Neural nets and the local predictor method used to predict the time series of chemical reactions

Lebender, D.; Schneider, F. W.

doi:10.1021/j100136a019

Cited by 4 publications

(4 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While Figure demonstrates that there is very little distinction of the atomic velocities between class 1 and class 2 trajectories at the transition state, and while it is perhaps remarkable that machine learning algorithms can achieve a correlative model to nearly 70% accuracy, it is possible that the machine learning algorithms are inhibited from improved prediction with initial velocities at the transition state due the large number of possible starting configurations. Additionally, it is also possible that machine learning cannot have improved prediction because very small changes of these starting configurations result in different trajectory outcomes . One hallmark of a chaotic system is that it shows significant sensitivity to small changes in the initial conditions.…”

Section: Results and Discussionmentioning

confidence: 99%

“…Additionally, it is also possible that machine learning cannot have improved prediction because very small changes of these starting configurations result in different trajectory outcomes. 47 One hallmark of a chaotic system is that it shows significant sensitivity to small changes in the initial conditions. In general, a chaotic system is difficult for a machine learning classifier to accurately predict outcomes.…”

Section: ■ Computational Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Machine Learning Analysis of Direct Dynamics Trajectory Outcomes for Thermal Deazetization of 2,3-Diazabicyclo[2.2.1]hept-2-ene

et al. 2020

View full text Add to dashboard Cite

Experimentally, the thermal gas-phase deazetization of 2,3-diazabicyclo[2.2.1]hept-2-ene (1) results in the loss of N 2 and the formation of bicyclo products 3 (exo) and 4 (endo) in a nonstatistical ratio, with preference for the exo product. Here, we report unrestricted M06-2X quasiclassical trajectories initialized from the concerted N 2 ejection transition state that were able to replicate the experimental preference to form 3. We found that the 3:4 ratio results from the relative amounts of very fast (ballistic) exotype trajectories versus trajectories that lead to the 1,3-diradical intermediate 2. These quasiclassical trajectories provided a set of transition-state vibrational, velocity, momenta, and geometric features for the machine learning analysis. A selection of popular supervised classification algorithms (e.g., random forest) provided poor prediction of trajectory outcomes based on only transitionstate vibrational quanta and energy features. However, these machine learning models provided more accurate predictions using atomic velocities and atomic positions, attaining ∼70% accuracy using initial conditions and between 85 and 95% accuracy at later reaction time steps. This increased accuracy allowed the feature importance analysis to reveal that, at the later-time analysis, the methylene bridge out-of-plane bending is correlated with trajectory outcomes for the formation of either the exo product or toward the diradical intermediate. Possible reasons for the struggle of machine learning algorithms to classify trajectories based on transitionstate features is the heavily overlapping feature values, the finite but very large possible vibrational mode combinations, and the possibility of chaos as trajectories propagate. We examined this chaos by comparing a set of nearly identical trajectories that differed by only a very small scaling of the kinetic energies resulting from the transition-state reaction coordinate.

show abstract

Section: Results and Discussionmentioning

confidence: 99%

Section: ■ Computational Methodsmentioning

confidence: 99%

Machine Learning Analysis of Direct Dynamics Trajectory Outcomes for Thermal Deazetization of 2,3-Diazabicyclo[2.2.1]hept-2-ene

et al. 2020

View full text Add to dashboard Cite

show abstract

“…This limitation of the FFNN has also been observed in the prediction of chaotic time series generated by nonlinear chemical reactions. 35 The predictive power of fuzzy logic may be improved by increasing the number of fuzzy sets. In plot (2) of Figure 11, the red trace represents the prediction achieved by a fuzzy logic system having four input variables and three bell-shaped fuzzy sets for each of them.…”

Section: B Prediction Of the Time Seriesmentioning

confidence: 99%

Analysis and prediction of aperiodic hydrodynamic oscillatory time series by feed-forward neural networks, fuzzy logic, and a local nonlinear predictor

Gentili

Gotoda

Dolnik

et al. 2015

Chaos: An Interdisciplinary Journal of Nonlinear Science

View full text Add to dashboard Cite

Forecasting of aperiodic time series is a compelling challenge for science. In this work, we analyze aperiodic spectrophotometric data, proportional to the concentrations of two forms of a thermoreversible photochromic spiro-oxazine, that are generated when a cuvette containing a solution of the spiro-oxazine undergoes photoreaction and convection due to localized ultraviolet illumination. We construct the phase space for the system using Takens' theorem and we calculate the Lyapunov exponents and the correlation dimensions to ascertain the chaotic character of the time series. Finally, we predict the time series using three distinct methods: a feed-forward neural network, fuzzy logic, and a local nonlinear predictor. We compare the performances of these three methods.

show abstract

“…For more complex problems one or more hidden layers are necessary between the input and output layer. Recently we trained such a feed-forward net containing 40 neurons including two hidden layers (3) to predict deterministic chaos in the . We also calculated the four logical functions AND, OR, NAND, and NOR using the bistable region in the NFT mode!…”

mentioning

confidence: 99%

Chemical Visualization of Boolean Functions: A Simple Chemical Computer

1995

Self Cite

View full text Add to dashboard Cite

We present a chemical realization of the Boolean functions AND, OR, NAND, and NOR with a neutralization reaction carried out in three coupled continuous flow stirred tank reactors (CSTR). Two of these CSTR's are used as input reactors, the third reactor marks the output. The chemical reaction is the neutralization of hydrochloric acid (HCl) with sodium hydroxide (NaOH) in the presence of phenolphtalein as an indicator, which is red in alkaline solutions and colorless in acidic solutions representing the two binary states 1 and 0, respectively. The time required for a "chemical computation" is determined by the flow rate of reactant solutions into the reactors since the neutralization reaction itself is very fast. While the acid flow to all reactors is equal and constant, the flow rate of NaOH solution controls the states of the input reactors. The connectivities between the input and output reactors determine the flow rate of NaOH solution into the output reactor, according to the chosen Boolean function. Thus the state of the output reactor depends on the states of the input reactors.

show abstract

Neural nets and the local predictor method used to predict the time series of chemical reactions

Cited by 4 publications

References 9 publications

Machine Learning Analysis of Direct Dynamics Trajectory Outcomes for Thermal Deazetization of 2,3-Diazabicyclo[2.2.1]hept-2-ene

Machine Learning Analysis of Direct Dynamics Trajectory Outcomes for Thermal Deazetization of 2,3-Diazabicyclo[2.2.1]hept-2-ene

Analysis and prediction of aperiodic hydrodynamic oscillatory time series by feed-forward neural networks, fuzzy logic, and a local nonlinear predictor

Chemical Visualization of Boolean Functions: A Simple Chemical Computer

Contact Info

Product

Resources

About