In this paper, we propose delayed Long Short-Term Memory (dLSTM), an anomaly detection method for time-series data. We first build a predictive model from normal (non-anomalous) training data, then perform based on the prediction error observed However, there are multiple states in waveforms of which may lower accuracy. To deal with problem, utilize models LSTM detection. scheme, accuracy strongly depends selecting proper possible models. novel to determine Our approach provides predicted...

A small-gate-count 8 bit bidirectional phase-domain MAC (PMAC) circuit is proposed to minimize both area and energy consumption of extremely energy-efficient deep neural network (DNN) accelerators, targeting the Internet-of-Things (IoT) edge devices operating with very strict power budgets (e.g., harvesting). PMAC consumes significantly less than standard fully digital MACs, due its efficient analog accumulation nature based on gated-ring oscillator (GRO). The architectural analysis...

In this paper, we propose a method to cluster the spacial patterns of visual field in glaucoma patients analyze progression glaucoma. The degree can be divided into several regions by straight line boundaries, call specific structure Direct Product Structure paper. Since observe direct product fields, bottom-up hierarchical clustering embed structure. our method, according minimum description length (MDL) principle, select best division so that total code required for encoding data as well...

PhaseMAC (PMAC), a phase domain Gated-Ring-Oscillator (GRO) based 8bit MAC circuit, is proposed to minimize both area and power consumption of deep learning accelerators. PMAC composes only digital cells consumes significantly smaller than standard designs, owing its efficient analog accumulation nature. It occupies 26.6 times conventional which competitive circuits. achieves peak efficiency 14 TOPS/W, best reported 48% higher arts. Results in anomaly detection tasks are demonstrated, the...

We consider the prediction of glaucomatous visual field loss based on patient datasets. It is critically important to predict how rapidly disease progressing in an individual patient. However, number measurements for each so small that a reliable predictor cannot be constructed from data single alone. In this paper, we propose novel multi-task learning approach issue. Patient consist three features: ID, 74-dimensional values, and inspection time. reduce problem into one matrix completion...

There is a need for forecasting of short-range future values in data streams such as traffic flows, stock prices, and electricity consumption. However, concept drift non-stationary an important problem. We propose online prediction method called OPOSSAM streams. manages time-series segments short-term memory long-term memory, forecasts by local regression based on the similarity segments. In particular, keeps consistent reducing redundant samples with large errors, automatically adjusts...

It is crucial to appropriately maintain automatic ticket gates (ATGs) keep transportation operating smoothly in urban areas. Although the average failure rate of new ATGs extremely low, continuous operation for many years might lead unstable performance due deterioration, and need periodic maintenance avoid fatal faults halt operations extended periods. To detect anomalies at an early stage, “anomaly signs” can be utilized flag by service engineers before occur. In addition, minimize cost...

Shapelets are discriminative segments used to classify time-series instances. Shapelet methods that jointly learn both classifiers and shapelets have been studied in recent years because such provide interpretable results superior accuracy. The partial area under the receiver operating characteristic curve (pAUC) for a low range of false-positive rates (FPR) is an important performance measure practical cases industries as medicine, manufacturing, maintenance. In this article, we propose...

In a semiconductor fabrication plant, various types of sensors are installed at equipment processes to monitor the quality products and progress processes. These generate huge volume time-series waveform data, which used for failure prognostics, diagnosis, anomaly detection using supervised or unsupervised classifiers. For this purpose, automatic extraction features from data is needed; however, effective automatically compound having multiple spikes, complex states transitions difficult. If...

It is crucial for automatic ticket gates (ATGs) on railways, also known as fare collection systems, to detect anomalies at an early stage, especially in the separation module multiple tickets. required efficient and low-cost monitoring without any additional sensors old-type ATGs that need be maintained frequently. However, failure rate basically very low, data contain various kinds of normal status indicators depending complicated mechatronics controls. In addition, it hard collect high...

PhaseMAC (PMAC), a phase domain Gated-Ring-Oscillator (GRO) based 8bit MAC circuit, is proposed to minimize both area and power consumption of deep learning accelerators. PMAC composes only digital cells consumes significantly smaller than standard designs, owing its efficient analog accumulation nature. It occupies 26.6 times conventional which competitive circuits. achieves peak efficiency 14 TOPS/W, best reported 48% higher arts. Results in anomaly detection tasks are demonstrated, the...