Search NASA⌕ Search

SEARCH · Search NASA

Results for “gradient descent algorithm”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 19 records
All sourcesNASA NTRSDOE OSTI
NASA NTRS2021

Multi-Pass Sequential Mini-Batch Stochastic Gradient Descent Algorithms for Noise Covariance Estimation in Adaptive Kalman Filtering

Estimation of unknown noise covariances in a Kalman filter is a problem of significant practical interest in a wide array of applications. Although this problem has a long history, reliable algorithms for their estimation were scant, and necessary and sufficient conditions for identifiability of the covariances were in dispute until recently. Necessary and sufficient conditions for covariance estimation and a batch estimation algorithm were presented in our previous study. This paper presents stochastic gradient descent algorithms for noise covariance estimation in adaptive Kalman filters that are an order of magnitude faster than the batch method for similar or better root mean square error. More significantly, these algorithms are applicable to non-stationary systems where the noise covariances can occasionally jump up or down by an unknown magnitude. The computational efficiency of the new algorithms stems from adaptive thresholds for convergence, recursive fading memory estimation of the sample cross-correlations of the innovations, and accelerated stochastic gradient descent algorithms. The comparative evaluation of the proposed methods on a number of test cases demonstrates their computational efficiency and accuracy.

Adaptive Kalman filtering↗
NASA NTRS1994

Multigrid one shot methods for optimal control problems: Infinite dimensional control

The multigrid one shot method for optimal control problems, governed by elliptic systems, is introduced for the infinite dimensional control space. ln this case, the control variable is a function whose discrete representation involves_an increasing number of variables with grid refinement. The minimization algorithm uses Lagrange multipliers to calculate sensitivity gradients. A preconditioned gradient descent algorithm is accelerated by a set of coarse grids. It optimizes for different scales in the representation of the control variable on different discretization levels. An analysis which reduces the problem to the boundary is introduced. It is used to approximate the two level asymptotic convergence rate, to determine the amplitude of the minimization steps, and the choice of a high pass filter to be used when necessary. The effectiveness of the method is demonstrated on a series of test problems. The new method enables the solutions of optimal control problems at the same cost of solving the corresponding analysis problems just a few times.

Arian, Eyal↗
NASA NTRS1992

Fast temporal neural learning using teacher forcing

A neural network is trained to output a time dependent target vector defined over a predetermined time interval in response to a time dependent input vector defined over the same time interval by applying corresponding elements of the error vector, or difference between the target vector and the actual neuron output vector, to the inputs of corresponding output neurons of the network as corrective feedback. This feedback decreases the error and quickens the learning process, so that a much smaller number of training cycles are required to complete the learning process. A conventional gradient descent algorithm is employed to update the neural network parameters at the end of the predetermined time interval. The foregoing process is repeated in repetitive cycles until the actual output vector corresponds to the target vector. In the preferred embodiment, as the overall error of the neural network output decreasing during successive training cycles, the portion of the error fed back to the output neurons is decreased accordingly, allowing the network to learn with greater freedom from teacher forcing as the network parameters converge to their optimum values. The invention may also be used to train a neural network with stationary training and target vectors.

Toomarian, Nikzad↗
NASA NTRS1991

Dynamic compensatory pattern matching in a fuzzy rule-based control system

A dynamic compensatory matching procedure is suggested as a method to generate an aggregated measure for evaluating the appropriateness of rules for control systems. It is a dynamic weighted matching technique which takes into account incomplete information under real-time requirements. The initial weights of importance of variables are generated with a generalized neural network architecture and a gradient descent algorithm. An intuitive compensatory scheme based on correlations among input variables of training data is adopted so that the system is coherent to a noisy environment.

Sun, Chuen-Tsai↗
NASA NTRS1991

Fuzzy modeling based on generalized neural networks and fuzzy clustering objective functions

An approach to the formulation of fuzzy if-then rules based on clustering objective functions is proposed. The membership functions are then calibrated with the generalized neural networks technique to achieve a desired input-output mapping. The learning procedure is basically a gradient-descent algorithm. A Kalman filter algorithm is used to improve the overall performance.

Sun, Chuen-Tsai↗
NASA NTRS1995

Fast temporal neural learning using teacher forcing

A neural network is trained to output a time dependent target vector defined over a predetermined time interval in response to a time dependent input vector defined over the same time interval by applying corresponding elements of the error vector, or difference between the target vector and the actual neuron output vector, to the inputs of corresponding output neurons of the network as corrective feedback. This feedback decreases the error and quickens the learning process, so that a much smaller number of training cycles are required to complete the learning process. A conventional gradient descent algorithm is employed to update the neural network parameters at the end of the predetermined time interval. The foregoing process is repeated in repetitive cycles until the actual output vector corresponds to the target vector. In the preferred embodiment, as the overall error of the neural network output decreasing during successive training cycles, the portion of the error fed back to the output neurons is decreased accordingly, allowing the network to learn with greater freedom from teacher forcing as the network parameters converge to their optimum values. The invention may also be used to train a neural network with stationary training and target vectors.

Toomarian, Nikzad↗
NASA NTRS1999

Hybrid Active-Passive Systems for Control of Aircraft Interior Noise

Previous work has demonstrated the large potential for hybrid active-passive systems for attenuating interior noise in aircraft fuselages. The main advantage of an active-passive system is, by utilizing the natural dynamics of the actuator system, the control actuator power and weight is markedly reduced and stability/robustness is enhanced. Three different active-passive approaches were studied in the past year. The first technique utilizes multiple tunable vibration absorbers (ATVA) for reducing narrow band sound radiated from panels and transmitted through fuselage structures. The focus is on reducing interior noise due to propeller or turbo fan harmonic excitation. Two types of tunable vibration absorbers were investigated; a solid state system based upon a piezoelectric mechanical exciter and an electromechanical system based upon a Motran shaker. Both of these systems utilize a mass-spring dynamic effect to maximize tile output force near resonance of the shaker system and so can also be used as vibration absorbers. The dynamic properties of the absorbers (i.e. resonance frequency) were modified using a feedback signal from an accelerometer mounted on the active mass, passed through a compensator and fed into the drive component of the shaker system (piezoelectric element or voice coil respectively). The feedback loop consisted of a two coefficient FIR filter, implemented on a DSP, where the input is acceleration of tile ATVA mass and the output is a force acting in parallel with the stiffness of the absorber. By separating the feedback signal into real and imaginary components, the effective natural frequency and damping of the ATVA can be altered independently. This approach gave control of the resonance frequencies while also allowing the simultaneous removal of damping from the ATVA, thus increasing the ease of controllability and effectiveness. In order to obtain a "tuned" vibration absorber the chosen resonant frequency was set to the excitation frequency. In order to minimize sound radiation a gradient descent algorithm was developed which globally adapted the resonance frequencies of multiple ATVA's while minimizing a cost based upon the radiated sound power or sound energy obtained from an array of microphones.

Fuller, Chris R.↗
NASA NTRS2005

Remote sensing of sea ice thickness by a combined spatial and frequency domain interferometer : formulations, instrument design & development

The thickness of Arctic sea ice plays a critical role in Earth's climate and ocean circulation. An accurate measurement of this parameter on synoptic scales at regular intervals would enable characterization of this important component for the understanding of ocean circulation and the global heat balance. Presented in this paper is a low frequency VHF interferometer technique and associated radar instrument design to measure sea ice thickness based on the use of backscatter correlation functions. The sea ice medium is represented as a multi-layered medium consisting of snow, seaice and sea water, with the interfaces between layers characterized as rough surfaces. This technique utilizes the correlation of two radar waves of different frequencies and incident and observation angles, scattered from the sea ice medium. The correlation functions relate information about the sea ice thickness. Inversion techniques such as the genetic algorithm, gradient descent, and least square methods, are used to derive sea ice thickness from the phase information related by the correlation functions.

remote sensing↗
NASA NTRS2004

Real-Time Adaptive Control of Flow-Induced Cavity Tones

An adaptive generalized predictive control (GPC) algorithm was formulated and applied to the cavity flow-tone problem. The algorithm employs gradient descent to update the GPC coefficients at each time step. The adaptive control algorithm demonstrated multiple Rossiter mode suppression at fixed Mach numbers ranging from 0.275 to 0.38. The algorithm was also able t o maintain suppression of multiple cavity tones as the freestream Mach number was varied over a modest range (0.275 to 0.29). Controller performance was evaluated with a measure of output disturbance rejection and an input sensitivity transfer function. The results suggest that disturbances entering the cavity flow are colocated with the control input at the cavity leading edge. In that case, only tonal components of the cavity wall-pressure fluctuations can be suppressed and arbitrary broadband pressure reduction is not possible. In the control-algorithm development, the cavity dynamics are treated as linear and time invariant (LTI) for a fixed Mach number. The experimental results lend support this treatment.

Kegerise, Michael A.↗
NASA NTRS2006

Real-Time Feedback Control of Flow-Induced Cavity Tones: Adaptive Control - Part 2

An adaptive generalized predictive control (GPC) algorithm was formulated and applied to the cavity flow-tone problem. The algorithm employs gradient descent to update the GPC coefficients at each time step. Past input-output data and an estimate of the open-loop pulse response sequence are all that is needed to implement the algorithm for application at fixed Mach numbers. Transient measurements made during controller adaptation revealed that the controller coefficients converged to a steady state in the mean, and this implies that adaptation can be turned off at some point with no degradation in control performance. When converged, the control algorithm demonstrated multiple Rossiter mode suppression at fixed Mach numbers ranging from 0.275 to 0.38. However, as in the case of fixed-gain GPC, the adaptive GPC performance was limited by spillover in sidebands around the suppressed Rossiter modes. The algorithm was also able to maintain suppression of multiple cavity tones as the freestream Mach number was varied over a modest range (0.275 to 0.29). Beyond this range, stable operation of the control algorithm was not possible due to the fixed plant model in the algorithm.

Kegerise, M. A.↗
NASA NTRS2009

Sequential Principal Component Analysis -An Optimal and Hardware-Implementable Transform for Image Compression

This paper presents the JPL-developed Sequential Principal Component Analysis (SPCA) algorithm for feature extraction / image compression, based on "dominant-term selection" unsupervised learning technique that requires an order-of-magnitude lesser computation and has simpler architecture compared to the state of the art gradient-descent techniques. This algorithm is inherently amenable to a compact, low power and high speed VLSI hardware embodiment. The paper compares the lossless image compression performance of the JPL's SPCA algorithm with the state of the art JPEG2000, widely used due to its simplified hardware implementability. JPEG2000 is not an optimal data compression technique because of its fixed transform characteristics, regardless of its data structure. On the other hand, conventional Principal Component Analysis based transform (PCA-transform) is a data-dependent-structure transform. However, it is not easy to implement the PCA in compact VLSI hardware, due to its highly computational and architectural complexity. In contrast, the JPL's "dominant-term selection" SPCA algorithm allows, for the first time, a compact, low-power hardware implementation of the powerful PCA algorithm. This paper presents a direct comparison of the JPL's SPCA versus JPEG2000, incorporating the Huffman and arithmetic coding for completeness of the data compression operation. The simulation results show that JPL's SPCA algorithm is superior as an optimal data-dependent-transform over the state of the art JPEG2000. When implemented in hardware, this technique is projected to be ideally suited to future NASA missions for autonomous on-board image data processing to improve the bandwidth of communication.

Duong, Tuan A.↗
NASA NTRS2005

A Comparative Study of Probability Collectives Based Multi-agent Systems and Genetic Algorithms

We compare Genetic Algorithms (GA's) with Probability Collectives (PC), a new framework for distributed optimization and control. In contrast to GA's, PC-based methods do not update populations of solutions. Instead they update an explicitly parameterized probability distribution p over the space of solutions. That updating of p arises as the optimization of a functional of p. The functional is chosen so that any p that optimizes it should be p peaked about good solutions. The PC approach works in both continuous and discrete problems. It does not suffer from the resolution limitation of the finite bit length encoding of parameters into GA alleles. It also has deep connections with both game theory and statistical physics. We review the PC approach using its motivation as the information theoretic formulation of bounded rationality for multi-agent systems. It is then compared with GA's on a diverse set of problems. To handle high dimensional surfaces, in the PC method investigated here p is restricted to a product distribution. Each distribution in that product is controlled by a separate agent. The test functions were selected for their difficulty using either traditional gradient descent or genetic algorithms. On those functions the PC-based approach significantly outperforms traditional GA's in both rate of descent, trapping in false minima, and long term optimization.

Huang, Chien-Feng↗
NASA NTRS2004

Algorithm for Training a Recurrent Multilayer Perceptron

An improved algorithm has been devised for training a recurrent multilayer perceptron (RMLP) for optimal performance in predicting the behavior of a complex, dynamic, and noisy system multiple time steps into the future. [An RMLP is a computational neural network with self-feedback and cross-talk (both delayed by one time step) among neurons in hidden layers]. Like other neural-network-training algorithms, this algorithm adjusts network biases and synaptic-connection weights according to a gradient-descent rule. The distinguishing feature of this algorithm is a combination of global feedback (the use of predictions as well as the current output value in computing the gradient at each time step) and recursiveness. The recursive aspect of the algorithm lies in the inclusion of the gradient of predictions at each time step with respect to the predictions at the preceding time step; this recursion enables the RMLP to learn the dynamics. It has been conjectured that carrying the recursion to even earlier time steps would enable the RMLP to represent a noisier, more complex system.

Parlos, Alexander G.↗
NASA NTRS1976

Function-space quasi-Newton algorithms for optimal control problems with bounded controls and singular arcs

Two existing function-space quasi-Newton algorithms, the Davidon algorithm and the projected gradient algorithm, are modified so that they may handle directly control-variable inequality constraints. A third quasi-Newton-type algorithm, developed by Broyden, is extended to optimal control problems. The Broyden algorithm is further modified so that it may handle directly control-variable inequality constraints. From a computational viewpoint, dyadic operator implementation of quasi-Newton methods is shown to be superior to the integral kernel representation. The quasi-Newton methods, along with the steepest descent method and two conjugate gradient algorithms, are simulated on three relatively simple (yet representative) bounded control problems, two of which possess singular subarcs. Overall, the Broyden algorithm was found to be superior. The most notable result of the simulations was the clear superiority of the Broyden and Davidon algorithms in producing a sharp singular control subarc.

Edge, E. R.↗
NASA NTRS1992

Learning and optimization with cascaded VLSI neural network building-block chips

To demonstrate the versatility of the building-block approach, two neural network applications were implemented on cascaded analog VLSI chips. Weights were implemented using 7-b multiplying digital-to-analog converter (MDAC) synapse circuits, with 31 x 32 and 32 x 32 synapses per chip. A novel learning algorithm compatible with analog VLSI was applied to the two-input parity problem. The algorithm combines dynamically evolving architecture with limited gradient-descent backpropagation for efficient and versatile supervised learning. To implement the learning algorithm in hardware, synapse circuits were paralleled for additional quantization levels. The hardware-in-the-loop learning system allocated 2-5 hidden neurons for parity problems. Also, a 7 x 7 assignment problem was mapped onto a cascaded 64-neuron fully connected feedback network. In 100 randomly selected problems, the network found optimal or good solutions in most cases, with settling times in the range of 7-100 microseconds.

Duong, T.↗
NASA NTRSDate not supplied

Efficient Global Optimization with Gradient Finish for Design Under Uncertainty

The Efficient Global Optimization (EGO) algorithm is extended to include a gradient-descent-based finish upon reaching a threshold value of the expected improvement function. Emphasis is placed on efficient evaluation of local gradients using Kriging models during the gradient-based finish to enable application to design under uncertainty (DUU) problems. The modified algorithm is applied to both the well-known Rosenbrock function and a more challenging hypersonic inlet design under uncertainty problem. Results demonstrate improvement in locating the global optimum compared to the classical implementation of EGO, as well as a reduced number of true function evaluations compared to pure gradient-based algorithms. For the Rosenbrock function, a global optimum is returned using an average of 12% fewer function calls than a gradient-based optimizer with comparable tolerances. For the design under uncertainty problem, a global optimum is found using an average of 75% fewer function calls than a gradient-based optimizer. Global Pareto fronts of multiobjective DUU problems are obtained at little additional cost after a single optimization is complete.

Optimization↗
NASA NTRS1992

Learning and tuning fuzzy logic controllers through reinforcements

A new method for learning and tuning a fuzzy logic controller based on reinforcements from a dynamic system is presented. In particular, our Generalized Approximate Reasoning-based Intelligent Control (GARIC) architecture: (1) learns and tunes a fuzzy logic controller even when only weak reinforcements, such as a binary failure signal, is available; (2) introduces a new conjunction operator in computing the rule strengths of fuzzy control rules; (3) introduces a new localized mean of maximum (LMOM) method in combining the conclusions of several firing control rules; and (4) learns to produce real-valued control actions. Learning is achieved by integrating fuzzy inference into a feedforward network, which can then adaptively improve performance by using gradient descent methods. We extend the AHC algorithm of Barto, Sutton, and Anderson to include the prior control knowledge of human operators. The GARIC architecture is applied to a cart-pole balancing system and has demonstrated significant improvements in terms of the speed of learning and robustness to changes in the dynamic system's parameters over previous schemes for cart-pole balancing.

Berenji, Hamid R.↗