This chapter contains sections titled: Web Search Programs, Federated Digital Libraries, Research... more This chapter contains sections titled: Web Search Programs, Federated Digital Libraries, Research on Alternative Approaches to Distributed Searching, Beyond Searching
5th European Conference on Speech Communication and Technology (Eurospeech 1997)
Page 1. ENSEMBLE METHODS FOR CONNECTIONIST ACOUSTIC MODELLING GD Cook SR Waterhouse AJ Robinson C... more Page 1. ENSEMBLE METHODS FOR CONNECTIONIST ACOUSTIC MODELLING GD Cook SR Waterhouse AJ Robinson Cambridge University Engineering Department Trumpington Street, Cambridge, UK. ABSTRACT In this ...
We present Orchid: a decentralized market for anonymous communication and virtual private network... more We present Orchid: a decentralized market for anonymous communication and virtual private networking. Extant privacy solutions are either opaque commercial services with concomitant centralization risks or free peer to peer networks which lack the proper aligned incentives for service quality and economic security at scale. Orchid is a bandwidth market where node providers stake tokens to advertise their services using the Ethereum blockchain. Clients construct single or multi-hop onion routed circuits by selecting nodes randomly weighted on stake and filtered on secondary criteria (price, location, etc.). Staking aligns incentives against operator malfeasance and linear stake weighting in particular neutralizes sybil attacks. Orchid uses a probabilistic payment system which scales to millions of transactions per second, enabling a highly liquid bandwidth market without a trusted central party. Payments at packet scale allow high frequency trustless interactions by reducing the implicit floated balances between transactors to miniscule levels.
Summary
This thesis investigates a recent tool in statistical analysis: the mixtures-of-experts ... more Summary
This thesis investigates a recent tool in statistical analysis: the mixtures-of-experts model for classifica- tion and regression. The aim of the thesis is to place mixtures-of-experts models in context with other statistical models. The hope of doing this is that we may better understand their advantages and dis- advantages over other models. The thesis first considers mixtures-of-experts models from a theoretical perspective and compares them with other models such as trees, switching regression models and modu- lar networks. Two extensions of the mixtures-of-experts model are then proposed. The first extension is a constructive algorithm for learning model architecture and parameters which is insipired by recursive partitioning. The second extension uses Bayesian methods for learning the parameters of the model. These extensions are compared empirically with the standard mixtures-of-experts model and with other statistical models on small to medium sized data sets. In the second part of the thesis the mixtures-of- experts framework is applied to acoustic modelling within a large vocabulary speech recognition system. The mixtures-of-experts is shown to give an advantage over standard single neural network approaches on this task. The results of both of these sets of comparisons indicate that mixtures-of-experts models are competitive with other state-of-the-art statistical models.
Ą Text, by hitting a Moreover News database Ą Stock quotes, by sending the query to Yahoo! Financ... more Ą Text, by hitting a Moreover News database Ą Stock quotes, by sending the query to Yahoo! Finance Ą Pictures, by using OnlinePhoto Lab Ą Arithmetic, by sending the query to a calculator Nearly a year later, the Infrasearch team is now part of Sun's Project JXTA ...
This thesis is not substantially the same as any other that I have submitted for a degree or dipl... more This thesis is not substantially the same as any other that I have submitted for a degree or diploma or other qualification at any other university. This thesis is entirely the result of my own work over the last four years at Cambridge University Engineering Department. The length of this ...
... T. Kurnik a , * , Jonathan J. Oliver 1 , b , Steven R. Waterhouse b , Tim Dunn a , Yalia Jaya... more ... T. Kurnik a , * , Jonathan J. Oliver 1 , b , Steven R. Waterhouse b , Tim Dunn a , Yalia Jayalakshmi a , Matt Lesho a ... Appropriate alternatives include the many types of neural networks [11, 12], Multivariate Adaptive Regression Splines (MARS) [13], Adaptive Nearest Neighbor ...
This chapter contains sections titled: Web Search Programs, Federated Digital Libraries, Research... more This chapter contains sections titled: Web Search Programs, Federated Digital Libraries, Research on Alternative Approaches to Distributed Searching, Beyond Searching
5th European Conference on Speech Communication and Technology (Eurospeech 1997)
Page 1. ENSEMBLE METHODS FOR CONNECTIONIST ACOUSTIC MODELLING GD Cook SR Waterhouse AJ Robinson C... more Page 1. ENSEMBLE METHODS FOR CONNECTIONIST ACOUSTIC MODELLING GD Cook SR Waterhouse AJ Robinson Cambridge University Engineering Department Trumpington Street, Cambridge, UK. ABSTRACT In this ...
We present Orchid: a decentralized market for anonymous communication and virtual private network... more We present Orchid: a decentralized market for anonymous communication and virtual private networking. Extant privacy solutions are either opaque commercial services with concomitant centralization risks or free peer to peer networks which lack the proper aligned incentives for service quality and economic security at scale. Orchid is a bandwidth market where node providers stake tokens to advertise their services using the Ethereum blockchain. Clients construct single or multi-hop onion routed circuits by selecting nodes randomly weighted on stake and filtered on secondary criteria (price, location, etc.). Staking aligns incentives against operator malfeasance and linear stake weighting in particular neutralizes sybil attacks. Orchid uses a probabilistic payment system which scales to millions of transactions per second, enabling a highly liquid bandwidth market without a trusted central party. Payments at packet scale allow high frequency trustless interactions by reducing the implicit floated balances between transactors to miniscule levels.
Summary
This thesis investigates a recent tool in statistical analysis: the mixtures-of-experts ... more Summary
This thesis investigates a recent tool in statistical analysis: the mixtures-of-experts model for classifica- tion and regression. The aim of the thesis is to place mixtures-of-experts models in context with other statistical models. The hope of doing this is that we may better understand their advantages and dis- advantages over other models. The thesis first considers mixtures-of-experts models from a theoretical perspective and compares them with other models such as trees, switching regression models and modu- lar networks. Two extensions of the mixtures-of-experts model are then proposed. The first extension is a constructive algorithm for learning model architecture and parameters which is insipired by recursive partitioning. The second extension uses Bayesian methods for learning the parameters of the model. These extensions are compared empirically with the standard mixtures-of-experts model and with other statistical models on small to medium sized data sets. In the second part of the thesis the mixtures-of- experts framework is applied to acoustic modelling within a large vocabulary speech recognition system. The mixtures-of-experts is shown to give an advantage over standard single neural network approaches on this task. The results of both of these sets of comparisons indicate that mixtures-of-experts models are competitive with other state-of-the-art statistical models.
Ą Text, by hitting a Moreover News database Ą Stock quotes, by sending the query to Yahoo! Financ... more Ą Text, by hitting a Moreover News database Ą Stock quotes, by sending the query to Yahoo! Finance Ą Pictures, by using OnlinePhoto Lab Ą Arithmetic, by sending the query to a calculator Nearly a year later, the Infrasearch team is now part of Sun's Project JXTA ...
This thesis is not substantially the same as any other that I have submitted for a degree or dipl... more This thesis is not substantially the same as any other that I have submitted for a degree or diploma or other qualification at any other university. This thesis is entirely the result of my own work over the last four years at Cambridge University Engineering Department. The length of this ...
... T. Kurnik a , * , Jonathan J. Oliver 1 , b , Steven R. Waterhouse b , Tim Dunn a , Yalia Jaya... more ... T. Kurnik a , * , Jonathan J. Oliver 1 , b , Steven R. Waterhouse b , Tim Dunn a , Yalia Jayalakshmi a , Matt Lesho a ... Appropriate alternatives include the many types of neural networks [11, 12], Multivariate Adaptive Regression Splines (MARS) [13], Adaptive Nearest Neighbor ...
Uploads
Papers by Steven Waterhouse
This thesis investigates a recent tool in statistical analysis: the mixtures-of-experts model for classifica- tion and regression. The aim of the thesis is to place mixtures-of-experts models in context with other statistical models. The hope of doing this is that we may better understand their advantages and dis- advantages over other models. The thesis first considers mixtures-of-experts models from a theoretical perspective and compares them with other models such as trees, switching regression models and modu- lar networks. Two extensions of the mixtures-of-experts model are then proposed. The first extension is a constructive algorithm for learning model architecture and parameters which is insipired by recursive partitioning. The second extension uses Bayesian methods for learning the parameters of the model. These extensions are compared empirically with the standard mixtures-of-experts model and with other statistical models on small to medium sized data sets. In the second part of the thesis the mixtures-of- experts framework is applied to acoustic modelling within a large vocabulary speech recognition system. The mixtures-of-experts is shown to give an advantage over standard single neural network approaches on this task. The results of both of these sets of comparisons indicate that mixtures-of-experts models are competitive with other state-of-the-art statistical models.
This thesis investigates a recent tool in statistical analysis: the mixtures-of-experts model for classifica- tion and regression. The aim of the thesis is to place mixtures-of-experts models in context with other statistical models. The hope of doing this is that we may better understand their advantages and dis- advantages over other models. The thesis first considers mixtures-of-experts models from a theoretical perspective and compares them with other models such as trees, switching regression models and modu- lar networks. Two extensions of the mixtures-of-experts model are then proposed. The first extension is a constructive algorithm for learning model architecture and parameters which is insipired by recursive partitioning. The second extension uses Bayesian methods for learning the parameters of the model. These extensions are compared empirically with the standard mixtures-of-experts model and with other statistical models on small to medium sized data sets. In the second part of the thesis the mixtures-of- experts framework is applied to acoustic modelling within a large vocabulary speech recognition system. The mixtures-of-experts is shown to give an advantage over standard single neural network approaches on this task. The results of both of these sets of comparisons indicate that mixtures-of-experts models are competitive with other state-of-the-art statistical models.