Abstract: Techniques for using enhanced logit values for classifying utterances and messages input to chatbot systems in natural language processing. A method can include a chatbot system receiving an utterance generated by a user interacting with the chatbot system and inputting the utterance into a machine-learning model including a series of network layers. A final network layer of the series of network layers can include a logit function. The machine-learning model can map a first probability for a resolvable class to a first logit value using the logit function. The machine-learning model can map a second probability for a unresolvable class to an enhanced logit value. The method can also include the chatbot system classifying the utterance as the resolvable class or the unresolvable class based on the first logit value and the enhanced logit value.

Abstract: Techniques for using logit values for classifying utterances and messages input to chatbot systems in natural language processing. A method can include a chatbot system receiving an utterance generated by a user interacting with the chatbot system. The chatbot system can input the utterance into a machine-learning model including a set of binary classifiers. Each binary classifier of the set of binary classifiers can be associated with a modified logit function. The method can also include the machine-learning model using the modified logit function to generate a set of distance-based logit values for the utterance. The method can also include the machine-learning model applying an enhanced activation function to the set of distance-based logit values to generate a predicted output. The method can also include the chatbot system classifying, based on the predicted output, the utterance as being associated with the particular class.

Abstract: The present invention is directed to tricyclic compounds, pharmaceutically acceptable compositions comprising compounds of the invention and methods of using said compositions in the treatment of various disorders.

Abstract: Techniques are disclosed herein for objective function optimization in target based hyperparameter tuning. In one aspect, a computer-implemented method is provided that includes initializing a machine learning algorithm with a set of hyperparameter values and obtaining a hyperparameter objective function that comprises a domain score for each domain that is calculated based on a number of instances within an evaluation dataset that are correctly or incorrectly predicted by the machine learning algorithm during a given trial. For each trial of a hyperparameter tuning process: training the machine learning algorithm to generate a machine learning model, running the machine learning model in different domains using the set of hyperparameter values, evaluating the machine learning model for each domain, and once the machine learning model has reached convergence, outputting at least one machine learning model.

Abstract: Techniques are disclosed herein for continuous hyperparameter tuning with automatic domain weight adjustment based on periodic performance checkpoints. In one aspect, a method is provided that includes initializing a machine learning algorithm with a set of hyperparameter values and obtaining a hyperparameter objective function that is defined at least in part on a plurality of domains of a search space that is associated with the machine learning algorithm. For each trial of a hyperparameter tuning process: running the machine learning algorithm in different domains using the set of hyperparameter values, periodically checking a performance of the machine learning algorithm in the different domains based on the hyperparameter objective function; and continuing hyperparameter tuning with a new set of hyperparameter values after automatically adjusting the domain weights according to a regression status of the different domains.

Abstract: Provided are a method and apparatus for evaluating image relative definition, a device and a medium, relating to technologies such as computer vision, deep learning and intelligent medical. A specific implementation solution is: extracting a multi-scale feature of each image in an image set, where the multi-scale feature is used for representing definition features of objects having different sizes in an image; and scoring relative definition of each image in the image set according to the multi-scale feature by using a relative definition scoring model pre-trained, where the purpose for training the relative definition scoring model is to learn a feature related to image definition in the multi-scale feature.

Abstract: A method to achieve full densification and grain size control for sintering metal materials, wherein raw material powder is deagglomerated to obtain deagglomerated powder with dispersion. The deagglomerated powder is granulated by spray granulation. The granulated particles are processed by high-pressure die pressing and cold isostatic pressing. The powder compact is sintered by two-step pressureless sintering. The first step is to heat up the powder compact to a higher temperature and hold for a short time to obtain 75-85% theoretical density; the second step is to cool down powder compact to a lower temperature and hold for a long time. The two-step sintering can decrease the sintering temperature, so that the powder compact can be densified at a lower temperature. Thus, the obtained refractory metal product is densified, with ultrafine grains, uniform grain size distribution, and outstanding mechanical properties.

Abstract: The present disclosure provides a touch device and a manufacturing method thereof. The touch device includes a substrate, a touch electrode layer, a protective layer, and a plurality of wires. The substrate includes a first region and a second region, in which the second region is adjacent to the first region. The touch electrode layer is disposed in the first region and is completely covered by the protective layer. The protective layer has a plurality of openings. The openings expose a portion of the touch electrode layer and extend from the first region to the second region. Each wire is formed in the corresponding openings and extends from the portion of the touch electrode layer to the second region, in which each opening is partially filled with one of the wires, and thereby a recess is defined in each opening.

Abstract: The disclosure discloses an algebraic decoding method and a decoder for a (n, n(n?1), n?1) permutation group code in a communication modulation system. The basic principle of the decoding method is: assuming that two code elements p(r1)=s1 and p(r2)=s2 can be correctly detected in a received real vector with a length of n, including their element values s1, s2 and position indices r1, r2 in the vector, an intermediate parameter w is determined by solving an equation (r1?r2)w=(s1?s2)(mod n); and each code element is calculated by w according to p(i)=(s1+(n?r1+i)w)(mod n), i=1, 2, . . . , n. The decoder is mainly composed of multiple n-dimensional registers, a w calculator, n code element calculators, and a code element buffer. In the disclosure, in a case where a receiver only correctly detects two code elements in a transmitted codeword with a length of n, the codeword can be correctly decoded by using the received information of the two code elements.

Abstract: The present disclosure provides an encoding method and an encoder for a (n, n(n?1), n?1) permutation group code in a communication modulation system, in which 2k k-length binary information sequences are mapped to 2k n-length permutation codeword signal points in a n-dimensional modulation constellation ?n. The constellation ?n with the coset characteristics is formed by selecting 2k n-length permutation codewords from n(n?1) permutation codewords of a code set Pn,xi of the (n, n(n?1), n?1) permutation group code based on coset partition. The constellation ?n is a coset code in which 2k1 cosets are included and each coset includes 2k2 permutation codewords, where k=k1+k2, and 2k?n(n?1). The present disclosure utilizes the coset characteristics to realize one-to-one correspondence mapping of the binary information sequence set to the permutation code constellation, so that the time complexity of executing the encoder is at most the linear complexity of the code length n.

Abstract: A security method for monitoring an optical module and a three-dimensional sensor using the same apply electromagnetic induction to the three-dimensional sensor to monitor the optical module and a light source module. Two inductive coils corresponding to each other are arranged on the light source module and the optical module. An alternative current is inputted to one of the inductive coils and another of the inductive coils generates an inductive current. The value of the inductive current is continuously detected. When the value of the inductive current varies, the abnormality of the optical module is determined to shut down the light source module, thereby completing the security mechanism of the three-dimensional sensor.

Abstract: The present disclosure provides a touch device and a manufacturing method thereof. The touch device includes a substrate, a touch electrode layer, a protective layer, and a plurality of wires. The substrate includes a first region and a second region, in which the second region is adjacent to the first region. The touch electrode layer is disposed in the first region and is completely covered by the protective layer. The protective layer has a plurality of openings. The openings expose a portion of the touch electrode layer and extend from the first region to the second region. Each wire is formed in the corresponding openings and extends from the portion of the touch electrode layer to the second region, in which each opening is partially filled with one of the wires, and thereby a recess is defined in each opening.

Abstract: The present disclosure provides an encoding method and an encoder for a (n, n(n?1), n?1) permutation group code in a communication modulation system, in which 2k k-length binary information sequences are mapped to 2k n-length permutation codeword signal points in a n-dimensional modulation constellation ?n. The constellation ?n with the coset characteristics is formed by selecting 2k n-length permutation codewords from n(n?1) permutation codewords of a code set Pn,xi of the (n, n(n?1), n?1) permutation group code based on coset partition. The constellation ?n is a coset code in which 2k1 cosets are included and each coset includes 2k2 permutation codewords, where k=k1+k2, and 2k?n(n?1). The present disclosure utilizes the coset characteristics to realize one-to-one correspondence mapping of the binary information sequence set to the permutation code constellation, so that the time complexity of executing the encoder is at most the linear complexity of the code length n.

Abstract: The disclosure discloses an algebraic decoding method and a decoder for a (n, n(n?1), n?1) permutation group code in a communication modulation system. The basic principle of the decoding method is: assuming that two code elements p(r1)=s1 and p(r2)=s2 can be correctly detected in a received real vector with a length of n, including their element values s1, s2 and position indices r1, r2 in the vector, an intermediate parameter w is determined by solving an equation (r1?r2)w=(s1?s2)(mod n); and each code element is calculated by w according to p(i)=(s1+(n?r1+i)w)(mod n), i=1, 2, . . . , n. The decoder is mainly composed of multiple n-dimensional registers, a w calculator, n code element calculators, and a code element buffer. In the disclosure, in a case where a receiver only correctly detects two code elements in a transmitted codeword with a length of n, the codeword can be correctly decoded by using the received information of the two code elements.

Abstract: A security method for monitoring an optical module and a three-dimensional sensor using the same apply electromagnetic induction to the three-dimensional sensor to monitor the optical module and a light source module. Two inductive coils corresponding to each other are arranged on the light source module and the optical module. An alternative current is inputted to one of the inductive coils and another of the inductive coils generates an inductive current. The value of the inductive current is continuously detected. When the value of the inductive current varies, the abnormality of the optical module is determined to shut down the light source module, thereby completing the security mechanism of the three-dimensional sensor.

Abstract: An analog circuit fault mode classification method comprises the following implementation steps: (1) collecting M groups of voltage signal sample vectors Vij to each of fault modes Fi of the analog circuit by using a data collection board; (2) sequentially extracting fault characteristic vectors VijF of the voltage signal sample vectors Vij by using subspace projection; (3) standardizing the extracted fault characteristic vectors VijF to obtain standardized fault characteristic vectors; (4) constructing a fault mode classifier based on a support vector machine, inputting the standardized fault characteristic vectors, performing learning and training on the classifier, and determining structure parameters of the classifier; and (5) completing determination of fault modes according to fault mode determination rules. The fault mode classifier of the present invention is simple in learning and training and reliable in mode classification accuracy.

Abstract: The present invention provides a method to achieve full densification and grain size control for sintering metal materials. First, raw material powder is deagglomerated to obtain deagglomerated powder with dispersion. The deagglomerated powder is granulated by spray granulation. The granulated particles are processed by high-pressure die pressing and cold isostatic pressing. The powder compact is sintered by two-step pressureless sintering. The first step is to heat up the powder compact to a higher temperature and hold for a short time to obtain 75-85% theoretical density; the second step is to cool down powder compact to a lower temperature and hold for a long time. The two-step sintering can decrease the sintering temperature, so that the powder compact can be densified at a lower temperature. Thus, the obtained refractory metal product is densified, with ultrafine grains, uniform grain size distribution, and outstanding mechanical properties.

Abstract: An analog circuit fault diagnosis method using a single testable node comprises the following steps: (1) obtaining prior sample data vectors under each fault mode; (2) computing a statistical average of the prior sample data vectors under each of the fault modes; (3) decomposing a signal by an orthogonal Haar wavelet filter set; (4) extracting the feature factor of the prior sample fault modes; (5) extracting a fault-mode-to-be-tested feature factor; (6) computing a correlation coefficient matrix and correlation metric parameters between the feature factor of the prior sample fault modes and the feature factor of the fault-mode-to-be-tested; and (7) determining a fault mode according to a maximal correlation principle by comparing the correlation metric parameters.

Abstract: An analog circuit fault mode classification method comprises the following implementation steps: (1) collecting M groups of voltage signal sample vectors Vij to each of fault modes Fi of the analog circuit by using a data collection board; (2) sequentially extracting fault characteristic vectors VijF of the voltage signal sample vectors Vij by using subspace projection; (3) standardizing the extracted fault characteristic vectors VijF to obtain standardized fault characteristic vectors; (4) constructing a fault mode classifier based on a support vector machine, inputting the standardized fault characteristic vectors, performing learning and training on the classifier, and determining structure parameters of the classifier; and (5) completing determination of fault modes according to fault mode determination rules. The fault mode classifier of the present invention is simple in learning and training and reliable in mode classification accuracy.

Abstract: An analog circuit fault diagnosis method using a single testable node comprises the following steps: (1) obtaining prior sample data vectors under each fault mode; (2) computing a statistical average of the prior sample data vectors under each of the fault modes; (3) decomposing a signal by an orthogonal Haar wavelet filter set; (4) extracting the feature factor of the prior sample fault modes; (5) extracting a fault-mode-to-be-tested feature factor; (6) computing a correlation coefficient matrix and correlation metric parameters between the feature factor of the prior sample fault modes and the feature factor of the fault-mode-to-be-tested; and (7) detennining a fault mode according to a maximal correlation principle by comparing the correlation metric parameters.