Convolutional Sequence to Sequence Learning
ثبت نشده
چکیده
A. Weight Initialization We derive a weight initialization scheme tailored to the GLU activation function similar to Glorot & Bengio (2010); He et al. (2015b) by focusing on the variance of activations within the network for both forward and backward passes. We also detail how we modify the weight initialization for dropout. A.1. Forward Pass Assuming that the inputs x l of a convolutional layer l and its weights W l are independent and identically distributed (i.i.d.), the variance of its output, computed as y l =W l x l +b l , is
منابع مشابه
Seismic Data Forecasting: A Sequence Prediction or a Sequence Recognition Task
In this paper, we have tried to predict earthquake events in a cluster of seismic data on pacific ring of fire, using multivariate adaptive regression splines (MARS). The model is employed as either a predictor for a sequence prediction task, or a binary classifier for a sequence recognition problem, which could alternatively help to predict an event. Here, we explain that sequence prediction/r...
متن کاملBLIND PARAMETER ESTIMATION OF A RATE k/n CONVOLUTIONAL CODE IN NOISELESS CASE
This paper concerns to blind identification of a convolutional code with desired rate in a noiseless transmission scenario. To the best of our knowledge, blind estimation of convolutional code based on only the received bitstream doesn’t lead to a unique solution. Hence, without loss of generality, we will assume that the transmitter employs a non-catastrophic encoder. Moreover, we consider a c...
متن کاملOperation Sequencing Optimization in CAPP Using Hybrid Teaching-Learning Based Optimization (HTLBO)
Computer-aided process planning (CAPP) is an essential component in linking computer-aided design (CAD) and computer-aided manufacturing (CAM). Operation sequencing in CAPP is an essential activity. Each sequence of production operations which is produced in a process plan cannot be the best possible sequence every time in a changing production environment. As the complexity of the product incr...
متن کاملLot Streaming in No-wait Multi Product Flowshop Considering Sequence Dependent Setup Times and Position Based Learning Factors
This paper considers a no-wait multi product flowshop scheduling problem with sequence dependent setup times. Lot streaming divide the lots of products into portions called sublots in order to reduce the lead times and work-in-process, and increase the machine utilization rates. The objective is to minimize the makespan. To clarify the system, mathematical model of the problem is presented. Sin...
متن کاملزمانبندی گروهی با در نظر گرفتن اثر یادگیری در سیستم تولید سلولی
The group scheduling problem in the cellular manufacturing system is comprised of two levels of scheduling. At the first level, the sequence of parts in each part-family is determined, and then at the second level the sequence of part-families is determined. In this paper, the flow shop group scheduling is investigated in order to minimize the makespan. In traditional group scheduling problems,...
متن کاملConvolutional Sequence Modeling Revisited
Although both convolutional and recurrent architectures have a long history in sequence prediction, the current “default” mindset in much of the deep learning community is that generic sequence modeling is best handled using recurrent networks. Yet recent results indicate that convolutional architectures can outperform recurrent networks on tasks such as audio synthesis and machine translation....
متن کامل