Abstract: Implementations relate to fine-tuning a generative model using training instances that phrase task(s) as instruction(s), and subsequently utilizing the fine-tuned generative model to respond to user input(s). The training instances can include a first training instance. The first training instance can include a first training instance input that includes a formulated user input, side information that provides information to generate a response to the formulated user input, and a complex instruction. The complex instruction is a multi-part instruction including a first description that instructs to respond utilizing side information and a second description that instructs to only use inherent knowledge to respond when the side information provides no useful information for responding. The first training instance can further include a first ground truth response derived from the side information responsive to the formulated user input.

Abstract: A method includes receiving distillation data including a plurality of out-of-domain training utterances. For each particular out-of-domain training utterance of the distillation data, the method includes generating a corresponding augmented out-of-domain training utterance, and generating, using a teacher ASR model trained on training data corresponding to a target domain, a pseudo-label corresponding to the corresponding augmented out-of-domain training utterance. The method also includes distilling a student ASR model from the teacher ASR model by training the student ASR model using the corresponding augmented out-of-domain training utterances paired with the corresponding pseudo-labels generated by the teacher ASR model.

Abstract: A method (400) for using anti-context examples for personalizing a speech recognition model (132) includes receiving audio data (104) corresponding to an utterance (102) spoken by a user (10), and processing, using the speech recognition model, the audio data to generate a transcription (106) of the utterance. The transcription including a misrecognized phrase (144) that was misrecognized in the transcription by the speech recognition model. The method also includes receiving user-corrected text (141) including a corrected phrase (146) that replaces the misrecognized phrase that was misrecognized in the transcription. Based on the misrecognized phrase, the method includes generating an anti-context example (305) including anti-context text (310) containing the misrecognized phrase paired with text-to-speech (TTS) audio data (315) corresponding to a synthesized speech representation of the anti-context text. The method also includes personalizing the speech recognition model based on the anti-context example.

Abstract: Implementations disclosed herein are directed to techniques for enabling decentralized learning of global language models (LMs). Remote processor(s) of a remote system can obtain a global LM that includes a global embedding matrix, generate a global embedding mask for the global embedding matrix using a masking technique, apply the global embedding mask to global embedding matrix to generate a sparsified global LM that includes a masked global embedding matrix that is a masked version of the global embedding matrix, transmit the sparsified global LM to computing device(s) that are participating in a given round of decentralized learning for the global language model, receive corresponding updates from the computing device(s), and cause the global LM to be updated based on the corresponding updates. By generating the global embedding mask and applying it to the global embedding matrix, the transferable size of the global LM is reduced thereby enabling decentralized learning thereof.

Abstract: Implementations described herein identify and correct automatic speech recognition (ASR) misrecognitions. For example, on-device processor(s) of a client device may generate a predicted textual segment that is predicted to correspond to spoken utterance of a user of the client device, and may receive further input that modifies the predicted textual segment to an alternate textual segment. Further, the on-device processor(s) may store these textual segments in on-device storage as a candidate correction pair, and transmit the candidate correction pair to a remote system. Moreover, remote processor(s) of the remote system may determine that the candidate correction pair is an actual correction pair, and may cause client devices to generate updates for a global ASR model for the candidate correction pair. Additionally, the remote processor(s) may distribute the global ASR model to the client devices and/or additional client devices.

Abstract: A method includes receiving a conversational training dataset including a plurality of conversational training samples, each training sample associated with a corresponding conversation and including: corresponding audio data characterizing a corresponding current utterance spoken by a user during a current turn in the corresponding conversation; a corresponding context for the corresponding current utterance including a transcript of a previous turn in the corresponding conversation that precedes the current turn; a corresponding ground-truth transcription of the corresponding current utterance; and a CoT annotation representing a corresponding logical relationship between the corresponding current utterance and the previous turn.

Abstract: Implementations described herein are directed to techniques for mitigating and/or eliminating catastrophic forgetting of a global machine learning (ML) model during decentralized learning thereof. Remote processor(s) of a remote system can initially train a global ML model based on server data that is accessible by the remote system. In subsequent decentralized learning of the global ML model, the remote processor(s) can utilize various checkpoint averaging techniques. As described herein, these various checkpoint averaging techniques can include, but are not limited to, a static checkpoint averaging technique, a dynamic checkpoint averaging techniques, and/or a mixed centralized and decentralized training technique.

Abstract: Implementations are directed to efficient federated learning of machine learning (ML) model(s) through on-the-fly decompression and compression of model parameters, of the ML model(s), when facilitating forward propagation and/or back propagation at client device(s). For example, implementations can transmit, from a remote system to a client device, a compressed on-device ML model that includes some compressed parameters. Further, the client device can, in performing forward propagation and/or back propagation using the on-device ML model, decompress those compressed parameters on-the-fly as the parameters are needed for the propagation. The propagation will utilize the decompressed parameters that were decompressed on the fly.

Abstract: Implementations disclosed herein are directed to techniques for enabling decentralized learning of global language models (LMs). Remote processor(s) of a remote system can obtain a global LM that includes a global embedding matrix, generate a global embedding mask for the global embedding matrix using a masking technique, apply the global embedding mask to global embedding matrix to generate a sparsified global LM that includes a masked global embedding matrix that is a masked version of the global embedding matrix, transmit the sparsified global LM to computing device(s) that are participating in a given round of decentralized learning for the global language model, receive corresponding updates from the computing device(s), and cause the global LM to be updated based on the corresponding updates. By generating the global embedding mask and applying it to the global embedding matrix, the transferable size of the global LM is reduced thereby enabling decentralized learning thereof.

Abstract: A method includes receiving distillation data including a plurality of out-of-domain training utterances. For each particular out-of-domain training utterance of the distillation data, the method includes generating a corresponding augmented out-of-domain training utterance, and generating, using a teacher ASR model trained on training data corresponding to a target domain, a pseudo-label corresponding to the corresponding augmented out-of-domain training utterance. The method also includes distilling a student ASR model from the teacher ASR model by training the student ASR model using the corresponding augmented out-of-domain training utterances paired with the corresponding pseudo-labels generated by the teacher ASR model.

Abstract: Information can be distilled from a global automatic speech recognition (ASR) model to a client ASR model. Many implementations include using an RNN-T model as the ASR model, where the global ASR model includes a global encoder, a joint network, a prediction network, and where the client ASR model includes a client encoder, the joint network, and the prediction network. Various implementations include using principal component analysis (PCA) while training the global ASR model to learn a mean vector and a set of principal components corresponding to the global ASR model. Additional or alternative implementations include training the client ASR model to generate one or more predicted coefficients of the global ASR model.

Abstract: A method includes receiving distillation data including a plurality of out-of-domain training utterances. For each particular out-of-domain training utterance of the distillation data, the method includes generating a corresponding augmented out-of-domain training utterance, and generating, using a teacher ASR model trained on training data corresponding to a target domain, a pseudo-label corresponding to the corresponding augmented out-of-domain training utterance. The method also includes distilling a student ASR model from the teacher ASR model by training the student ASR model using the corresponding augmented out-of-domain training utterances paired with the corresponding pseudo-labels generated by the teacher ASR model.

Abstract: Implementations described herein identify and correct automatic speech recognition (ASR) misrecognitions. For example, on-device processor(s) of a client device may generate a predicted textual segment that is predicted to correspond to spoken utterance of a user of the client device, and may receive further input that modifies the predicted textual segment to an alternate textual segment. Further, the on-device processor(s) may store these textual segments in on-device storage as a candidate correction pair, and transmit the candidate correction pair to a remote system. Moreover, remote processor(s) of the remote system may determine that the candidate correction pair is an actual correction pair, and may cause client devices to generate updates for a global ASR model for the candidate correction pair. Additionally, the remote processor(s) may distribute the global ASR model to the client devices and/or additional client devices.

Abstract: Example aspects of the present disclosure provide a novel, resource-efficient approach for federated machine learning techniques with PTNs. The system can determine a first set of training parameters from a plurality of parameters of the global model. Additionally, the system can generate a random seed, using a random number generator, based on a set of frozen parameters. Moreover, the system can transmit, respectively to a plurality of client computing devices, a first set of training parameters and the random seed. Furthermore, the system can receive, respectively from the plurality of client computing devices, updates to one or more parameters in the first set of training parameters. Subsequently, the system can aggregate the updates to one or more parameters that are respectively received from the plurality of client computing devices. The system can modify one or more global parameters of the global model based on the aggregation.

Abstract: The present disclosure discloses polyester composites and their preparation methods, and belongs to the technical field of polymer processing and modification. The polyester composites of the present disclosure comprise 65 to 90 parts of polyester, 5 to 35 parts of an elastomer, 0.05 to 3 parts of a chain extender and 0.01 to 5 parts of a functional additive. The polyester composites of the present disclosure not only have ultra-high toughness, but also can maintain high tensile strength, have excellent hydrolysis resistance, can be matched with an antibacterial agent or an antistatic agent to have good antibacterial or antistatic additional functions, can be widely applied to the fields of fibers and fabrics, plastic structural parts, plastic packages or automobile interior parts, and have a wide prospect.

Abstract: Methods and systems for artificial intelligence based medical image segmentation are disclosed. In a method for autonomous artificial intelligence based medical image segmentation, a medical image of a patient is received. A current segmentation context is automatically determined based on the medical image and at least one segmentation algorithm is automatically selected from a plurality of segmentation algorithms based on the current segmentation context. A target anatomical structure is segmented in the medical image using the selected at least one segmentation algorithm.

Abstract: The invention discloses a preparation method of a catechol group modified biomacromolecular scaffold material, comprising: grafting a catechol-containing compound by amidation to obtain modified biomacromolecules; then, allowing dopamine to perform oxidized self-polymerization in a weakly alkaline buffer solution to form polydopamine (PDA) particles with a uniform particle size; next, forming a scaffold which has three cross-linking structures, namely modified biomacromolecules, modified biomacromolecules/PDA, and biomacromolecules/PDA, through interaction between catechol groups, interaction between catechol groups and PDA particles, and interaction between macromolecules and PDA particles in the modified macromolecules respectively; and cross-linking the scaffold with calcium ions, adipic dihydrazide or genipin to further adjust the degree of cross-linking and porosity of the scaffold.

Abstract: Systems and methods are provided for performing medical imaging analysis. Input medical imaging data is received for performing a particular one of a plurality of medical imaging analyses. An output that provides a result of the particular medical imaging analysis on the input medical imaging data is generated using a neural network trained to perform the plurality of medical imaging analyses. The neural network is trained by learning one or more weights associated with the particular medical imaging analysis using one or more weights associated with a different one of the plurality of medical imaging analyses. The generated output is outputted for performing the particular medical imaging analysis.

Abstract: Methods and apparatus for automated medical image analysis using deep learning networks are disclosed. In a method of automatically performing a medical image analysis task on a medical image of a patient, a medical image of a patient is received. The medical image is input to a trained deep neural network. An output model that provides a result of a target medical image analysis task on the input medical image is automatically estimated using the trained deep neural network. The trained deep neural network is trained in one of a discriminative adversarial network or a deep image-to-image dual inverse network.

Abstract: Methods and systems for artificial intelligence based medical image segmentation are disclosed. In a method for autonomous artificial intelligence based medical image segmentation, a medical image of a patient is received. A current segmentation context is automatically determined based on the medical image and at least one segmentation algorithm is automatically selected from a plurality of segmentation algorithms based on the current segmentation context. A target anatomical structure is segmented in the medical image using the selected at least one segmentation algorithm.