SYSTEM AND METHOD FOR INCREMENTAL LEARNING FOR MACHINE LEARNING TYPE MODELS TO OPTIMISE TRAINING SPEED - Patent application

Abstract

The disclosed systems and methods may include a computer-implemented method for efficient machine learning model training. The method may include obtaining a plurality of training samples for the machine learning model. The method may include, for one or more first training iterations, training the machine learning model using one or more respective first training samples of the plurality of training samples based at least in part on a first regularization scale configured to control the relative effect of one or more regularization techniques. The method may include, for one or more second training iterations, training the machine learning model using one or more respective second training samples of the plurality of training samples based at least in part on a second regularization scale larger than the first regularization scale.

Description

[Technical field]

[0001] Related Applications This application claims priority to and the benefit of U.S. Provisional Patent Application No. 63 / 145,830, which is incorporated herein by reference in its entirety.

[0002] The present disclosure relates generally to the incremental learning of machine learning models. More particularly, the present disclosure relates to the incremental adjustment of regularization during training of machine learning models to optimize training speed. [Background technology]

[0003] Recent advances in machine learning have substantially increased the size and complexity of both machine learning models (e.g., neural networks) and the data used to train them. As an example, training state-of-the-art deep learning models can sometimes require the use of thousands of graphics processing units for weeks at a time, and thus present prohibitive computational costs. Other networks can train quickly, but with costly overhead for large numbers of parameters. Thus, ways to increase training speed and parameter efficiency would substantially increase the availability of computational resources for other tasks. [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] Tan, M. and Le, Q. V. EfficientNet: Rethinking model scaling for convolutional neural networks. ICML, 2019a [Non-Patent Document 2] Gupta, S. and Tan, M. EfficientNet-EdgeTPU: Creating accelerator-optimized neural networks with automl. https: / / ai.googleblog.com / 2019 / 08 / efficientnetedgetpu-creating.html, 2019 [Non-Patent Document 3] Sandler et al., Mobilenetv2: Inverted residuals and linear bottlenecks. CVPR, 2018 [Non-Patent Document 4] Tan et al., Mnasnet: Platform-aware neural architecture search for mobile. CVPR, 2019 [Non-Patent Document 5] Srivastava et al., Dropout: a simple way to prevent neural networks from overfitting, The Journal of Machine Learning Research, 15(1):1929–1958, 2014 [Non-Patent Document 6] Cubuk et al., Randaugment: Practical automated data augmentation with a reduced search space. ECCV, 2020 [Non-Patent Document 7] Zhang et al., Mixup: Beyond empirical risk minimization. ICLR, 2018 Summary of the Invention [Means for solving the problem]

[0005] Aspects and advantages of embodiments of the present disclosure will be set forth in part in the description that follows, or may be learned from the description, or may be learned through practice of the embodiments.

[0006] One exemplary aspect of the present disclosure is directed to a computer-implemented method for efficient machine learning model training. The method may include obtaining a plurality of training samples for the machine learning model by a computing system including one or more computing devices. The method may include, for one or more first training iterations, training the machine learning model using one or more respective first training samples of the plurality of training samples based at least in part on a first regularization scale configured to control the relative effect of one or more regularization techniques by the computing system. The method may include, for one or more second training iterations, training the machine learning model using one or more respective second training samples of the plurality of training samples based at least in part on a second regularization scale larger than the first regularization scale by the computing system.

[0007] Another exemplary aspect of the present disclosure is directed to a computing system for determining a model with an optimized training rate. The computing system may include one or more processors. The computing system may include one or more tangible non-transitory computer-readable media storing computer-readable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations. The operations may include generating a first machine learning model from a defined model search space, the defined model search space including one or more searchable parameters, the first machine learning model including one or more first values ​​for the one or more searchable parameters. The operations may include performing a model training process on the first machine learning model to obtain first training data describing the first training rate. The operations may include generating a second machine learning model from the defined model search space based at least in part on the first training data, the second machine learning model including one or more second values ​​for the one or more searchable parameters, at least one of the one or more second values ​​being different from the one or more first values. The operations may include performing a model training process on a second machine learning based model to obtain second training data describing a second training speed, the second training speed being faster than the first training speed.

[0008] Another exemplary aspect of the present disclosure is directed to one or more tangible non-transitory computer-readable media storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform operations. The operations may include generating a first machine learning model from a defined model search space, the defined model search space including one or more searchable parameters, the first machine learning model including one or more first values ​​for the one or more searchable parameters. The operations may include performing a model training process on the first machine learning model to obtain first training data describing a first training speed. The operations may include generating a second machine learning model from the defined model search space based at least in part on the first training data, the second machine learning model including one or more second values ​​for the one or more searchable parameters, at least one of the one or more second values ​​being different from the one or more first values, the second machine learning model including multiple sequential model stages, each model stage including one or more model layers, the first model stage including fewer model layers than a second model stage of the multiple model stages. The operations may include performing a model training process on a second machine learning based model to obtain second training data describing a second training speed, the second training speed being faster than the first training speed.

[0009] Another exemplary embodiment is directed to one or more tangible non-transitory computer-readable media, the computer-readable media storing a machine learning model including a first sequence of a plurality of Fused-MBConv stages and a second sequence of a plurality of MBConv stages, the second sequence of the plurality of MBConv stages following the first sequence of the plurality of Fused-MBConv stages, and computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform operations, the operations including obtaining model inputs and processing the model inputs with the machine learning model to generate model outputs. In some implementations, the plurality of Fused-MBConv stages are comprised of three Fused-MBConv stages. In some implementations, the three Fused-MBConv stages include a first, a second, and a third Fused-MBConv stage having 2, 4, and 4 layers, respectively. In some implementations, the three Fused-MBConv stages include a first, a second, and a third Fused-MBConv stage having 24, 48, and 64 channels, respectively. In some implementations, the three Fused-MBConv stages include a first, a second, and a third Fused-MBConv stage each having a 3×3 kernel.

[0010] Other aspects of the present disclosure are directed to various systems, apparatus, non-transitory computer-readable media, user interfaces, and electronic devices.

[0011] These and other features, aspects, and advantages of various embodiments of the present disclosure will become better understood with reference to the following description and appended claims. The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the present disclosure and, together with the description, serve to explain related principles.

[0012] Detailed descriptions of embodiments directed to persons skilled in the art are set forth herein and refer to the accompanying figures. [Brief description of the drawings]

[0013] [Figure 1A] FIG. 1 is a block diagram of an example computing system for performing model training using progressive regularization, according to an example embodiment of the present disclosure. [Figure 1B] FIG. 1 is a block diagram of an example computing device that performs model training using progressive regularization, according to an example embodiment of the present disclosure. [Figure 1C] FIG. 1 is a block diagram of an exemplary computing device for implementing generation of a machine learning model with optimized training speed, in accordance with an exemplary embodiment of the present disclosure. [Figure 2A] FIG. 2 illustrates a Fused-MBConv and MBConv architecture according to an exemplary embodiment of the present disclosure. [Figure 2B] FIG. 1 is a block diagram of an example machine learning model generated through an architecture exploration technique for increasing training speed, in accordance with an example embodiment of the present disclosure. [Diagram 3] FIG. 1 is a graphical diagram of an example neural architecture search approach for increasing training speed and accuracy, according to an example embodiment of the present disclosure. [Figure 4] FIG. 1 is a dataflow diagram of an example method for generating a machine learning model with optimized training rate. [Diagram 5] FIG. 1 is a flowchart diagram of an example method for performing model training using progressive regularization, according to an example embodiment of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] Reference numbers that are repeated among the figures identify like features in various implementations.

[0015] overview In general, the present disclosure is directed to the incremental learning of machine learning models. More specifically, the present disclosure relates to the incremental adjustment of regularization during training of a machine learning model to optimize training speed. As an example, a plurality of training samples (e.g., training images, training datasets, etc.) can be obtained for a machine learning model (e.g., a convolutional neural network, a deep learning network, etc.). For one or more training iterations, the machine learning model can be trained with one or more of these training samples based on a first regularization scale. The first regularization scale can be configured to control the relative effect of one or more regularization techniques (e.g., model dropout, training data augmentation, etc.). For one or more second training iterations, the model can be trained at least in part based on a second regularization scale that is larger than the first regularization scale. Furthermore, in some implementations, the complexity of the training samples (e.g., image size, dataset size, etc.) can be incrementally increased in a substantially similar manner. By initially training the model with relatively weak levels of regularization and data complexity and then progressively increasing both parameters, the systems and methods of the present disclosure substantially reduce the computational resources required during earlier training iterations, thus simultaneously increasing the accuracy of the model while increasing the overall speed of training the model.

[0016] To further increase the training efficiency of the machine learning model, the architecture of the model may be selected based on architecture search (e.g., neural architecture search, etc.) that enhances the training efficiency. As an example, a first machine learning model including one or more parameters may be generated from the defined model search space. More specifically, the first machine learning model may include one or more values ​​for one or more parameters. A model training process (e.g., the aforementioned progressive regularization, etc.) may be applied to the first machine learning model to obtain training data describing the first training speed. A second machine learning model including one or more second values ​​different from the one or more first values ​​may be generated from the defined model search space based at least in part on the first training data. A model training process may be performed on the second machine learning model to obtain the second training data. The second training data may describe a second training speed that is faster than the first training speed. In this way, an architecture search technique may be used to optimize for the training speed, thus significantly increasing the overall training speed of the machine learning model.

[0017] In some implementations, the complexity of the training data may be gradually increased according to the regularization scale during training. As an example, a first complexity (e.g., image size, etc.) may be determined for a first set of training images (e.g., 280×280 pixels, etc.). The first set of training images may be used in the training process to train a machine learning model at the first regularization scale. A second complexity greater than the first complexity may be determined for a second set of training images according to a second regularization scale greater than the first one (e.g., 1080×720 pixels, etc.). The second set of training images may be used in the training process to train a machine learning model based on the second regularization scale. In this way, both accuracy and training efficiency (e.g., speed, number of training epochs, etc.) may be substantially increased.

[0018] In some implementations, the training complexity may represent or otherwise describe the difficulty associated with processing the training sample. As an example, the training sample may be image data. Increasing the training complexity of the image data may include augmenting one or more characteristics of the image data (e.g., increasing the resolution, increasing the color of the image data, increasing the number of image features, adding noise to the image, rotating the image, augmenting the image with a second image, etc.). As another example, if the training sample is a polygonal mesh, increasing the training complexity of the polygonal mesh may include increasing the number of polygons included. Thus, it should be broadly understood that adjusting the training complexity of the training sample may be accomplished in any manner that adjusts the difficulty associated with processing the training sample by the respective machine learning model.

[0019] In some implementations, the models of the present disclosure may be or otherwise include attention-based models. For example, the models of the present disclosure may include one or more self-attention mechanisms and / or one or more attention level layers. For example, the models of the present disclosure may be or otherwise include transformer models. In some implementations, the regularization technique may include augmenting the attention scope in these models. For example, the regularization technique may augment the full self-attention layer in some manner (e.g., adjusting the attention weights, modifying the attention architecture, modifying the scope in which attention is determined between the attention heads of the layer, etc.).

[0020] The systems and methods described herein provide several technical effects and benefits. As an example, by gradually increasing the scale of regularization during training, the systems and methods of the present disclosure can substantially reduce the overall complexity and computational resources required to train a model (e.g., less processing power, less memory usage, less power consumption, etc.) compared to previous training techniques that use, for example, continuous-scale regularization. Thus, the proposed techniques can enable more efficient training of machine learning models that use fewer computational resources.

[0021] Moreover, by generating models with architectures optimized for training speed and efficiency, the disclosed systems and methods can further reduce the overall time and computational resources required for model training. For example, by utilizing architecture search techniques to generate segmented machine learning models with increasing complexity (e.g., number of layers, etc.) per segment, the disclosed systems and methods can further leverage progressive regularization during training, thus further reducing computational resource requirements.

[0022] As another exemplary technical effect and benefit, it is known that the incremental adjustment of training complexity (e.g., the size of the training data, the relative "problem" of correctly processing the training data, etc.) generally increases training speed. However, this may also lead to a reduction in model accuracy. By incrementally increasing the training regularization in a corresponding manner, the increase in training speed from incremental training complexity can be enhanced while mitigating any reduction in model accuracy. Thus, the incremental regularization of the model during training can enhance other methods for increasing training speed while substantially reducing or eliminating any reduction in accuracy from those methods.

[0023] As another exemplary technical effect and benefit, meta-learning (e.g., neural architecture search, etc.) techniques often use "early truncation" techniques when generating machine learning models. For example, meta-learning techniques may implement early stopping for a cycle of training iterations based on early results of the training iterations. By using progressive regularization techniques, training iterations that are relatively early in a training cycle have relatively low computational cost. By reducing the overall computational cost associated with early training iterations, the systems and methods of the present disclosure can substantially reduce the adverse effects (e.g., wasted computational resources, etc.) associated with early truncation in meta-learning techniques.

[0024] Thus, exemplary embodiments of the present disclosure are directed to a new family of convolutional networks that have faster training speed and better parameter efficiency than previous models. To develop these models, the exemplary system described herein can use a combination of training-aware neural architecture search and scaling to jointly optimize training speed and parameter efficiency. The models were searched from a search space enriched with new options such as Fused-MBConv. Experiments show that the models proposed herein train much faster than state-of-the-art models and are up to 6.8 times smaller.

[0025] Furthermore, model training can be further sped up by incrementally increasing image size during training. However, this often results in a loss of accuracy. To compensate for this loss of accuracy, the present disclosure proposes improved methods of incremental learning that adaptively adjust regularization (e.g., data augmentation) in addition to image size.

[0026] More specifically, training efficiency has received a lot of attention recently. For example, NFNet aims to improve training efficiency by removing costly batch normalization, some recent works focus on improving training speed by adding attention layers to convolutional networks (ConvNets), and Vision Transformer improves training efficiency on large datasets by using transformer blocks.

[0027] However, these methods often entail costly overhead at large parameter sizes. In contrast, this disclosure uses a combination of training-aware neural architecture search (NAS) and scaling to improve both training speed and parameter efficiency. Given the parameter efficiency of several models known as EfficientNet (see Tan, M. and Le, Q. V. EfficientNet: Rethinking model scaling for convolutional neural networks. ICML, 2019a), this disclosure systematically examines the training bottlenecks in EfficientNet. These studies show that in EfficientNet, (1) training is slow for very large image sizes, (2) deep-wise convolutions are slow in early layers, and (3) scaling up every stage equally is suboptimal.

[0028] Based on these observations, this disclosure provides an enriched search space with additional operations such as Fused-MBConv. This disclosure also applies training-aware NAS and scaling to jointly optimize model accuracy, training speed, and parameter size. The resulting network, which we call EfficientNetV2, trains up to 4x faster than previous models while using up to 6.8x smaller parameter size.

[0029] In some exemplary implementations, training can be further accelerated by gradually increasing the image size during training. Many previous works use smaller image sizes in training, but keep the same regularization for all image sizes, causing a decrease in accuracy. Therefore, keeping the same regularization for different image sizes is not ideal; for the same network, small image sizes lead to small network capacity and therefore require weak regularization, and vice versa, large image sizes require stronger regularization to combat overfitting.

[0030] Based on this insight, the present disclosure proposes an improved method of incremental learning, where in early training epochs, the network may be trained with small image sizes and weak regularization (e.g., dropout and data augmentation), and then the image size may be gradually increased and stronger regularization may be added. This approach can speed up training without causing a loss in accuracy.

[0031] Using improved incremental learning, the proposed example implementation of EfficientNetV2 achieves strong results on ImageNet, CIFAR-10, CIFAR-100, Cars, and Flowers datasets. On ImageNet, EfficientNet V2 achieves the highest accuracy of 85.7% while training 3x-9x faster and up to 6.8x smaller than previous models. EfficientNetV2 and incremental learning also make it easier to train models on larger datasets. For example, ImageNet21k is about 10x larger than ImageNet ILSVRC2012, but EfficientNetV2 can finish training within 2 days using modest computational resources of 32 TPUv3 cores. By pre-training on the publicly available ImageNet21k, EfficientNetV2 achieves the highest accuracy of 87.3% on ImageNet ILSVRC2012, outperforming the state-of-the-art ViT-L / 16 by 2.0% accuracy while training 5x-11x faster.

[0032] Thus, this disclosure provides an advancement in EfficientNetV2, a new lineage of smaller and faster models. As seen by our training-aware NAS and scaling, EfficientNetV2 outperforms previous models in both training speed and parameter efficiency. An improved method of incremental learning adaptively adjusts regularization with image size. This method speeds up training and improves accuracy at the same time. The proposed approach demonstrates up to 11x faster training speed and up to 6.8x better parameter efficiency than prior art on ImageNet, CIFAR, Cars, and Flowers datasets.

[0033] Exemplary EfficientNetV2 Architecture Design In this section, we discuss the training bottleneck of EfficientNet and introduce the proposed training-aware NAS and scaling, as well as the EfficientNetV2 model.

[0034] Overview of EfficientNets EfficientNet is a family of models optimized for FLOPs and parameter efficiency. It leverages NAS to search for a baseline EfficientNet-B0 with a better tradeoff in accuracy and FLOPs. The baseline model is then scaled up with a composite scaling strategy to obtain a family of models B1-B7. Recent studies claim large gains in training or inference speed, but are often worse than EfficientNet in parameter and FLOPs efficiency. The present disclosure improves training speed while preserving parameter efficiency.

[0035] Understanding Training Efficiency This section describes the training bottleneck of EfficientNet (hereafter referred to as EfficientNetV1) and some simple techniques to improve the training speed. Training with very large image sizes is slow, and as pointed out by previous studies, the large image size of EfficientNetV1 results in heavy memory usage. Since the total memory in GPU / TPU is fixed, these models are generally trained with smaller batch sizes, which drastically slows down training. A simple improvement is to apply FixRes by using smaller image sizes for training than for inference. Smaller image sizes lead to less computation and allow larger batch sizes, thus improving the training speed by up to 2.2x. Using smaller image sizes for training also leads to slightly better accuracy. In some implementations, no layers are fine-tuned after training. Another training bottleneck of EfficientNetV1 comes from the extensive deep wise convolutions, which are slow in early layers but effective in later stages. Deep unit convolution has fewer parameters and FLOPs than normal convolution, but often cannot fully utilize advanced accelerators. Recently, Fused-MBConv has been used to better utilize mobile or server accelerators. Fused-MBConv is described in Gupta, S. and Tan, M. EfficientNet-EdgeTPU: Creating accelerator-optimized neural networks with automl. https: / / ai.googleblog.com / 2019 / 08 / efficientnetedgetpu-creating.html, 2019. Fused-MBConv replaces the deep unit conv3×3 and augmented conv1×1 in MBConv with a single normal conv3×3.MBConv is described in Sandler et al., Mobilenetv2: Inverted residuals and linear bottlenecks. CVPR, 2018 and Tan, M. and Le, Q. V. EfficientNet: Rethinking model scaling for convolutional neural networks. ICML, 2019a. The Fused-MBConv and MBConv architectures are shown in Figure 2A.

[0036] To systematically compare these two building blocks, in the exemplary experiments, the original MBConv in EfficientNet-B4 was gradually replaced by Fused-MBConv. When applied in early stages 1-3, Fused-MBConv can improve the training speed with small overhead in parameters and FLOPs, but replacing all blocks with Fused-MBConv (stages 1-7) will significantly increase parameters and FLOPs, as well as slow down training. Finding the right combination of these two building blocks, i.e., MBConv and Fused-MBConv, is not trivial, and is a problem solved by the present disclosure through the use of neural architecture search to automatically search for the best combination. Scaling up every stage equally is suboptimal, and EfficientNetV1 scales up all stages equally using a simple compound scaling rule. For example, when the depth factor is 2, all stages in the network double the number of layers. However, these stages do not contribute equally to the training speed and parameter efficiency. The exemplary implementations of the present disclosure use a non-uniform scaling strategy to gradually add more layers in later stages. Furthermore, v1 EfficientNet scales up the image size aggressively, leading to high memory consumption and slow training. To address this issue, the exemplary implementations of the present disclosure slightly modify the scaling rules and limit the maximum image size to a smaller value.

[0037] Exemplary Training Cognitive NAS and Scaling The exemplary implementations of this disclosure provide multiple design options to improve training speed. To explore the best combination of these options, this section proposes a training-aware NAS. The NAS search, i.e., the exemplary training-aware NAS framework proposed by this disclosure, aims to jointly optimize accuracy, parameter efficiency, and training efficiency in a progressive accelerator. Specifically, the NAS uses EfficientNet as its backbone. The search space can be a stage-based factorization space similar to Tan et al., Mnasnet: Platform-aware neural architecture search for mobile. CVPR, 2019, which consists of design options for convolution operation type {MBConv, Fused-MBConv}, number of layers, kernel size {3×3, 5×5}, and expansion ratio {1, 4, 6}. Meanwhile, the search space size can be optionally reduced by (1) removing unnecessary search options, such as the pooling skip option, since they are never used in the original EfficientNet, and (2) reusing the same channel size from the backbone, since it has already been explored in the original EfficientNet. Since the search space is smaller, the search process can apply reinforcement learning or simply random search in a much larger network with a size comparable to EfficientNet-B4. Specifically, the exemplary search method can sample up to 1000 models and train each model for only about 10 epochs with the reduced image size for training. The exemplary search reward is calculated by the simple weighted product A·S of the model accuracy A, the normalized training step time S, and the parameter size P. w P v where w=-0.07 and v=-0.05 are empirically determined to balance the tradeoffs.

[0038] The following table shows an example architecture for the explored model EfficientNetV2-S. Compared with the EfficientNet backbone, the proposed explored EfficientNetV2 has several key differences. (1) The first difference is that EfficientNetV2 uses both MBConv and the newly added Fused-MBConv extensively in early layers. (2) Second, EfficientNetV2 prefers a smaller expansion ratio for MBConv, since smaller expansion ratios tend to have less memory access overhead. (3) Third, EfficientNetV2 prefers a smaller 3×3 kernel size, which adds more layers to compensate for the reduced receptive field resulting from the smaller kernel size. (4) Finally, EfficientNetV2 completely removes the last stride 1 stage in the original EfficientNet, possibly due to the large parameter size and memory access overhead.

[0039] [Table 1]

[0040] Another exemplary architecture is shown in the table below.

[0041] [Table 2]

[0042] EfficientNetV2 Scaling: Some example implementations can scale up EfficientNetV2-S to obtain EfficientNetV2-M / L using a composite scaling similar to that described in EfficientNetV1, with some additional optimizations made, namely: (1) the maximum inference image size can be limited to 480 because very large images often lead to costly memory and training speed overhead; and (2) as a heuristic, more layers can be gradually added in later stages (e.g., stages 5 and 6) to increase the network capacity without adding significant runtime overhead.

[0043] Example-Based Incremental Learning Method Example Derivation As discussed in the previous section, image size plays an important role in training efficiency. In addition to FixRes, many other works dynamically change the image size during training, which often causes a decrease in accuracy. This decrease in accuracy is likely due to unbalanced regularization, and when training with different image sizes, it is still best to adjust the regularization strength accordingly (rather than using a fixed regularization as in previous works). In fact, it is common for larger models to require stronger regularization to combat overfitting, for example, EfficientNet-B7 uses larger dropout and stronger data augmentation than B0. This disclosure suggests that even for the same network, smaller image sizes lead to smaller network capacity and therefore require weaker regularization, and vice versa, larger image sizes lead to more computation with larger capacity and therefore are more vulnerable to overfitting.

[0044] Incremental Learning by Example with Adaptive Regularization One exemplary training process using improved incremental learning is as follows: in early training epochs, the network is trained with smaller images and weak regularization, so that the network can learn simple representations easily and quickly. Then, the image size can be gradually increased, but by adding stronger regularization, learning is also made more difficult.

[0045] Formally, the entire training has N total steps, and the target image size is the regularization scale

[0046]

number

[0047] S with a list of e Assume that k = 1 , where k represents a type of regularization, such as a dropout rate or a mix-up rate value. Some example implementations divide the training into M stages, and for each stage 1 ≦ i ≦ M, the model trains a training set of 1 ≦ i ≦ M for an image size S i and the regularization scale

[0048]

number

[0049] The final stage M can be trained on the target image size S e and the regularization Φ e For simplicity, some example implementations heuristically choose the initial image size S0 and regularization Φ0, and then use linear interpolation to determine the values ​​for each stage. One example algorithm is given below. At the beginning of each stage, the network inherits all weights from the previous stage. Unlike Transformers, where weights (e.g., position embeddings) may depend on the input length, ConvNet weights do not depend on image size and can therefore be easily inherited.

[0050] Example algorithm for incremental learning with adaptive regularization: Input: initial image size S0 and regularization

[0051]

number

[0052] . Input: Final image size S e and regularization

[0053]

number

[0054] . Input: The total number of training steps, N, and the phase, M. for i=0 to M-1 do Image Size:

[0055]

number

[0056] Regularization:

[0057]

number

[0058] The model,

[0059]

number

[0060] Only steps, S i and R i Train in. end for

[0061] The proposed exemplary implementation of incremental learning is generally compatible with any existing regularization. By way of example, the following types of regularization can be adapted incrementally as described herein:

[0062] Dropout: A network-level regularization that reduces cross-adaptation by randomly dropping channels. Incremental learning can be applied to adjust the dropout rate γ. Dropout is described in Srivastava et al., Dropout: a simple way to prevent neural networks from overfitting, The Journal of Machine Learning Research, 15(1):1929-1958, 2014.

[0063] RandAugment: Image-wise data augmentation with adjustable scale ε. Incremental learning can be applied to adjust the scale. RandAugment is described in Cubuk et al., Randaugment: Practical automated data augmentation with a reduced search space. ECCV, 2020.

[0064] Mixup: Cross image data augmentation. Label (x i ,y i ) and (x j ,y j Given two images with,λ, we combine them with mixup ratio,λ, i.e.,

[0065]

number

[0066] and

[0067]

number

[0068] Incremental learning can be applied to adjust the mixup ratio λ during training. Mixup is described in Zhang et al., Mixup: Beyond empirical risk minimization. ICLR, 2018.

[0069] Exemplary Devices and Systems 1A illustrates a block diagram of an exemplary computing system 100 for performing model training using progressive regularization, according to an exemplary embodiment of the present disclosure. The system 100 includes a user computing device 102, a server computing system 130, and a training computing system 150, communicatively coupled via a network 180.

[0070] The user computing device 102 may be any type of computing device, such as, for example, a personal computing device (e.g., a laptop or desktop), a mobile computing device (e.g., a smartphone or tablet), a game console or controller, a wearable computing device, an embedded computing device, or any other type of computing device.

[0071] The user computing device 102 includes one or more processors 112 and memory 114. The one or more processors 112 may be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller, etc.) and may be a single processor or multiple processors operatively connected. The memory 114 may include one or more non-transitory computer-readable storage media, such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. The memory 114 may store data 116 and instructions 118 that are executed by the processor 112 to cause the user computing device 102 to perform operations.

[0072] In some implementations, the user computing device 102 can store or include one or more machine learning models 120. For example, the machine learning models 120 can be or otherwise include various machine learning models, such as neural networks (e.g., deep neural networks) or other types of machine learning models including nonlinear and / or linear models. The neural networks can include feedforward neural networks, recurrent neural networks (e.g., long short-term memory recurrent neural networks), convolutional neural networks, or other forms of neural networks. Some exemplary machine learning models can leverage attention mechanisms such as self-attention. For example, some exemplary machine learning models can include multi-headed self-attention models (e.g., transformer models).

[0073] In some implementations, one or more machine learning based models 120 may be received from the server computing system 130 over the network 180, stored in the user computing device memory 114, and then used or otherwise implemented by one or more processors 112. In some implementations, the user computing device 102 may implement multiple parallel instances of a single machine learning based model 120 (e.g., for parallel incremental regularization training across multiple instances of the model).

[0074] Additionally or alternatively, one or more machine learning based models 140 may be included in or otherwise stored and implemented by a server computing system 130 that communicates with the user computing device 102 according to a client-server relationship. For example, the machine learning based models 140 may be implemented by the server computing system 140 as part of a web service (e.g., a classification service, a prediction service, etc.). Thus, one or more models 120 may be stored and implemented at the user computing device 102 and / or one or more models 140 may be stored and implemented at the server computing system 130.

[0075] The user computing device 102 may also include one or more user input components 122 that receive user input. For example, the user input component 122 may be a touch-sensitive component (e.g., a touch-sensitive display screen or touchpad) that is sensitive to the touch of a user input object (e.g., a finger or stylus). The touch-sensitive component may be useful for implementing a virtual keyboard. Other exemplary user input components include a microphone, a traditional keyboard, or other means by which a user may provide user input.

[0076] The server computing system 130 includes one or more processors 132 and a memory 134. The one or more processors 132 may be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller, etc.) and may be a single processor or multiple processors operatively connected. The memory 134 may include one or more non-transitory computer-readable storage media, such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. The memory 134 may store data 136 and instructions 138 that are executed by the processor 132 to cause the server computing system 130 to perform operations.

[0077] In some implementations, the server computing system 130 includes or is otherwise implemented by one or more server computing devices. In instances in which the server computing system 130 includes multiple server computing devices, such server computing devices may operate according to a sequential computing architecture, a parallel computing architecture, or some combination thereof.

[0078] As described above, the server computing system 130 can store or otherwise include one or more machine learning models 140. For example, the models 140 can be or otherwise include a variety of machine learning models. Exemplary machine learning models include neural networks or other multi-layer nonlinear models. Exemplary neural networks include feedforward neural networks, deep neural networks, recurrent neural networks, and convolutional neural networks. Some exemplary machine learning models can leverage attention mechanisms such as self-attention. For example, some exemplary machine learning models can include a multi-headed self-attention model (e.g., a transformer model).

[0079] The user computing device 102 and / or the server computing system 130 can train the models 120 and / or 140 by interacting with a training computing system 150 that is communicatively coupled via a network 180. The training computing system 150 may be separate from the server computing system 130 or may be part of the server computing system 130.

[0080] The training computing system 150 includes one or more processors 152 and a memory 154. The one or more processors 152 may be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller, etc.) and may be a single processor or multiple processors operatively connected. The memory 154 may include one or more non-transitory computer-readable storage media, such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. The memory 154 may store data 156 and instructions 158 that are executed by the processor 152 to cause the training computing system 150 to perform operations. In some implementations, the training computing system 150 includes or is otherwise implemented by one or more server computing devices.

[0081] The training computing system 150 may include a model trainer 160 that trains the machine learning based models 120 and / or 140 stored in the user computing device 102 and / or the server computing system 130 using various training or learning techniques, such as, for example, backpropagation. For example, a loss function may be backpropagated through the model to update one or more parameters of the model (e.g., based on the gradient of the loss function). Various loss functions may be used, such as mean squared error, likelihood loss, cross entropy loss, hinge loss, and / or various other loss functions. Gradient descent techniques may be used to iteratively update the parameters for several training iterations.

[0082] In some implementations, performing the error backpropagation may include performing truncated backpropagation over time. The model trainer 160 can implement several generalization techniques (e.g., weight decay, dropout, etc.) to improve the generalization ability of the trained model.

[0083] In particular, the model trainer 160 can train the machine learning model 120 and / or 140 based on a set of training data 162. As an example, the training data 162 can include a plurality of training samples (e.g., training images, training datasets, etc.) for the machine learning model (e.g., model 120, model 140, etc.). For one or more training iterations, the machine learning model 120 / 140 can be trained (e.g., using the model trainer 160) using one or more of these training samples 162 based on a first regularization scale. The first regularization scale can be configured for the model trainer 160 to control the relative effect of one or more regularization techniques (e.g., model dropout, training data augmentation, etc.). For one or more second training iterations, the model 120 / 140 can be trained by the model trainer 160 based at least in part on a second regularization scale that is larger than the first regularization scale. Furthermore, in some implementations, the complexity of the training samples 162 (e.g., image size, dataset size, etc.) may be incrementally increased in a substantially similar manner (e.g., with the model trainer 160). By incrementally increasing the regularization scale over the number of training iterations, the accuracy of the model 120 / 140 may be increased while simultaneously increasing the overall training efficiency of the model 120 / 140.

[0084] In some implementations, if the user has given consent, the training examples may be provided by the user computing device 102. Thus, in such implementations, the model 120 provided to the user computing device 102 may be trained by the training computing system 150 against user-specific data received from the user computing device 102. In some instances, this process may be referred to as individualizing the model.

[0085] The model trainer 160 includes computer logic used to provide the desired functionality. The model trainer 160 can be implemented in hardware, firmware, and / or software controlling a general-purpose processor. For example, in some implementations, the model trainer 160 includes program files stored on a storage device, loaded into memory, and executed by one or more processors. In other implementations, the model trainer 160 includes one or more sets of computer-executable instructions stored in a tangible computer-readable storage medium, such as a RAM, a hard disk, or an optical or magnetic medium.

[0086] Network 180 may be any type of communications network, such as a local area network (e.g., an intranet), a wide area network (e.g., the Internet), or some combination thereof, and may include any number of wired or wireless links. In general, communications over network 180 may be carried over any type of wired and / or wireless connections, using a wide variety of communications protocols (e.g., TCP / IP, HTTP, SMTP, FTP), encodings or formats (e.g., HTML, XML), and / or protection schemes (e.g., VPN, Secure HTTP, SSL).

[0087] In some implementations, the input to the machine learning based model of the present disclosure may be image data. The machine learning based model may process the image data to generate an output. As an example, the machine learning based model may process the image data to generate an image recognition output (e.g., recognition of the image data, latent embedding of the image data, encoded representation of the image data, hash of the image data, etc.). As another example, the machine learning based model may process the image data to generate an image segmentation output. As another example, the machine learning based model may process the image data to generate an image classification output. As another example, the machine learning based model may process the image data to generate an image data modification output (e.g., alteration of the image data, etc.). As another example, the machine learning based model may process the image data to generate an encoded image data output (e.g., encoded and / or compressed representation of the image data, etc.). As another example, the machine learning based model may process the image data to generate an upscaled image data output. As another example, the machine learning based model may process the image data to generate a prediction output.

[0088] In some implementations, the input to the machine learning based model of the present disclosure may be text or natural language data. The machine learning based model may process the text or natural language data to generate an output. As an example, the machine learning based model may process the natural language data to generate a language encoding output. As another example, the machine learning based model may process the text or natural language data to generate a latent text embedding output. As another example, the machine learning based model may process the text or natural language data to generate a transformation output. As another example, the machine learning based model may process the text or natural language data to generate a classification output. As another example, the machine learning based model may process the text or natural language data to generate a text segmentation output. As another example, the machine learning based model may process the text or natural language data to generate a semantic intent output. As another example, the machine learning based model may process the text or natural language data to generate an upscaled text or natural language output (e.g., text or natural language data that is of higher quality than the input text or natural language, etc.). As another example, the machine learning based model may process the text or natural language data to generate a prediction output.

[0089] In some implementations, the input to the machine learning based model of the present disclosure may be audio data. The machine learning based model may process the audio data to generate an output. As an example, the machine learning based model may process the audio data to generate a speech recognition output. As another example, the machine learning based model may process the audio data to generate a speech conversion output. As another example, the machine learning based model may process the audio data to generate a latent embedding output. As another example, the machine learning based model may process the audio data to generate an encoded audio output (e.g., an encoded and / or compressed representation of the audio data, etc.). As another example, the machine learning based model may process the audio data to generate an upscaled audio output (e.g., audio data of higher quality than the input audio data, etc.). As another example, the machine learning based model may process the audio data to generate a text representation output (e.g., a text representation of the input audio data, etc.). As another example, the machine learning based model may process the audio data to generate a predicted output.

[0090] In some implementations, the input to the machine learning based model of the present disclosure may be latent coding data (e.g., a latent space representation of the input, etc.). The machine learning based model may process the latent coding data to generate an output. As an example, the machine learning based model may process the latent coding data to generate a recognition output. As another example, the machine learning based model may process the latent coding data to generate a reconstruction output. As another example, the machine learning based model may process the latent coding data to generate an exploration output. As another example, the machine learning based model may process the latent coding data to generate a reclustering output. As another example, the machine learning based model may process the latent coding data to generate a prediction output.

[0091] In some implementations, the input to the machine learning based model of the present disclosure may be statistical data. The statistical data may be, represent, or otherwise include data calculated and / or derived from some other data source. The machine learning based model may process the statistical data to generate an output. As an example, the machine learning based model may process the statistical data to generate a recognition output. As another example, the machine learning based model may process the statistical data to generate a prediction output. As another example, the machine learning based model may process the statistical data to generate a classification output. As another example, the machine learning based model may process the statistical data to generate a segmentation output. As another example, the machine learning based model may process the statistical data to generate a visualization output. As another example, the machine learning based model may process the statistical data to generate a diagnostic output.

[0092] In some implementations, the input to the machine learning based models of the present disclosure may be sensor data. The machine learning based models may process the sensor data to generate an output. As an example, the machine learning based models may process the sensor data to generate a recognition output. As another example, the machine learning based models may process the sensor data to generate a prediction output. As another example, the machine learning based models may process the sensor data to generate a classification output. As another example, the machine learning based models may process the sensor data to generate a segmentation output. As another example, the machine learning based models may process the sensor data to generate a visualization output. As another example, the machine learning based models may process the sensor data to generate a diagnostic output. As another example, the machine learning based models may process the sensor data to generate a detection output.

[0093] In some cases, the machine learning based model may be configured to perform a task that includes encoding input data for reliable and / or efficient transmission or storage (and / or corresponding decoding). For example, the task may be an audio compression task. The input may include audio data and the output may include compressed audio data. In another example, the input may include visual data (e.g., one or more images or videos) and the output may include compressed visual data, and the task is a visual data compression task. In another example, the task may include generating an embedding for the input data (e.g., input audio or visual data).

[0094] In some cases, the input includes visual data and the task is a computer vision task. In some cases, the input includes pixel data for one or more images and the task is an image processing task. For example, the image processing task may be image classification, where the output is a set of scores, each score corresponding to a different object class and representing a likelihood that one or more images show an object belonging to the object class. The image processing task may be object detection, where the image processing output identifies one or more regions in one or more images and, for each region, a likelihood that the region shows an object of interest. As another example, the image processing task may be image segmentation, where the image processing output defines, for each pixel in one or more images, a respective likelihood for each category in a predetermined set of categories. For example, the set of categories may be foreground and background. As another example, the set of categories may be object classes. As another example, the image processing task may be depth estimation, where the image processing output defines, for each pixel in one or more images, a respective depth value. As another example, the image processing task may be motion estimation, where the network input includes multiple images and the image processing output defines, for each pixel of one of the input images, the scene motion represented in pixels between the images in the network input.

[0095] In some cases, the input includes audio data representing speech and the task is a speech recognition task. The output may include text output that is mapped to the speech. In some cases, the task includes encrypting or decrypting input data. In some cases, the task includes a microprocessor-implemented task such as branch prediction or memory address translation.

[0096] 1A illustrates one exemplary computing system that can be used to implement the present disclosure. Other computing systems may be used. For example, in some implementations, the user computing device 102 may include a model trainer 160 and a training dataset 162. In such implementations, the model 120 can be both trained and used locally on the user computing device 102. In some of such implementations, the user computing device 102 can implement the model trainer 160 to personalize the model 120 based on user-specific data.

[0097] 1B illustrates a block diagram of an exemplary computing device 10 for performing model training using progressive regularization according to an exemplary embodiment of the present disclosure. The computing device 10 may be a user computing device or a server computing device.

[0098] The computing device 10 includes several applications (e.g., applications 1-N). Each application includes its own machine learning library and machine learning model. For example, each application may include a machine learning model. Exemplary applications include a text messaging application, an email application, a dictation application, a virtual keyboard application, a browser application, etc.

[0099] 1B, each application may communicate with several other components of the computing device, such as one or more sensors, a context manager, a device state component, and / or additional components. In some implementations, each application may communicate with each device component using an API (e.g., a public API). In some implementations, the API used by each application is specific to that application.

[0100] 1C illustrates a block diagram of an exemplary computing device 50 for implementing the generation of a machine learning model with an optimized training rate according to an exemplary embodiment of the present disclosure. The computing device 50 may be a user computing device or a server computing device.

[0101] Computing device 50 includes several applications (e.g., applications 1-N). Each application communicates with a central intelligence layer. Exemplary applications include a text messaging application, an email application, a dictation application, a virtual keyboard application, a browser application, etc. In some implementations, each application can communicate with the central intelligence layer (and the models stored therein) using an API (e.g., a common API across all applications).

[0102] The central intelligence layer includes several machine learning models. For example, as shown in FIG. 1C, a respective machine learning model may be provided to each application and managed by the central intelligence layer. In other implementations, two or more applications may share a single machine learning model. For example, in some implementations, the central intelligence layer may provide a single model to all of the applications. In some implementations, the central intelligence layer is included in or otherwise implemented by the operating system of the computing device 50.

[0103] The central intelligence layer can communicate with a central device data layer, which can be a centralized repository of data for computing devices 50. As shown in FIG. 1C, the central device data layer can communicate with several other components of the computing device, such as one or more sensors, a context manager, a device state component, and / or additional components. In some implementations, the central device data layer can communicate with each device component using an API (e.g., a private API).

[0104] Example model layout 2B illustrates a block diagram of an exemplary machine learning model 202 generated through an architecture search technique for increasing training speed, according to an exemplary embodiment of the present disclosure. In some implementations, the machine learning model 202 receives a set of input data 204 (e.g., image data, statistical data, video data, etc.) and is trained to provide output data 210 as a result of receiving the input data 204. More specifically, the machine learning model 202 may include multiple stages. The machine learning model 202 may include a first model stage 206 and a second model stage 208.

[0105] More specifically, the machine learning model 202 may be generated using a neural architecture search method configured to increase training speed. By way of example, the model may include two stages (e.g., a first model stage 206, a second model stage 208, etc.). The first model stage 206 may include two model layers 206A and 206B (e.g., a convolutional layer, a fused convolutional layer, etc.). The second model stage 208 may include three layers 208A-208C.

[0106] More specifically, the architecture of the machine learning model 202 may be configured to facilitate model training using an incrementally increasing regularization scale (e.g., the degree to which a regularization technique affects the training of the model 202, etc.). By way of example, the machine learning model 202 may be trained using an initial regularization scale corresponding to the number of layers (e.g., 206A, 206B, etc.) included in the first model stage 206. The regularization scale may be increased to a second scale, greater than the initial scale, corresponding to the number of layers (e.g., 208A, 208B, 208C, etc.) included in the second model stage 208. In that way, the architecture of the model 202 may include multiple stages, each including a number of layers corresponding to a particular scale of regularization, thus increasing the effectiveness of the incremental regularization during training.

[0107] 3 illustrates a graphical illustration 300 of an exemplary neural architecture search approach for enhancing training speed and accuracy, according to an exemplary embodiment of the present disclosure. More specifically, a training controller 302 can determine the magnitude of regularization to be implemented in a trainer 304. For example, the controller 302 can progressively regularize the training process (e.g., modifying training data, model dropout, etc.) implemented by the trainer 304 over several iterations. The trainer 304 can train the machine learning model 306 for several iterations with the regularization magnitude indicated by the controller 302.

[0108] After an initial number of training iterations, the multi-reward objective function 308 may evaluate the accuracy and training time (e.g., as reported by the trainer 304, etc.) of the machine learning based model 306. Based on this evaluation, the multi-reward objective function 308 may provide feedback to the controller 302, which may then adjust the magnitude of regularization for additional training iterations of the machine learning based model 306.

[0109] As an example, the controller 302 can obtain multiple training samples and provide the training samples to the trainer 304. The controller 302 can then determine an initial magnitude of regularization for implementation in the trainer 304, which is configured to control the relative effect of one or more regularization techniques (e.g., model dropout, training data augmentation, etc.). For example, the controller may determine a relatively weak magnitude of regularization for training and then gradually increase the regularization over time. In some implementations, the controller 302 may also determine an initial complexity of the training samples, corresponding to the magnitude of regularization. As an example, the controller 302 may determine a relatively weak level of training sample complexity, corresponding to a relatively weak level of regularization. For example, if the training samples include image data, the controller may downscale the images (e.g., from 800×600 to 80×60, etc.) to reduce the overall level of complexity.

[0110] The trainer 304 may train the machine learning model 306 based at least in part on the regularization magnitude determined by the controller 302 for one or more initial training iterations. After one or more initial training iterations, the trainer 304 may provide model accuracy and training speed data for evaluation using a multi-reward objective function 308. In some implementations, the multi-reward objective function 308 may be configured to reward both training speed and accuracy, thereby increasing training speed while maintaining a particular, threshold level of accuracy.

[0111] Based on this evaluation, feedback may be provided to the controller 302. Depending on the feedback, the controller 302 may then gradually increase the regularization magnitude and the complexity of the training samples in the trainer 304. The controller 302 may then gradually increase the regularization and sample complexity implemented by the trainer 304 during model training. Thus, by starting with relatively weak regularization and complexity, the trainer 304 may train the machine learning model 306 substantially faster than would be possible using a static level of regularization and sample complexity, thus increasing the training speed and substantially reducing the associated computational resource costs.

[0112] 4 illustrates a data flow diagram 400 of an exemplary method for generating a machine learning model with an optimized training rate. More specifically, a model search architecture 402 (e.g., a neural search architecture, etc.) may include or otherwise access a defined model search space 402A. The defined model search space 402A may be or otherwise include one or more searchable parameters (e.g., number of layers, type of layer (e.g., convolutional layer, fused convolutional layer, fused MB-CONV layer, etc.), learning rate, hyperparameters, layer size, channel size, kernel size, etc.). It should be broadly understood that one or more searchable parameters of the defined model search space 402A may define or otherwise control any particular aspect or implementation of a machine learning model (e.g., 404, 410, etc.).

[0113] Based on the defined model search space 402A, the model search architecture 402 can generate a first machine learning model 404. The first machine learning model 404 can include one or more values ​​for one or more searchable parameters. The first machine learning model 404 can then be trained using a model training process 406 to obtain first training data. The first training data can describe a first training speed (e.g., an amount of time required for training, etc.) for the training process 406 performed on the first machine learning model 404.

[0114] The first training data 408 may be provided to the model search architecture 402. Based on the first training data 408, the model search architecture may generate a second machine learning based model 410 from the defined model search space 402A. The second machine learning based model 410 may include one or more values ​​for one or more respective searchable parameters that are different from one or more values ​​of the first machine learning based model 404. As an example, one value of the first machine learning based model 404 for a searchable parameter may specify that a first type of layer (e.g., a standard convolutional layer, etc.) is used in the first machine learning based model 404. The second machine learning based model may instead include a value for the same parameter that specifies that a different type of layer (e.g., a fused convolutional layer, etc.) is used in the second machine learning based model 410.

[0115] The second machine learning model 410 may be trained using the same model training process 406 to obtain second training data 412. The second training data 412 may describe a second training speed that is faster than the first training speed described by the first training data 408. In that way, the model search architecture may iteratively generate machine learning models with faster and faster training speeds by sampling the defined model search space 402A, thus generating a machine learning model (such as, for example, the second machine learning model 410) with optimal training speed characteristics.

[0116] 5 shows a flow chart diagram of an exemplary method performed in accordance with an exemplary embodiment of the present disclosure. Although FIG. 5 shows steps performed in a specific order for purposes of explanation and discussion, the method of the present disclosure is not limited to the specifically shown order or sequence. Various steps of method 500 may be omitted, rearranged, combined, and / or adapted in various ways without departing from the scope of the present disclosure.

[0117] At 502, the computing system may obtain a plurality of training samples for a machine learning based model. More specifically, the computing system may obtain a plurality of training samples (e.g., image data, dataset data, etc.) of a certain complexity that may be used to train a machine learning based model for one or more tasks.

[0118] At 504, the computing system may train the machine learning based model for the one or more first training iterations based at least in part on the first regularization magnitude. More specifically, for the one or more first training iterations, the computing system may train the machine learning based model with one or more respective first training samples of the plurality of training samples based at least in part on a first regularization magnitude configured to control a relative effect of the one or more regularization techniques.

[0119] At 506, the computing system can train the machine learning based model based at least in part on the second regularization scale for the one or more second training iterations. More specifically, the computing system can train the machine learning based model with one or more respective second training samples of the plurality of training samples based at least in part on a second regularization scale that is larger than the first regularization scale for the one or more second training iterations.

[0120] Additional Disclosures The techniques described herein refer to servers, databases, software applications, and other computer-based systems, as well as actions taken and information sent to and from such systems. The inherent flexibility of computer-based systems allows for a wide variety of possible configurations, combinations, and divisions of tasks and functions among components. For example, the processes described herein may be implemented using a single device or component, or multiple devices or components working in combination. Databases and applications may be implemented on a single system or distributed across multiple systems. Distributed components may operate sequentially or in parallel.

[0121] Although the present subject matter has been described in detail with respect to various specific exemplary embodiments thereof, each example is provided as an explanation, not a limitation of the present disclosure. Upon understanding the above, those skilled in the art can easily create modifications, variations, and equivalents of such embodiments. Thus, the present disclosure does not exclude such modifications, variations, and / or additions to the present subject matter as would be readily apparent to those skilled in the art. For example, features illustrated or described as part of one embodiment may also be used with another embodiment to produce yet a further embodiment. Thus, it is intended that the present disclosure cover such modifications, variations, and equivalents. [Explanation of symbols]

[0122] 10. Computing Devices 50 Computing Devices 100 Computing Systems 102 User Computing Devices 112 processors 114 Memory, computing device memory 120 models, machine learning models 122 User Input Components 130 Server Computing System 132 processors 134 Memory 140 models, machine learning models 150 Training Computing System 152 processors 154 Memory 160 Model Trainer 180 Network 202 Machine Learning Models 302 Training Controller 304 Trainer 306 Machine Learning Models 402 Model Exploration Architecture 404 Machine Learning Models 410 Machine Learning Models

Claims

1. 1. A computer-implemented method for efficient machine learning model training, comprising: obtaining, by a computing system comprising one or more computing devices, a plurality of training samples for a machine learning based model; For one or more first training iterations, training, by the computing system, the machine learning based model using a first training sample of one or more of the plurality of training samples based at least in part on a first regularization magnitude configured to control a relative effect of one or more regularization techniques; For one or more second training iterations, training, by the computing system, the machine learning based model using one or more respective second training samples of the plurality of training samples based at least in part on a second regularization scale that is greater than the first regularization scale; The one or more regularization techniques include: adjusting, by the computing system, a number of model channels of at least one layer of the machine learning model; or adjusting, by the computing system, at least one characteristic of one or more training samples of the plurality of training samples. Computer-implemented method.

2. obtaining the plurality of training samples for the machine learning model further includes determining, by the computing system, a first sample complexity for the one or more first training samples; 2. The computer-implemented method of claim 1, wherein prior to training the machine learning model with the one or more respective second training samples, the method includes determining, by the computing system, a second sample complexity for the one or more second training samples, the second sample complexity being greater than the first sample complexity.

3. The plurality of training samples includes a respective plurality of training images; 3. The computer-implemented method of claim 2, wherein determining the second sample complexity for the one or more second training samples comprises adjusting, by the computing system, a size of one or more second training images, wherein the size of the one or more second training images is greater than a size of one or more first training images.

4. Prior to obtaining the plurality of training samples for the machine learning model, generating, by the computing system using a machine learning model search architecture, an initial machine learning model including one or more first values ​​for one or more respective parameters; determining, by the computing system, a first training speed of the initial machine learning based model; generating, by the computing system using the machine learning model exploration architecture, the machine learning model including one or more second values ​​for the one or more respective parameters, at least one of the one or more second values ​​being different from the one or more first values.

5. 5. The computer-implemented method of claim 4, further comprising determining, by the computing system, a second training rate of the machine learning based model, the second training rate being greater than the first training rate.

6. 6. The computer-implemented method of claim 4 or 5, wherein the machine learning model comprises a plurality of sequential model stages, each model stage comprising one or more layers, and a first model stage comprises fewer layers than a second model stage of the plurality of sequential model stages.

7. The computer-implemented method of claim 1, wherein the second regularization scale is based at least in part on one or more respective training outputs from the one or more first training iterations.

8. 1. A computing system for determining a model having an optimized training rate, comprising: one or more processors; and one or more computer-readable storage media storing computer-readable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, including: generating a first machine learning model from a defined model search space, the defined model search space including one or more searchable parameters, the first machine learning model including one or more first values ​​for the one or more searchable parameters; performing a model training process on the first machine learning model to obtain first training data describing a first training speed; generating a second machine learning based model from the defined model search space based at least in part on the first training data, the second machine learning based model including one or more second values ​​for the one or more explorable parameters, at least one of the one or more second values ​​being different from the one or more first values; performing the model training process on the second machine learning model to obtain second training data describing a second training rate, the second training rate being faster than the first training rate; The model training process is progressively regularized. Computing system.

9. The model layers of the defined model search space include: Convolutional layers, or The computing system of claim 8 , further comprising at least one of a fused convolutional layer.

10. 10. The computing system of claim 8 or 9, wherein the second machine learning type model includes a plurality of sequential model stages, each model stage including one or more model layers, and a first model stage includes fewer model layers than a second model stage of the plurality of sequential model stages.

11. 11. The computing system of claim 8, wherein the first training data further describes a first training accuracy, and the second training data further describes a second training accuracy that is greater than the first training accuracy.

12. 12. The computing system of claim 11, wherein generating the second machine learning based model from the defined model search space is further based at least in part on the first training accuracy.

13. Performing a model training process on the first machine learning model includes: Obtaining a plurality of training samples for the first machine learning model; For one or more first training iterations, training the first machine learning based model using one or more respective first training samples of the plurality of training samples based at least in part on a first regularization magnitude configured to control a relative effect of one or more regularization techniques; For one or more second training iterations, and training the first machine learning based model using one or more respective second training samples of the plurality of training samples based at least in part on a second regularization scale that is greater than the first regularization scale.

14. The plurality of training samples includes a respective plurality of training images; 14. The computing system of claim 13, wherein determining a second sample complexity for the one or more second training samples includes adjusting a size of one or more second training images, the size of the one or more second training images being greater than a size of the one or more first training images.

15. The plurality of training samples includes a respective plurality of training images; 15. The computing system of claim 14, wherein determining the second sample complexity for the one or more second training samples comprises adjusting a size of one or more second training images, the size of the one or more second training images being greater than a size of one or more first training images.

16. 16. The computing system of claim 8, wherein at least one of the one or more parameters is configured to control one or more of a type of model layer or a number of model layers included in the machine learning based model.

17. 17. The computing system of claim 8, wherein the operations further include providing the second machine learning based model as an output.

18. 18. The computing system of claim 8, wherein generating the second machine learning model from the defined model search space further comprises determining a plurality of sequential processing stages for the second machine learning model, each of the plurality of sequential processing stages being associated with one or more model layers, and a number of model layers associated with a second processing stage of the plurality of sequential processing stages being greater than a number of model layers associated with a first processing stage of the plurality of sequential processing stages.

Images