摘要:
An optimization in machine learning is achieved using Newton's algorithm together with an efficient technique for solving linear equations, such as the method of conjugate residuals. The techniques of the present invention are applicable to learning language models, predicting classes of objects from images and videos, and classifying financial transactions for prevention of fraud. Other uses include determining a function from a sequence of words to a relevant web page for a search engine, or to inverting arbitrary output values of an analyzed system into an internally running simulation.
摘要:
An apparatus and a method are provided for learning a program with a large number of parameters. In one embodiment, a method not only distorts the input values, but also distorts some of the parameters in the program model. Such an approach not only forces the learned program to acquire parameter values to predict missing or desired data, but also to correct errors in the input data and the program parameters themselves, thereby rendering the learned program more resilient to overfitting and falling into local optima.
摘要:
One embodiment of the present invention provides a system characterizes a document with respect to clusters of conceptually related words. Upon receiving a document containing a set of words, the system selects “candidate clusters” of conceptually related words that are related to the set of words. These candidate clusters are selected using a model that explains how sets of words are generated from clusters of conceptually related words. Next, the system constructs a set of components to characterize the document, wherein the set of components includes components for candidate clusters. Each component in the set of components indicates a degree to which a corresponding candidate cluster is related to the set of words.
摘要:
A method and an apparatus allow learning a program that is characterized by a set of parameters. In addition to carrying out operations of the program based on an input vector and the values of the parameters, the method also carries out automatic differentiation steps over the operations of the program to compute derivatives of the output vector with respect to the parameters to any desired order. Based on the computed derivatives, the values of the parameters of the program are updated.
摘要:
A method and a system allow accessing several of a user's controlled access accounts by presenting the credentials of only one of the accounts. The method may include (a) storing the credentials for each of the user's accounts; (b) receiving from the user credentials corresponding to any of the user's accounts; (c) presenting the received credentials to access the corresponding account; and (d) upon successful access of the corresponding account, using the stored credentials to access one or more of the user's accounts without requiring the user to present the corresponding credentials. For each of the user's accounts, the credentials are stored encrypted, using a randomly generated key, common to all the encrypted credentials. In addition, the randomly generated key is encrypted using the credentials of each of the accounts. In that manner, plain-text copies of neither the random key nor the credentials of the accounts need to be stored.
摘要:
An instant messaging (IM) system has an IM server connected to a communication network, and logic operable on the IM server enabling a first user to browse contacts associated with a second user. In some cases the first user can add and list contacts accessed by browsing the contacts associated with the second user.
摘要:
One embodiment of the present invention provides a system that learns a generative model for textual documents. During operation, the system receives a current model, which contains terminal nodes representing random variables for words and cluster nodes representing clusters of conceptually related words. Within the current model, nodes are coupled together by weighted links, so that if a cluster node in the probabilistic model fires, a weighted link from the cluster node to another node causes the other node to fire with a probability proportionate to the link weight. The system also receives a set of training documents, wherein each training document contains a set of words. Next, the system applies the set of training documents to the current model to produce a new model.
摘要:
Methods and apparatus consistent with the invention provide improved organization of documents responsive to a search query. In one embodiment, a search query is received and a list of responsive documents is identified. The responsive documents are organized based in whole or in part on usage statistics.
摘要:
An instant messaging (IM) system has an IM server connected to a communication network, and logic operable on the IM server enabling a first user to browse contacts associated with a second user. In some cases the first user can add and list contacts accessed by browsing the contacts associated with the second user.
摘要:
Methods and apparatus consistent with the invention provide improved organization of documents responsive to a search query. In one embodiment, a search query is received and a list of responsive documents is identified. The responsive documents are organized based in whole or in part on usage statistics.