Credit Assignment through Time : Alternatives

نویسنده

  • Yoshua Bengio
چکیده

Learning to recognize or predict sequences using long-term context has many applications. However, practical and theoretical problems are found in training recurrent neural networks to perform tasks in which input/output dependencies span long intervals. Starting from a mathematical analysis of the problem, we consider and compare alternative algorithms and architectures on tasks for which the span of the input/output dependencies can be controlled. Results on the new algorithms show performance qualitatively superior to that obtained with backpropagation.

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Credit Assignment through Time: Alternatives to Backpropagation

Learning to recognize or predict sequences using long-term context has many applications. However, practical and theoretical problems are found in training recurrent neural networks to perform tasks in which input/output dependencies span long intervals. Starting from a mathematical analysis of the problem, we consider and compare alternative algorithms and architectures on tasks for which the ...

متن کامل

Discounting of Letters of Credit; a Legal Analysis

Letter of Credit is an international payment instrument whereby the issuing bank undertakes to pay the beneficiary, against presentation of certain stipulated documents, according to the conditions of the Letter of Credit. Discounting of LC for the short-term financing of the seller, due to the independent and irrevocable undertaking of the bank to make payment, is prevalent. Beneficiary gets t...

متن کامل

Credit Assignment through Time : Alternatives to

Learning to recognize or predict sequences using long-term context has many applications. However, practical and theoretical problems are found in training recurrent neural networks to perform tasks in which input/output dependencies span long intervals. Starting from a mathematical analysis of the problem, we consider and compare alternative algorithms and architectures on tasks for which the ...

متن کامل

Alternatives for Classifier System Credit Assignment

Classifier systems are production rule systems that automatically generate populations of rules cooperating to accomplish desired tasks. The genetic algorithm is the systems' discovery mechanism, and its effectiveness is dependent in part on the accurate estimation of the relative merit of each of the rules (classifiers) in the current population. Merit is estimated conventionally by use of the...

متن کامل

QUICR-Learning for Multi-Agent Coordination

Coordinating multiple agents that need to perform a sequence of actions to maximize a system level reward requires solving two distinct credit assignment problems. First, credit must be assigned for an action taken at time step t that results in a reward at time step t′ > t. Second, credit must be assigned for the contribution of agent i to the overall system performance. The first credit assig...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 2005