured values. An ISE is used to investigate the concentration of a specific ion from aqueous solution, by measuring the Nernst potential along the glass membrane. However, application of ISE on a mixture of multiple ion has some problem. First problem is a chemical artifact which is called ion interference effect. Electrically charged particles interact with each other and flows through the glass membrane of different ISEs. Second problem is the kinetic artifact caused by the movement of the liquid. Water molecules collide with the glass membrane causing abnormal peak values of voltage. The last artifact is the interference of ISEs. When multiple ISEs are dipped into same solution, one electrode's signal emission interference voltage measurement of other electrodes. Therefore, an ISE is recommended to be applied on single-ion solution, without any other sensors applied at the same time. Deep learning approach can remove both 3 artifacts at the same time. The proposed method used 5 layers of artificial neural networks to regress correct signal to remove complex artifacts with one-shot calculation. Its MAPE was less than 1.8% and R2 of regression was 0.997. A randomly chosen value of AI-processed data has MAPE less than 5% (p-value 0.016).

Pan-artifact Removing with Deep Learning, on ISEs

This paper presents a signal-processing method to remove pan-artifact on ISEs with artificial neural networks. An Ion Selective Electrode is used to investigate the concentration of a specific ion from aqueous solution, by measuring the Nernst potential along the glass membrane. However, Application of ISE on a multi-ion solution has problem. First problem is a chemicophysical artifact which is called ion interference effect. Electrically charged particles interact with each other and flows through the glass membrane of different ISEs. Second problem is that movement of liquid directly interfere the glass membrane, causing inaccurate voltage measurement. When multiple ISEs are dipped into same solution, a sensor's signal emission interference voltage measurement of other sensors. Therefore, an ISE is recommended to applied on single-ion solution, without any other sensors applied at the same time. Deep learning approach can remove both artifacts at the same time. The proposed method is designed to remove complex artifacts with one-shot calculation, with MAPE less than 1.8%, and R2 as 0.997. A randomly chosen value of AI-predicted value has MAPE less than 5% (p-value 0.016).

Drug-disease Graph: Predicting Adverse Drug Reaction Signals via Graph Neural Network with Clinical Data

Adverse Drug Reaction (ADR) is a significant public health concern world-wide. Numerous graph-based methods have been applied to biomedical graphs for predicting ADRs in pre-marketing phases. ADR detection in post-market surveillance is no less important than pre-marketing assessment, and ADR detection with large-scale clinical data have attracted much attention in recent years. However, there are not many studies considering graph structures from clinical data for detecting an ADR signal, which is a pair of a prescription and a diagnosis that might be a potential ADR. In this study, we develop a novel graph-based framework for ADR signal detection using healthcare claims data. We construct a Drug-disease graph with nodes representing the medical codes. The edges are given as the relationships between two codes, computed using the data. We apply Graph Neural Network to predict ADR signals, using labels from the Side Effect Resource database. The model shows improved AUROC and AUPRC performance of 0.795 and 0.775, compared to other algorithms, showing that it successfully learns node representations expressive of those relationships. Furthermore, our model predicts ADR pairs that do not exist in the established ADR database, showing its capability to supplement the ADR database.

Machine learning approach to remove ion interference effect in agricultural nutrient solutions

High concentration agricultural facilities such as vertical farms or plant factories consider hydroponic techniques as optimal solutions. Although closed-system dramatically reduces water consumption and pollution issues, it has ion-ratio related problem. As the root absorbs individual ions with different rate, ion rate in a nutrient solution should be adjusted periodically. But traditional method only considers pH and electrical conductivity to adjust the nutrient solution, leading to ion imbalance and accumulation of excessive salts. To avoid those problems, some researchers have proposed ion-balancing methods which measure and control each ion concentration. However, those approaches do not overcome the innate limitations of ISEs, especially ion interference effect. An anion sensor is affected by other anions, and the error grows larger in higher concentration solution. A machine learning approach to modify ISE data distorted by ion interference effect is proposed in this paper. As measurement of TDS value is relatively robust than any other signals, we applied TDS as key parameter to build a readjustment function to remove the artifact. Once a readjustment model is established, application on ISE data can be done in real time. Readjusted data with proposed model showed about 91.6 ~ 98.3% accuracies. This method will enable the fields to apply recent methods in feasible status.