We develop a framework to incorporate unlabeled data in the Error-Correcting Output Coding (ECOC) setup by decomposing multiclass problems into multiple binary problems and then use Co-Training to learn the individual binary classification problems. We show that our method is especially useful for classification tasks involving a large number of categories where Co-training doesn’t perform very...

Changed hearing occurs when sensorineural loss is acquired or increases, when hearing aids or cochlear implants are first acquired, when hearing aids are reprogrammed, and when cochlear implants are remapped. The changes affect speech perception-a process in which decisions about a talker's language output are made on the basis of sensory and contextual evidence, using knowledge and skill. The ...

This paper studies how joint training of multiple support vector machines (SVMs) can improve the effectiveness and efficiency of automatic image annotation. We cast image annotation as an output-related multi-task learning framework, with the prediction of each tag’s presence as one individual task. Evidently, these tasks are related via correlations between tags. The proposed joint learning fr...

We demonstrate the utility of multi-task learning with recurrent neural networks in the context of a temporal regression task using real world gas turbine data. Our goal is to learn the input-output relationship of a task with temporal dependencies given only few data by exploiting additional data of related tasks. Therefore, we propose a multi-task learning approach with many inter-task and fe...

This study investigated the effects of different output-based task repetition conditions on EFL learners’ speech act production. Three intact classes of English-major students constituted three instructional groups: (1) the explicit task-repetition (ETR) group, (2) the implicit task-repetition (ITR) group, and (3) the no-input task repetition (NTR) group. All the three groups engaged in t...

Two sensorimotor tasks that share neither sensory nor motor modality can interfere with one another when they are performed simultaneously. A possible cause for this interference is the recruitment of common brain regions by these two tasks, thereby creating a bottleneck of information processing. This hypothesis predicts that such "bottleneck" regions would be activated by each task even when ...

This paper describes a generic lexicon tool which uses lexical representations and finite state transducers enhanced by arithmetic operations in DATR to generate individual output formats from a general phonological feature based representation. The tool was developed in connection with the lexicon component of a diagnostic evaluation toolkit, BEETLE, for a linguistic word recognition system. T...

This paper addresses verification and debugging tool for development of FPGA modules. Proposed tool is developed for educational purposes in teaching students on Digital Design and VHDL programming language. Main goal of the debugging module is to get/set signal values while the FPGA board is running the module of interest. Two PicoBlaze CPUs are used in order to synchronize the input and outpu...

A common way to improve the performance of deep learning is to train an ensemble of neural networks and combine them during decoding. However, this is computationally expensive in test time. In this paper, we propose an diversity-penalizing ensemble training (DPET) procedure, which trains an ensemble of DNNs, whose parameters were differently initialized, and penalizes differences between each ...