-
公开(公告)号：EP3798911A1
公开(公告)日：2021-03-31
申请号：EP19199297.3
申请日：2019-09-24
摘要： Some embodiments are directed to a method (500) to classify sensor data with improved robustness against label noise. A predicted label may be computed for a novel input with improved robustness against label noise by estimating a label which is most likely under repeated application of a base training function to the training labels incorporating noise according to a noise level and subsequent application of a base classifier configured according to the base prediction function to the novel input.
-
公开(公告)号：EP3798913A1
公开(公告)日：2021-03-31
申请号：EP19199318.7
申请日：2019-09-24
发明人： Kolter, Jeremy Zieg , Schmidt, Frank , Wong, Eric , Rice, Leslie
摘要： A computer-implemented method for training a classifier (60) for classifying input signals (x) obtained from a sensor (30), comprising the steps of:
- accessing, from a memory (146), the classifier (60), including providing initial values of parameters ( φ ) characterizing said classifier (60);
- providing a set of training data comprising pairs of training input signals ( x i ) and corresponding training output signals ( y i );
- initializing a set of perturbations ( δ i ) corresponding to each one of said training input signals ( x i );
- performing a first plurality of training epochs, each epoch comprising a second plurality ( N ) of training steps for each input signal ( x i ), each of said training steps comprising
- providing one of said input signals ( x i );
- providing an adversarial input signal x i adv depending on said input signal ( x i ) and the perturbation ( δ i ) that corresponds to said provided input signal ( x i );
- updating the perturbation ( δ i ) based on its value from the previous epoch and a value that characterizes a gradient with respect to said perturbation ( δ i ) to a loss function ( ℓ ) which characterizes a difference between the output signal provided by said classifier (60) if inputted said adversarial input signal x i adv and the training output signal ( y i );
- freezing the value of said updated perturbation ( δ i ) until its update in the next epoch;
- updating said parameters ( φ ) depending on a gradient with respect to said parameters ( φ ) of said loss function ( ℓ ).-
公开(公告)号：EP3772709A1
公开(公告)日：2021-02-10
申请号：EP19190237.8
申请日：2019-08-06
摘要： A neural network may comprise an iterative function ( z [ i +1] = f ( z [ i ] , θ , c ( x )). Such an iterative function is known in the field of machine learning to be representable by a stack of layers which have mutually shared weights. As described in this specification, this stack of layers may during training be replaced by the use of a numerical root-finding algorithm to find an equilibrium of the iterative function in which a further execution of the iterative function would not substantially further change the output of the iterative function. Effectively, the stack of layers may be replaced by a numerical equilibrium solver 480. The use of the numerical root-finding algorithm is demonstrated to greatly reduce the memory footprint during training while achieving similar accuracy as state-of-the-art prior art models.
-
公开(公告)号：EP3832550A1
公开(公告)日：2021-06-09
申请号：EP19212867.6
申请日：2019-12-02
摘要： A computer-implemented method for training a classifier (60), particularly a binary classifier, for classifying input signals ( x i ) to optimize performance according to a non-decomposable metric that measures an alignment between classifications ( y i ) corresponding to input signals ( x i ) of a set of training data and corresponding predicted classifications ( ŷ i ) of said input signals obtained from said classifier, comprising the steps of:
- providing weighting factors that characterize how said non-decomposable metric depends on a plurality of terms from a confusion matrix of said classifications ( y i ) and said predicted classifications ( ŷ i );
- training said classifier (60) depending on said provided weighting factors.-
公开(公告)号：EP3742345A1
公开(公告)日：2020-11-25
申请号：EP19176011.5
申请日：2019-05-22
摘要： A system (100) is disclosed for applying a neural network to an input instance. The neural network comprises an optimization layer for determining values of one or more output neurons from values of one or more input neurons by a joint optimization parametrized by one or more parameters. An input instance is obtained. The values of the one or more input neurons to the optimization layer are obtained and input vectors for the one or more input neurons are determined therefrom. Output vectors for the one or more output neurons are computed from the determined input vectors by jointly optimizing at least the output vectors with respect to the input vectors to solve a semidefinite program defined by the one or more parameters. The values of the one or more output neurons are determined from the respective computed output vectors.
-
公开(公告)号：EP3885848A1
公开(公告)日：2021-09-29
申请号：EP20165352.4
申请日：2020-03-24
发明人： Kolter, Jeremy Zieg , Donti, Priya L. , Fazlyab, Mahyar , Vinogradska, Julia , Roderick, Melrose
IPC分类号： G05B13/02
摘要： Some embodiments are directed to a controller for generating a control signal for a computer-controlled machine. A neural network may be applied to a current sensor signal, the neural network being configured to map the sensor signal to a raw control signal. A projection function may be applied to the raw control signal to obtain a stable control signal to control the computer-controllable machine.
-
公开(公告)号：EP3798862A1
公开(公告)日：2021-03-31
申请号：EP19199308.8
申请日：2019-09-24
摘要： System (300) and computer implemented method (200) for learning rules from a data base (100) comprising entities (110) and relations (120) between said entities, wherein an entity (110) is either a constant (C1, C2, C3, C4, C5, C6, C7) or a numerical value (N1, N2, N3), and a relation (120) between a constant (C6, C7) and a numerical value (N1, N2, N3) is a numerical relation (120b) and a relation (120) between two constants (C1, C2, C3, C4) is a non-numerical relation (120a), the method (200) comprising steps of:
deriving (210) aggregate values from said numerical and/or non-numerical relations;
deriving (220) non-numerical relations from said aggregate values;
adding (230) said derived non-numerical relations to the data base;
constructing (240) differentiable operators, wherein a differentiable operator refers to a non-numerical or a derived non-numerical relation of the data base, and extracting (250) rules from said differentiable operators.-
公开(公告)号：EP3838505A1
公开(公告)日：2021-06-23
申请号：EP19219055.1
申请日：2019-12-20
摘要： The invention relates to a computer-implemented method (700) of configuring a system which interacts with a physical environment. An action of the system in a state of the physical environment results in an updated state of the physical environment according to a transition probability. A safe set of state-action pairs known to be safely performable and an unsafe set of state-action pairs to be avoided are indicated. During an environment interaction, a safe set of state-action pairs is updated by estimating a transition probability for a state-action pair based on an empirical transition probability of a similar other state-action pair, and including the state-action pair in the safe set of state-action pairs only if the state-action pair is not labelled as unsafe and the safe set of state-action pairs can be reached with sufficient probability from the state-action pair based on the estimated transition probability.
-
公开(公告)号：EP3772707A1
公开(公告)日：2021-02-10
申请号：EP19190105.7
申请日：2019-08-05
摘要： A system and computer-implemented method are provided for training a dynamics model to learn the dynamics of a physical system. In particular, the dynamics model may be learned to be able to infer a future state of the physical system and/or its environment based on a current state of the physical system and/or its environment. The learned dynamics model is inherently globally stable. Namely, instead of learning a dynamics model and attempting to separately verify its stability, the learnable dynamics model comprises a learnable Lyapunov function which is jointly learned together with the nominal dynamics of the physical system. Accordingly, the learned dynamics model is highly suitable for real-life applications in which a physical system may assume a state which was unseen during training as the learned dynamics model is inherently globally stable.
-
公开(公告)号：EP3739516A1
公开(公告)日：2020-11-18
申请号：EP19175044.7
申请日：2019-05-17
IPC分类号： G06K9/62
摘要： A system (100) is disclosed for training a classification model to be robust against perturbations of multiple perturbation types. A perturbation type defines a set of allowed perturbations. The classification model is trained by, in an outer iteration, selecting a set of training instances of a training dataset; selecting, among perturbations allowed by the multiple perturbation types, one or more perturbations for perturbing the selected training instances to maximize a loss function; and updating the set of parameters of the classification model to decrease the loss for the perturbed instances. A perturbation is determined by, in an inner iteration, determining updated perturbations allowed by respective perturbation types of the multiple perturbation types and selecting an updated perturbation that most increases the loss of the classification model.
-
-
-
-
-
-
-
-
-
搜索结果
10 条记录 (耗时 0.015 秒)
