Abstract: A method of labeling a dataset includes inputting a testing set comprising a plurality of input data samples into a plurality of pre-trained machine learning models to generate a set of embeddings output by the plurality of pre-trained machine learning models. The method further includes performing an iterative cluster labeling algorithm that includes generating a plurality of clusterings from the set of embeddings, analyzing the plurality of clusterings to identify a target embedding with a highest duster quality, analyzing the target embedding to determine a compactness for each of the plurality of clusterings of the target embedding, and identifying a target cluster among the plurality of clusterings of the target embedding based on the compactness. The method further includes assigning pseudo-labels to the subset of the plurality of input data samples that are members of the target duster.

Abstract: A liquid desiccant regenerator configured to produce a first output stream with a higher concentration of a liquid desiccant than a first input stream. The regenerator also produces a second output stream with a lower concentration of the liquid desiccant than a second input stream. Regeneration of the liquid desiccant in the liquid desiccant regenerator decreases a temperature of the liquid desiccant regenerator. The system includes an air contactor coupled to the first output stream and exposing an input air stream to the first output stream. The first output stream absorbs water from the input air stream to form at least one diluted output desiccant stream. A heat pump of the system is thermally coupled to move the heat from the first output stream to the liquid desiccant regenerator. The heat moved to the liquid desiccant regenerator increases an efficiency of the liquid desiccant regenerator.

Abstract: A system includes an electrochemical regenerator configured to receive a first solution having a first salt concentration and output a second solution having a second salt concentration lower than the first salt concentration and a third solution having a third salt concentration higher than the first salt concentration. The first and second solutions are sent to first and second reservoirs respectively absorb and emit heat in response to a phase change of one of the solutions. The absorption or emission of heat can be used in a heat pump system.

Abstract: A system and method provide a combination of a modular message structure, a priority-based message packing scheme, and a data packet queue management system to optimize the information content of a transmitted message in, for example, the Ocean of Things (OoT) environment. The modular message structure starts with a header that provides critical information and reference points for time and location. The rest of the message is composed of modular data packets, each of which has a data ID section that the message decoder uses for reference when reconstructing the message contents, an optional size section that specifies the length of the following data section if it can contain data of variable length, and a data section that can be compressed in a manner unique to that data type.

Abstract: A system and method are provided wherein encrypted data is stored in modular lockers on a first storage volume associated with a high-power microprocessor deployed in a system. Each encrypted locker maps, for example, to a specific time segment (e.g., one day) which simplifies mounting of the encrypted volume for data access and reduces the locker size for external access of data while encrypted (e.g., via Wifi). New data in a second storage volume associated with and generated by a low-power microprocessor associated with the system gets transferred to the encrypted data store during wake cycles of the high-power microprocessor. To manage space on the first storage volume, time stamps associated with each encrypted locker allows simple removal of files older than a specified time period by removing of any encrypted lockers older than that threshold.

Abstract: A method of labeling a dataset includes inputting a testing set comprising a plurality of input data samples into a plurality of pre-trained machine learning models to generate a set of embeddings output by the plurality of pre-trained machine learning models. The method further includes performing an iterative cluster labeling algorithm that includes generating a plurality of clusterings from the set of embeddings, analyzing the plurality of clusterings to identify a target embedding with a highest duster quality, analyzing the target embedding to determine a compactness for each of the plurality of clusterings of the target embedding, and identifying a target cluster among the plurality of clusterings of the target embedding based on the compactness. The method further includes assigning pseudo-labels to the subset of the plurality of input data samples that are members of the target duster.

Abstract: A liquid desiccant regenerator configured to produce a first output stream with a higher concentration of a liquid desiccant than a first input stream. The regenerator also produces a second output stream with a lower concentration of the liquid desiccant than a second input stream. Regeneration of the liquid desiccant in the liquid desiccant regenerator decreases a temperature of the liquid desiccant regenerator. The system includes an air contactor coupled to the first output stream and exposing an input air stream to the first output stream. The first output stream absorbs water from the input air stream to form at least one diluted output desiccant stream. A heat pump of the system is thermally coupled to move the heat from the first output stream to the liquid desiccant regenerator. The heat moved to the liquid desiccant regenerator increases an efficiency of the liquid desiccant regenerator.

Abstract: A method of labeling a dataset of input samples for a machine learning task includes selecting a plurality of pre-trained machine learning models that are related to a machine learning task. The method further includes processing a plurality of input data samples through each of the pre-trained models to generate a set of embeddings. The method further includes generating a plurality of clusterings from the set of embeddings. The method further includes analyzing, by a processing device, the plurality of clusterings to extract superclusters. The method further includes assigning pseudo-labels to the input samples based on analysis.

Abstract: A liquid desiccant regenerator configured to produce a first output stream with a higher concentration of a liquid desiccant than a first input stream. The regenerator also produces a second output stream with a lower concentration of the liquid desiccant than a second input stream. Regeneration of the liquid desiccant in the liquid desiccant regenerator decreases a temperature of the liquid desiccant regenerator. The system includes an air contactor coupled to the first output stream and exposing an input air stream to the first output stream. The first output stream absorbs water from the input air stream to form at least one diluted output desiccant stream. A heat pump of the system is thermally coupled to move the heat from the first output stream to the liquid desiccant regenerator. The heat moved to the liquid desiccant regenerator increases an efficiency of the liquid desiccant regenerator.

Abstract: An electrochemical system includes a first reservoir comprising a first fluid and a catalyst, wherein the first fluid comprises a reaction mixture that reacts to form first and second products, and a second reservoir comprises a second fluid. A first electrode contacts a redox-active electrolyte material solution and has a reversible redox reaction with the electrolyte material to accept at least one ion. A second electrode contacts a redox-active electrolyte material solution and has a reversible redox reaction with the electrolyte material to drive at least one ion into the second fluid as an electrical potential is supplied. A diluted effluent comprising the second product and the catalyst exits the second reservoir, wherein the second product is removed from the first reservoir via electroosmosis, and optionally concurrently via osmosis, and a product stream comprising the first product exits the first reservoir.

Abstract: A method of labeling a dataset of input samples for a machine learning task includes selecting a plurality of pre-trained machine learning models that are related to a machine learning task. The method further includes processing a plurality of input data samples through each of the pre-trained models to generate a set of embeddings. The method further includes generating a plurality of clusterings from the set of embeddings. The method further includes analyzing, by a processing device, the plurality of clusterings to extract superclusters. The method further includes assigning pseudo-labels to the input samples based on analysis.

Abstract: Embodiments provide a system and method for performing collaborative learning of machine representations of a concept. During operation, the system can receive a user-specified object associated with a user's concept of interest. The system can compute a similarity score between a target feature vector associated with the user-specified object and a respective feature vector for a set of candidate objects. The system can determine, based on the similarity score, a first subset of candidate objects that satisfy a similarity threshold. The system can receive, via a GUI, a first user-feedback associated with a visual representation of the first subset of candidate objects. The first user-feedback can represent an elaboration of a current user's concept of interest. The system can then modify, based on the first user-feedback, the target feature vector and the similarity function, thereby providing an improved model for machine representations of a current user's concept of interest.

Abstract: A system comprises a liquid desiccant regeneration system, a first air contactor stage, and a second air contactor stage. The regeneration system has a first stage with a first concentration output and first diluted output, and a second stage with a second concentration output, different from the first concentration output, and a second diluted output. The first air contactor stage is coupled to the first concentrated output to form a first output air stream having a reduced water content and a first diluted air contactor output. The second air contactor stage is coupled to the second concentrated output to form a second output air stream having a reduced water content and a second diluted air contactor output. Both diluted air contactor outputs are recirculated into the regeneration system, and the output air streams are combined.

Abstract: Embodiments described herein provide a system for improving a classifier by computing a statistic for the utility of sharing data with a second party. The system may encrypt a set of class labels based on a public key/private key pair to obtain a set of encrypted class labels. The system may send a public key and the set of encrypted class labels to a second computing device. The system may receive an encrypted value computed by the second computing device based on the public key. The system may decrypt the encrypted value based on a private key to obtain a decrypted value. The system may then send a pair of encrypted values computed based on the decrypted value to the second computing device. The system may subsequently receive an encrypted utility statistic from the second computing device, and decrypt the encrypted utility statistic to obtain a decrypted utility statistic.

Abstract: A system comprises an electrochemical liquid desiccant regeneration system, a first air contactor, and a second air contactor. The regeneration system comprises a first output stream having a first concentration of liquid desiccant and a second output stream having a second, lower concentration. The first air contactor disposes a first input air stream having a first water concentration in fluid communication with the first output stream to form a first output air stream having a second, lower water concentration and a diluted output stream, which is circulated back into the regeneration system. The second air contactor disposes a second input air stream having a third water concentration in fluid communication with the liquid desiccant output stream to form a second output air stream having a fourth water concentration higher than the third water concentration and a concentrated output stream, which is circulated back into the regeneration system.

Abstract: An electrochemical device comprises a first type of membrane disposed between first and second reservoirs containing an input solution, and a second type of membrane, different from the first type, is disposed between a first redox-active electrolyte chamber and the first reservoir and disposed between a second redox-active electrolyte chamber and the second reservoir. The first type of membrane and one of the second type of membranes form a membrane pair and the pair has an area specific resistance below y=5065.3x3?1331.1x2+90.035x+39 Ohm cm2 when the pair is equilibrated in an electrolyte and for at least part of a range where 0<x<0.4 and x is the mass fraction of salt in the electrolyte.

Abstract: A system includes an electrochemical regenerator configured to receive a first solution having a first salt concentration and output a second solution having a second salt concentration lower than the first salt concentration and a third solution having a third salt concentration higher than the first salt concentration. The first and second solutions are sent to first and second reservoirs respectively absorb and emit heat in response to a phase change of one of the solutions. The absorption or emission of heat can be used in a heat pump system.

Abstract: A method of removing carbon dioxide from an atmosphere and generating hydrogen includes capturing carbon dioxide from the atmosphere in an alkaline capture solution, sending the alkaline capture solution to a series of electrolyzers in a CO2-rich path, wherein each electrolyzer cell raises the acidity of the input CO2-rich solution to produce an acidified CO2-rich solution, removing carbon dioxide from the acidified CO2-rich solution at a carbon dioxide removal unit operation to produce a CO2-poor solution, sending the CO2-poor solution to the series of electrolyzers in a return path, wherein each electrolyzer raises the alkalinity of the return CO2-poor solution to produce a basified CO2-poor solution, wherein a difference in pH between the CO2-rich solution and the CO2-poor solution within each electrolyzer is less than 3, and returning the basified CO2-poor solution to the carbon dioxide capture unit operation.

Abstract: A method includes receiving a user object specified by a user. A similarity score is computed using a similarity function between the user object and one or more candidate objects in a database based on respective feature vectors. A first subset of the one or more candidate objects is presented to the user based on the respective computed similarity scores. First feedback is received from the user about the first subset of candidate objects. The similarity function is adjusted based on the received first feedback.

Abstract: One embodiment provides a system for detecting and reporting information associated with RF signals. The system obtains in-phase and quadrature (IQ) data of the RF signals received at a predetermined center frequency and computes statistics associated with time-dependent changes of the IQ data. Computing the statistics includes: placing IQ data points included in the IQ data into a plurality of bins in an IQ plane; computing, for each IQ data point, a change in position in the IQ plane between the IQ data point and a corresponding IQ data point recorded prior to a predetermined time interval; and computing, for each bin, an average change in position of IQ data points inside the bin. The system assembles a data packet to be transmitted over a bandwidth-limited communication channel. The data packet includes a small number of bits representing the computed statistics associated with the time-dependent changes of the IQ data.