Abstract: Methods and systems for beam search decoding. One of the methods includes initializing beam data specifying a set of k candidate output sequences and a respective total score for each of the candidate output sequences; updating the beam data at each of a plurality of decoding steps, comprising, at each decoding step: generating a score distribution that comprises a respective score for each token in the vocabulary; identifying a plurality of expanded sequences; generating, for each expanded sequence, a respective backwards-looking score; generating, for each expanded sequence, a respective forward-looking score; computing, for each expanded sequence, a respective total score from the respective forward-looking score for the expanded sequence and the respective backwards-looking score for the expanded sequence; and updating the set of k candidate output sequences using the respective total scores for the expanded sequences.

Abstract: A liquid dispenser comprising a liquid pump. The liquid pump comprises a first liquid inlet configured to allow the introduction of a first liquid into a mixing chamber; a second liquid inlet configured to allow the introduction of a second liquid into the mixing chamber; an outlet valve configured to regulate the release of a mixed liquid from the mixing chamber, the mixed liquid being a blend of the first liquid and the second liquid; and a reciprocating member configured to effect a reciprocating movement along a longitudinal axis, the reciprocating member being configured to regulate the aperture of the outlet valve. The liquid dispenser comprises a piezoelectric detector arranged in such a way that when the liquid pump generates a shockwave, the shockwave is detected by the piezoelectric detector, the piezoelectric detector generating a voltage peak which is a function of the shockwave.

Abstract: There is described a neural network system for generating a graph, the graph comprising a set of nodes and edges. The system comprises one or more neural networks configured to represent a probability distribution over sequences of node generating decisions and/or edge generating decisions, and one or more computers configured to sample the probability distribution represented by the one or more neural networks to generate a graph.

Abstract: Systems, methods, and computer program products for monitoring a level of product in a reservoir of a chemical dispensing system. A lid configured to engage an opening of a container includes a level sensor having a transceiver that transmits an output signal and receives a portion of the output signal reflected back to the transceiver. The level sensor determines an amount of the product in the container by comparing a characteristic of the reflected signal, such as time the reflected signal was received, to the characteristic of the output signal, such as the time the output signal was transmitted. The level sensor may also determine the size of the container based on the reflected signal, and take this size into account when determining the amount of product in the container. The level sensor transmits product level data to a mobile device in response to receiving an interrogation signal therefrom.

Abstract: Systems, methods, and computer program products for monitoring a level of product in a reservoir of a chemical dispensing system. A lid configured to engage an opening of a container includes a level sensor having a transceiver that transmits an output signal and receives a portion of the output signal reflected back to the transceiver. The level sensor determines an amount of the product in the container by comparing a characteristic of the reflected signal, such as time the reflected signal was received, to the characteristic of the output signal, such as the time the output signal was transmitted. The level sensor may also determine the size of the container based on the reflected signal, and take this size into account when determining the amount of product in the container. The level sensor transmits product level data to a mobile device in response to receiving an interrogation signal therefrom.

Abstract: Methods and systems for beam search decoding. One of the methods includes initializing beam data specifying a set of k candidate output sequences and a respective total score for each of the candidate output sequences; updating the beam data at each of a plurality of decoding steps, comprising, at each decoding step: generating a score distribution that comprises a respective score for each token in the vocabulary; identifying a plurality of expanded sequences; generating, for each expanded sequence, a respective backwards-looking score; generating, for each expanded sequence, a respective forward-looking score; computing, for each expanded sequence, a respective total score from the respective forward-looking score for the expanded sequence and the respective backwards-looking score for the expanded sequence; and updating the set of k candidate output sequences using the respective total scores for the expanded sequences.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training a neural network to perform a machine learning task through reinforcement learning. In one aspect, the training uses importance weights generated using standardized absolute deviations of quality scores generated by the neural network for candidate network outputs.

Abstract: A liquid dispenser comprising a liquid pump. The liquid pump comprises a first liquid inlet configured to allow the introduction of a first liquid into a mixing chamber; a second liquid inlet configured to allow the introduction of a second liquid into the mixing chamber; an outlet valve configured to regulate the release of a mixed liquid from the mixing chamber, the mixed liquid being a blend of the first liquid and the second liquid; and a reciprocating member configured to effect a reciprocating movement along a longitudinal axis, the reciprocating member being configured to regulate the aperture of the outlet valve. The liquid dispenser comprises a piezoelectric detector arranged in such a way that when the liquid pump generates a shockwave, the shockwave is detected by the piezoelectric detector, the piezoelectric detector generating a voltage peak which is a function of the shockwave.

Abstract: There is described a neural network system for generating a graph, the graph comprising a set of nodes and edges. The system comprises one or more neural networks configured to represent a probability distribution over sequences of node generating decisions and/or edge generating decisions, and one or more computers configured to sample the probability distribution represented by the one or more neural networks to generate a graph.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for selecting an input vocabulary for a machine learning model using power indices. One of the methods includes computing a respective score for each of a plurality of text tokens in an initial vocabulary and then selecting the text tokens in the input vocabulary based on the respective scores.

Abstract: A liquid dispenser comprising a liquid pump. The liquid pump comprises a first liquid inlet configured to allow the introduction of a first liquid into a mixing chamber; a second liquid inlet configured to allow the introduction of a second liquid into the mixing chamber; an outlet valve configured to regulate the release of a mixed liquid from the mixing chamber, the mixed liquid being a blend of the first liquid and the second liquid; and a reciprocating member configured to effect a reciprocating movement along a longitudinal axis, the reciprocating member being configured to regulate the aperture of the outlet valve. The liquid dispenser comprises a piezoelectric detector arranged in such a way that when the liquid pump generates a shockwave, the shockwave is detected by the piezoelectric detector, the piezoelectric detector generating a voltage peak which is a function of the shockwave.

Abstract: There is described a neural network system for generating a graph, the graph comprising a set of nodes and edges. The system comprises one or more neural networks configured to represent a probability distribution over sequences of node generating decisions and/or edge generating decisions, and one or more computers configured to sample the probability distribution represented by the one or more neural networks to generate a graph.

Abstract: A system incorporating a fluid dispenser for dispensing chemicals includes at least one sensor for sensing fluid flow and a dispenser management unit. The dispenser management unit includes a processor and corresponding storage memory for storing firmware, at least one input for receiving sensor data from the or each sensor, a data transfer device comprising storage memory for storing the received sensor data, and a power source for powering at least the processor. The system further incorporates a portable terminal configured to receive data from the dispenser, and a remote server configured to receive the data from the portable terminal.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating a target sequence from an input sequence. In one aspect, a method comprises maintaining a set of current hypotheses, wherein each current hypothesis comprises an input prefix and an output prefix. For each possible combination of input and output prefix length, the method extends any current hypothesis that could reach the possible combination to generate respective extended hypotheses for each such current hypothesis; determines a respective direct score for each extended hypothesis using a direct model; determines a first number of highest-scoring hypotheses according to the direct scores; rescores the first number of highest-scoring hypotheses using a noisy channel model to generate a reduced number of hypotheses; and adds the reduced number of hypotheses to the set of current hypotheses.

Abstract: A liquid dispenser comprising a liquid pump. The liquid pump comprises a first liquid inlet configured to allow the introduction of a first liquid into a mixing chamber; a second liquid inlet configured to allow the introduction of a second liquid into the mixing chamber; an outlet valve configured to regulate the release of a mixed liquid from the mixing chamber, the mixed liquid being a blend of the first liquid and the second liquid; and a reciprocating member configured to effect a reciprocating movement along a longitudinal axis, the reciprocating member being configured to regulate the aperture of the outlet valve. The liquid dispenser comprises a piezoelectric detector arranged in such a way that when the liquid pump generates a shockwave, the shockwave is detected by the piezoelectric detector, the piezoelectric detector generating a voltage peak which is a function of the shockwave.

Abstract: The present invention relates to a system (300) for monitoring fluid dispensers, comprising: at least one fluid dispenser (100) for dispensing chemicals, comprising a sensor (116) for sensing fluid flow; a dispenser management unit (102) comprising: a processor (120), and corresponding storage memory for storing firmware; at least one input for receiving sensor data from the or each sensor; a data transfer device (126) comprising storage memory for storing said received sensor data; and a power source (122) for powering at least said processor; a portable terminal (302) comprising a device reader and a terminal communications module; and a remote server (304) comprising a server communications module, wherein said remote server is configured to communicate with said terminal via a communications link between said terminal communications module and said server communications module; wherein said terminal further comprises storage memory for storing sensor data received from the or each fluid dispenser data tra

Abstract: There is described a neural network system for generating a graph, the graph comprising a set of nodes and edges. The system comprises one or more neural networks configured to represent a probability distribution over sequences of node generating decisions and/or edge generating decisions, and one or more computers configured to sample the probability distribution represented by the one or more neural networks to generate a graph.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating a target sequence from an input sequence. In one aspect, a method comprises maintaining a set of current hypotheses, wherein each current hypothesis comprises an input prefix and an output prefix. For each possible combination of input and output prefix length, the method extends any current hypothesis that could reach the possible combination to generate respective extended hypotheses for each such current hypothesis; determines a respective direct score for each extended hypothesis using a direct model; determines a first number of highest-scoring hypotheses according to the direct scores; rescores the first number of highest-scoring hypotheses using a noisy channel model to generate a reduced number of hypotheses; and adds the reduced number of hypotheses to the set of current hypotheses.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating a target sequence from an input sequence. In one aspect, a method comprises maintaining a set of current hypotheses, wherein each current hypothesis comprises an input prefix and an output prefix. For each possible combination of input and output prefix length, the method extends any current hypothesis that could reach the possible combination to generate respective extended hypotheses for each such current hypothesis; determines a respective direct score for each extended hypothesis using a direct model; determines a first number of highest-scoring hypotheses according to the direct scores; rescores the first number of highest-scoring hypotheses using a noisy channel model to generate a reduced number of hypotheses; and adds the reduced number of hypotheses to the set of current hypotheses.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating a target sequence from an input sequence. In one aspect, a method comprises maintaining a set of current hypotheses, wherein each current hypothesis comprises an input prefix and an output prefix. For each possible combination of input and output prefix length, the method extends any current hypothesis that could reach the possible combination to generate respective extended hypotheses for each such current hypothesis; determines a respective direct score for each extended hypothesis using a direct model; determines a first number of highest-scoring hypotheses according to the direct scores; rescores the first number of highest-scoring hypotheses using a noisy channel model to generate a reduced number of hypotheses; and adds the reduced number of hypotheses to the set of current hypotheses.