Abstract: A computer-implemented method of training an ensemble machine learning system comprising a plurality of ensemble members. The method includes selecting a shared objective and an objective for each of the ensemble members. The method further includes training each of the ensemble members according to each objective on a training data set, connecting an output of each of the ensemble members to a joint optimization machine learning system to form a consolidated machine learning system, and training the consolidated machine learning system according to the shared objective and the objective for each of the ensemble members on the training data set. The ensemble members can be the same or different types of machine learning systems. Further, the joint optimization machine learning system can be the same or a different type of machine learning system than the ensemble members.

Abstract: A machine learning system includes a coach machine learning system that uses machine learning to help a student machine learning system learn its system. By monitoring the student learning system, the coach machine learning system can learn (through machine learning techniques) “hyperparameters” for the student learning system that control the machine learning process for the student learning system. The machine learning coach could also determine structural modifications for the student learning system architecture. The learning coach can also control data flow to the student learning system.

Abstract: Computer-implemented systems and methods build ensembles for deep learning through parallel data splitting by creating and training an ensemble of up to 2n ensemble members based on a single base network and a selection of n network elements. The ensemble members are created by the “blasting” process, in which training data are selected for each of the up to 2n ensemble members such that each of the ensemble members trains with updates in a different direction from each of the other ensemble members. The ensemble members may also be trained with joint optimization.

Abstract: A system and method for controlling a nodal network. The method includes estimating an effect on the objective caused by the existence or non-existence of a direct connection between a pair of nodes and changing a structure of the nodal network based at least in part on the estimate of the effect. A nodal network includes a strict partially ordered set, a weighted directed acyclic graph, an artificial neural network, and/or a layered feed-forward neural network.

Abstract: A system and method for controlling a nodal network. The method includes estimating an effect on the objective caused by the existence or non-existence of a direct connection between a pair of nodes and changing a structure of the nodal network based at least in part on the estimate of the effect. A nodal network includes a strict partially ordered set, a weighted directed acyclic graph, an artificial neural network, and/or a layered feed-forward neural network.

Abstract: A system and method for controlling a nodal network. The method includes estimating an effect on the objective caused by the existence or non-existence of a direct connection between a pair of nodes and changing a structure of the nodal network based at least in part on the estimate of the effect. A nodal network includes a strict partially ordered set, a weighted directed acyclic graph, an artificial neural network, and/or a layered feed-forward neural network.

Abstract: A computer system uses a pool of predefined functions and pre-trained networks to accelerate the process of building a large neural network or building a combination of (i) an ensemble of other machine learning systems with (ii) a deep neural network. Copies of a predefined function node or network may be placed in multiple locations in a network being built. In building a neural network using a pool of predefined networks, the computer system only needs to decide the relative location of each copy of a predefined network or function. The location may be determined by (i) the connections to a predefined network from source nodes and (ii) the connections from a predefined network to nodes in an upper network. The computer system may perform an iterative process of selecting trial locations for connecting arcs and evaluating the connections to choose the best ones.

Abstract: Computer systems and computer-implemented methods train a machine-learning regression system. The method comprises the step of generating, with a machine-learning generator, output patterns; distorting the output patterns of the generator by a scale factor to generate distorted output patterns; and training the machine-learning regression system to predict the scaling factor, where the regression system receives the distorted output patterns as input and learns and the scaling factor is a target value for the regression system. The method may further comprise, after training the machine-learning regression system, training a second machine-learning generator by back propagating partial derivatives of an error cost function from the regression system to the second machine-learning generator and training the second machine-learning generator using stochastic gradient descent.

Abstract: Machine-learning data generators use an additional objective to avoid generating data that is too similar to any previously known data example. This prevents plagiarism or simple copying of existing data examples, enhancing the ability of a generator to usefully generate novel data. A formulation of generative adversarial network (GAN) learning as the mixed strategy minimax solution of a zero-sum game solves the convergence and stability problem of GANs learning, without suffering mode collapse.

Abstract: A computer-implemented method for analyzing a first neural network via a second neural network according to a differentiable function. The method includes adding a derivative node to the first neural network that receives derivatives associated with a node of the first neural network. The derivative node is connected to the second neural network such that the second neural network can receive the derivatives from the derivative node. The method further includes feeding forward activations in the first neural network for a data item, back propagating a selected differentiable function, providing the derivatives from the derivative node to the second neural network as data, feeding forward the derivatives from the derivative node through the second neural network, and then back propagating a secondary objective through both neural networks. In various aspects, the learned parameters of one or both of the neural networks can be updated according to the back propagation calculations.

Abstract: A computer-implemented method of training an ensemble machine learning system comprising a plurality of ensemble members. The method includes selecting a shared objective and an objective for each of the ensemble members. The method further includes training each of the ensemble members according to each objective on a training data set, connecting an output of each of the ensemble members to a joint optimization machine learning system to form a consolidated machine learning system, and training the consolidated machine learning system according to the shared objective and the objective for each of the ensemble members on the training data set. The ensemble members can be the same or different types of machine learning systems. Further, the joint optimization machine learning system can be the same or a different type of machine learning system than the ensemble members.

Abstract: A deep neural network architecture comprises a stack of strata in which each stratum has its individual input and an individual objective, in addition to being activated from the system input through lower strata in the stack and receiving back propagation training from the system objective back propagated through higher strata in the stack of strata. The individual objective for a stratum may comprise an individualized target objective designed to achieve diversity among the strata. Each stratum may have a stratum support subnetwork with various specialized subnetworks. These specialized subnetworks may comprise a linear subnetwork to facilitate communication across strata and various specialized subnetworks that help encode features in a more compact way, not only to facilitate communication across strata but also to increase interpretability for human users and to facilitate communication with other machine learning systems.

Abstract: A system and method for controlling a nodal network. The method includes estimating an effect on the objective caused by the existence or non-existence of a direct connection between a pair of nodes and changing a structure of the nodal network based at least in part on the estimate of the effect. A nodal network includes a strict partially ordered set, a weighted directed acyclic graph, an artificial neural network, and/or a layered feed-forward neural network.

Abstract: Computer-based systems and methods guide the learning of features in middle layers of a deep neural network. The guidance can be provided by aligning sets of nodes or entire layers in a network being trained with sets of nodes in a reference system. This guidance facilitates the trained network to more efficiently learn features learned by the reference system using fewer parameters and with faster training. The guidance also enables training of a new system with a deeper network, i.e., more layers, which tend to perform better than shallow networks. Also, with fewer parameters, the new network has fewer tendencies to overfit the training data.

Abstract: Systems and methods improve the performance of a network that has converged such that the gradient of the network and all the partial derivatives are zero (or close to zero) by splitting the training data such that, on each subset of the split training data, some nodes or arcs (i.e., connections between a node and previous or subsequent layers of the network) have individual partial derivative values that are different from zero on the split subsets of the data, although their partial derivatives averaged over the whole set of training data is close to zero. The present system and method can create a new network by splitting the candidate nodes or arcs that diverge from zero and then trains the resulting network with each selected node trained on the corresponding cluster of the data. Because the direction of the gradient is different for each of the nodes or arcs that are split, the nodes and their arcs in the new network will train to be different. Therefore, the new network is not at a stationary point.

Abstract: A system and method for controlling a nodal network. The method includes estimating an effect on the objective caused by the existence or non-existence of a direct connection between a pair of nodes and changing a structure of the nodal network based at least in part on the estimate of the effect. A nodal network includes a strict partially ordered set, a weighted directed acyclic graph, an artificial neural network, and/or a layered feed-forward neural network.

Abstract: Systems and methods analyze and correct the vulnerability of individual nodes in a neural network to changes in the input data. The analysis comprises first changing the activation function of one or more nodes to make them more vulnerable. The vulnerability is then measured based on a norm on the vector of partial derivatives of the network objective evaluated on each training data item. The system is made less vulnerable by splitting the data based on the sign of the partial derivative of the network objective with respect to a vulnerable and training new ensemble members on selected subsets from the data split.

Abstract: Computer-implemented systems and methods soft-tie learned parameters of a neural network(s). The soft-tying comprises: applying a common label to the first and second learned parameters; and as part of the training, and in response to the first and second learned parameters having the common label, applying a regularization penalty to a loss function for the first learned parameter upon a determination that the first learned parameter is different than the second learned parameter. The learned parameters can be connection weights, node biases, and/or parametric model statistics. The application of the regularization penalty can be influenced by a soft-tying hyperparameter.

Abstract: Computer-based systems and methods add extra terms to the objective function of machine learning systems (e.g., neural networks) in an ensemble for selected items of training data. This selective training is designed to penalize and decrease any tendency for two or more members of the ensemble to make the same mistake on any item of training data, which should result in improved performance of the ensemble in operation.

Abstract: A machine learning system includes a coach machine learning system that uses machine learning to help a student machine learning system learn its system. By monitoring the student learning system, the coach machine learning system can learn (through machine learning techniques) “hyperparameters” for the student learning system that control the machine learning process for the student learning system. The machine learning coach could also determine structural modifications for the student learning system architecture. The learning coach can also control data flow to the student learning system.