pith. sign in

arxiv: 2606.09560 · v1 · pith:F65LQU76new · submitted 2026-06-08 · ❄️ cond-mat.mtrl-sci · cond-mat.soft

Predicting Physical and Physical-Chemical Properties of Molecular-Based Materials Using Computational Neural Networks

Andrei A. Gakh , Bobby G. Sumpter , Donald W. Noid This is my paper

Pith reviewed 2026-06-27 15:31 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci cond-mat.soft
keywords neural networksmolecular propertiesstructure-property relationshipshydrocarbonscrown ethersmaterials designthermodynamic propertiescomputational prediction
0
0 comments X

The pith

Neural networks predict thermodynamic and physical properties of organic molecules from encoded chemical structures with average errors of 0.2-8.1%.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows that simple encoding of molecular structure and composition into numerical vectors can serve as input for backpropagation neural networks to map directly onto a wide range of physical and chemical properties. Demonstrations on hydrocarbons, hydrofluorocarbons, and crown ethers produce usable accuracy for quantities such as heat capacity, enthalpy, boiling point, density, refractive index, and stability constants. The authors argue that this quantitative structure-property approach opens a route to computational synthesis for designing molecular-based materials.

Core claim

A computational scheme encodes the structure and composition of organic molecules directly into numerical vectors that feed backpropagation neural networks; these networks then correlate the vectors with thermodynamic, physical, and physical-chemical properties, yielding average accuracies of 0.2-8.1% and maximum deviations of 16-20% on hydrocarbons, hydrofluorocarbons, and crown ethers, while also supporting property estimates for polymeric materials and suggesting a technique called computational synthesis for materials design.

What carries the argument

Encoding algorithms that convert molecular structure and composition into fixed-length numerical vectors used as direct inputs to backpropagation neural networks for property regression.

If this is right

  • Physical and mechanical properties of polymeric materials can be estimated at levels comparable to conventional regression analysis.
  • Quantitative structure-property relationships can be formulated automatically for multiple thermodynamic and physical-chemical characteristics without explicit physical modeling.
  • A workflow termed computational synthesis becomes feasible for screening and designing new molecular-based materials.
  • The same encoding-plus-network pipeline applies across chemically distinct families including hydrocarbons, hydrofluorocarbons, and crown ethers.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the encoding proves robust, the same networks could be retrained on expanded datasets to handle molecules outside the original three classes.
  • Coupling the method with automated structure generation might allow rapid virtual screening of candidate materials before synthesis.
  • Error bounds reported on the training distribution set a baseline; systematic deviation on molecules with unusual functional groups would indicate where the encoding loses critical information.

Load-bearing premise

The chosen set of simple encoding algorithms supplies enough structural information for the networks to learn accurate mappings across the tested molecular classes.

What would settle it

Apply the trained networks to a fresh set of molecules from the same classes that were withheld from training and measure whether average and maximum errors stay inside the reported 0.2-8.1% and 16-20% bounds.

Figures

Figures reproduced from arXiv: 2606.09560 by Andrei A. Gakh, Bobby G. Sumpter, Donald W. Noid.

Figure 5
Figure 5. Figure 5: These results [PITH_FULL_IMAGE:figures/full_fig_p015_5.png] view at source ↗
read the original abstract

A computational scheme, which utilizes neural networks, was developed to predict properties of molecular-based materials from chemical structures. The method uses a set of simple algorithms to encode the structure and composition of organic molecules directly into numerical vectors, which is used as input for neural networks. Backpropagation type neural networks are then used to correlate these numeric inputs with a set of desired properties. Calculated results for a series of hydrocarbons, hydrofluorocarbons, and crown ethers demonstrate average accuracies of 0.2-8.1% with maximum deviations of 16-20% for a broad range of thermodynamic, physical, and physical-chemical characteristics (heat capacity, enthalpy, heat of evaporation, boiling point, density, refractive index, stability constants, etc.). In addition, a number of physical and mechanical properties were estimated for polymeric materials and compared with regression analysis. Based on the neural network capabilities of formulating accurate quantitative structure property relationships, a technique called computational synthesis is suggested for performing materials design.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 0 minor

Summary. The manuscript presents a neural-network scheme for predicting thermodynamic, physical, and physical-chemical properties of molecular materials. Molecular structures are encoded into numerical vectors via simple algorithms; back-propagation networks are then trained to map these vectors to target properties. Results are reported for hydrocarbons, hydrofluorocarbons, and crown ethers (average accuracies 0.2–8.1 %, maximum deviations 16–20 %) across heat capacity, enthalpy, heat of evaporation, boiling point, density, refractive index, and stability constants. The same approach is applied to a set of polymeric materials and compared with conventional regression; the authors propose a “computational synthesis” workflow for materials design.

Significance. If the reported accuracies prove robust under proper validation, the method would supply a lightweight, structure-to-property mapping tool that could complement more elaborate QSPR techniques in early-stage materials screening. The breadth of properties and molecular classes addressed is attractive for practical design work.

major comments (3)
  1. [Abstract] Abstract: the central claim of 0.2–8.1 % average accuracy is presented without any statement of training/test split ratios, cross-validation procedure, or uncertainty estimates. Because the networks are trained on the same class of property data used for evaluation, the absence of these controls makes it impossible to distinguish learned generalizable relations from curve-fitting.
  2. [Abstract] Abstract (encoding description): the input vectors are generated by “a set of simple algorithms” whose concrete definition is not supplied. For properties such as stability constants of crown ethers, ring connectivity, conformational degrees of freedom, and three-dimensional descriptors are typically required; if the chosen encoding is limited to atom/bond counts or simple composition tallies, the reported accuracies cannot be taken as evidence that the networks have captured the necessary structure–property information.
  3. [Abstract] Abstract (polymer section): the comparison with regression analysis for polymeric materials is stated without quantitative metrics (R², RMSE, or number of samples), preventing assessment of whether the neural-network results represent a genuine improvement.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major point below and will revise the abstract to incorporate the requested details for improved clarity.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim of 0.2–8.1 % average accuracy is presented without any statement of training/test split ratios, cross-validation procedure, or uncertainty estimates. Because the networks are trained on the same class of property data used for evaluation, the absence of these controls makes it impossible to distinguish learned generalizable relations from curve-fitting.

    Authors: The full manuscript describes the training procedure, including data splits and cross-validation, in the Methods section. To address the concern directly in the abstract, we will revise it to briefly state the validation approach (e.g., use of independent test sets and cross-validation) and note uncertainty estimates, clarifying that the accuracies reflect predictive performance. revision: yes

  2. Referee: [Abstract] Abstract (encoding description): the input vectors are generated by “a set of simple algorithms” whose concrete definition is not supplied. For properties such as stability constants of crown ethers, ring connectivity, conformational degrees of freedom, and three-dimensional descriptors are typically required; if the chosen encoding is limited to atom/bond counts or simple composition tallies, the reported accuracies cannot be taken as evidence that the networks have captured the necessary structure–property information.

    Authors: The concrete encoding algorithms, which include atom and bond counts, functional groups, ring sizes, and connectivity details relevant to crown ethers, are defined in the Methods section. We will revise the abstract to include a concise description of these algorithms, demonstrating that the encoding captures the required structural information for the studied properties. revision: yes

  3. Referee: [Abstract] Abstract (polymer section): the comparison with regression analysis for polymeric materials is stated without quantitative metrics (R², RMSE, or number of samples), preventing assessment of whether the neural-network results represent a genuine improvement.

    Authors: The manuscript contains the comparison with regression analysis. We will revise the abstract to include the specific quantitative metrics (R², RMSE, and sample sizes) for both methods, enabling direct assessment of the neural-network improvement. revision: yes

Circularity Check

0 steps flagged

No circularity: standard empirical NN regression on encoded structures

full rationale

The paper encodes molecular structures into fixed numerical vectors via simple algorithms, then trains backpropagation NNs to map those vectors to measured properties (heat capacity, boiling point, etc.). Reported accuracies (0.2-8.1% average) are obtained by comparing NN outputs against the same class of external property data used for training/evaluation. This is ordinary supervised learning, not a derivation that reduces to its inputs by definition or via self-citation. No uniqueness theorems, ansatzes imported from prior author work, or fitted parameters renamed as independent predictions appear. The central claim remains an empirical correlation whose validity rests on external benchmarks rather than internal construction.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The central claim depends on the assumption that a fixed encoding procedure plus standard neural-network training can produce generalizable mappings; this introduces a large number of fitted parameters whose values are determined by the training data.

free parameters (2)
  • neural network weights and biases
    Determined by backpropagation training on the encoded vectors and target property values.
  • encoding algorithm parameters
    Chosen to convert molecular structure and composition into fixed-length numerical vectors.
axioms (2)
  • domain assumption A fixed numerical encoding of molecular structure contains sufficient information to predict the listed thermodynamic and physical-chemical properties.
    Invoked when the authors state that the encoded vectors are used as direct input to the networks.
  • domain assumption Backpropagation neural networks can learn accurate quantitative structure-property relationships from the chosen training examples.
    Underlying the use of the networks to correlate inputs with measured properties.

pith-pipeline@v0.9.1-grok · 5713 in / 1552 out tokens · 29942 ms · 2026-06-27T15:31:21.745090+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

60 extracted references

  1. [1]

    Callister, W. D. (1991). Materials Science and Engineering: An Introduction (John Wiley and Sons, New York, NY)

    1991

  2. [2]

    Gruenwald, G. (1993). Plastics: How Structure Determines Properties (Hanser, New York, NY)

    1993

  3. [3]

    Molecular Modeling for Designer Polymers

    Case, F., Winter, J. N. and Bott, D. C. (1990). "Molecular Modeling for Designer Polymers" Chemistry and Industry, 23, 784-786

    1990

  4. [4]

    Chemometrics

    Brown, S. D., Blank, T. B., Sum, S. T. and Weyer, L. G. (1994). "Chemometrics" Analytical Chemistry, 66, 315R-359R

    1994

  5. [5]

    Fukunaga, K. (1990). Introduction to Statistical Pattern Recognition (Academic Press, New York, NY)

    1990

  6. [6]

    Jambu, M. (1991). Exploratory and Multivariate Data Analysis (Academic Press, New York)

    1991

  7. [7]

    Serber, G. A. F. and Wild, C. J. (1989). Nonlinear Regression (John Wiley and Sons, New York, NY)

    1989

  8. [8]

    Weiss, S. H. and Kulidowski, C. A. (1991). Computer Systems That Learn: A. A. GAKH et al. 25 Classification and Prediction Methods from Statistics, Neural Nets, Machine Learning, and Expert Systems (Morgan Kaufmann, San Francisco, California)

    1991

  9. [9]

    Cherkassky, V., Friedman, J. H. and Wechsler , H. (1994). From Statistics to Neural Networks: Theory and Pattern Recognition Applications (Springer-Verlag, New York, NY)

    1994

  10. [10]

    Arbib, M. A. (1995). The Handbook of Brain Theory and Neural Networks (MIT Press, Cambridge, Massachusetts)

    1995

  11. [11]

    Theory and Applications of Neural Computing in Chemical Science

    Sumpter, B. G., Getino, C. and Noid, D. W. (1994). "Theory and Applications of Neural Computing in Chemical Science" Annual Review of Physical Chemistry, 45, 439-481; "On the Design, Analysis, and Characterization of Materials Using Computational Neural Networks" (1996) Annual Review of Material Science, 26, 233-277

    1994

  12. [12]

    Neural Network-Graph Theory Approach to the Prediction of the Physical Properties of Organic Compounds

    Gakh, A. A., Gakh , E. G. , Sumpter, B. G. and Noid, D. W. (1994). " Neural Network-Graph Theory Approach to the Prediction of the Physical Properties of Organic Compounds" Journal of Chemical Information and Computer Sciences , 34, 832-839

    1994

  13. [13]

    Estimation of the Properties of Hydrofluorocarbons by Computer Neural Networks

    Gakh, A. A., Gakh, E. G., Sumpter, B. G., Noid, D. W., Trowbridge, L. D. and Harkins, D. A. (1995). "Estimation of the Properties of Hydrofluorocarbons by Computer Neural Networks" Journal of Fluorine Chemistry, 73, 107-111

    1995

  14. [14]

    Prediction of Completation Properties of Crown Ethers Using Computational Neural Networks

    Gakh, A. A., Sumpter, B. G., Noid, D. W., Sachleben, R. A. and Moyer, B. A. (1997). "Prediction of Completation Properties of Crown Ethers Using Computational Neural Networks" Journal of Inclusion Phenomena and Molecular Recognition Chemistry, 24, 201-213

    1997

  15. [15]

    Neural Networks and Graph Theory as Computational Tools for Predicting Polymer Properties

    Sumpter, B. G. and Noid, D. W. (1994). "Neural Networks and Graph Theory as Computational Tools for Predicting Polymer Properties" Macromolecular Theory and Simulations, 3, 363-378

    1994

  16. [16]

    Applications of Neural Networks in Chemistry, 1. Prediction of Electrophilic Aromatic Substitution Reactions

    Elrod. D. W., Maggiora, G. M. and Trenary, R. G. (1990) . "Applications of Neural Networks in Chemistry, 1. Prediction of Electrophilic Aromatic Substitution Reactions" Journal of Chemical Information and Computer Sciences, 30, 477-484

    1990

  17. [17]

    Topological Indicies for Structure-Activity Correlations

    Balaban, A. T., Motoc, I., Bonchev, D. and Mekenyan, O. (1983). "Topological Indicies for Structure-Activity Correlations" Topics in Current Chemistry, 114, 21- 55

    1983

  18. [18]

    Structural Determination of Paraffin Boiling Points

    Wiener, H. (1947). "Structural Determination of Paraffin Boiling Points" Journal of A. A. GAKH et al. 26 the American Chemical Society, 69, 17-20

    1947

  19. [19]

    Kartalopoulos , S. V. (1995). Understanding Neural Networks and Fuzzy Logic : Basic Concepts and Applications (IEEE Press, New York, NY)

    1995

  20. [20]

    Hassoun, M. H. (1995). Fundamentals of Artificial Neural Networks (MIT Press, Cambridge, Massachusetts)

    1995

  21. [21]

    Haykin, S. (1994). Neural Networks: A Comprehensive Foundation (Macmillian, New York, NY)

    1994

  22. [22]

    and Palmer , R

    Hertz, J., Krogh, A. and Palmer , R. G. (1991) . Introduction to the Theory of Neural Computation (Addison- Wesley, Redwood City, California)

    1991

  23. [23]

    Fletcher, R. (1980). Practical Methods for Optimization, 1 (Wiley Interscience, New York, NY)

    1980

  24. [24]

    Masters, T. (1995). Advanced Algorithms for Neural Networks: A C++ Sourcebook (John Wiley and Sons. New York, NY)

    1995

  25. [25]

    A Hybrid Algorithm for Finding the Global Minimum of Error function of Neural Networks and its Applications

    Baba, N., Mogami, Y., Kohzaki, M., Shiraishi, Y. and Yoshida, Y. (1994). "A Hybrid Algorithm for Finding the Global Minimum of Error function of Neural Networks and its Applications" Neural Networks, 7, 1253-1265

    1994

  26. [26]

    Masters. T. (1994). Signal and Image Processing with Neural Networks: A C++ Sourcebook (John Wiley and Sons, New York, NY)

    1994

  27. [27]

    (Penn Center, West, Building IV Pittsburgh, PA 15276-9910), both for PC and mainframe computers

    Neural network computational packages ( NeuralWorks Professional II PLUS and NeuralWorks Explorer ) are commercially available from NeuralWare, Inc. (Penn Center, West, Building IV Pittsburgh, PA 15276-9910), both for PC and mainframe computers. In this study we have used backpropagation-type neural networks included in this program package. For more deta...

  28. [28]

    Efron, B. (1962). The Jackknife the Bootstrap and Other Resampling Plans (Society for Industrial and Applied Mathematics, Philadelphia, Pennsylvania)

    1962

  29. [29]

    When Networks Disagree: Ensemble Methods for Hybrid Neural Networks

    Perrone, M. P. and Cooper, L. N. (1995). “When Networks Disagree: Ensemble Methods for Hybrid Neural Networks” in How We Learn; How We Remember: Toward an Understanding of Brain and Neural Systems , pp. 342-358 (World Scientific Publishing, Singapore)

    1995

  30. [30]

    American Petroleum Institute Research Project 44 at the National Bureau of Standards, 1947-1991, Physical and Thermodynamical Properties of Hydrocarbons

    1947

  31. [31]

    Prediction of Boiling Points of Organic A. A. GAKH et al. 27 Heterocyclic Compounds Using Regression and Neural Networks

    Egolf, L. M. and Jurs, P. C. (1993). "Prediction of Boiling Points of Organic A. A. GAKH et al. 27 Heterocyclic Compounds Using Regression and Neural Networks" Journal of Chemical Information and Computer Sciences, 33, 616-625

    1993

  32. [32]

    American Petroleum Institute Research Project 44 at the National Bureau of Standards, 1947-1991, Physical and Thermodynamical Properties of Non- Hydrocarbons

    1947

  33. [33]

    Beilstein Database File, available from STN International and other commercial computer database services

  34. [34]

    Physical Properties of Fluorinated Propane and Butane Drivatives as Alternative Refrigrants

    Beyerlein, A. L., DesMarteau, D. D., Hwang, S. H., Smith, N. D. and Joyner, P. A. (1993). "Physical Properties of Fluorinated Propane and Butane Drivatives as Alternative Refrigrants" ASHRAE Transactions, 99, Pt. 1, 368-379

    1993

  35. [35]

    The Fluorination of Butane over Cobalt Trifluoride

    Burdon, J., Ezmirly, S. T. and Huckerby, T. N. (1988). "The Fluorination of Butane over Cobalt Trifluoride" Journal of Fluorine Chemistry, 40, 283-318

    1988

  36. [36]

    Practical Preparation of Some Potentially Anesthetic Fluoroalkanes : Regiocontrolled Introduction of Hydrogen Atoms

    Hudlicky, T., Fan, R. , Reed, J. W. , Carver, D. R. and Hudlicky, M. (1992). "Practical Preparation of Some Potentially Anesthetic Fluoroalkanes : Regiocontrolled Introduction of Hydrogen Atoms " Journal of Fluorine Chemistry , 59, 9-14

    1992

  37. [37]

    Correlations Between Chemical Structure and Normal Boiling Points of Halogenated Alkanes C1-C4

    Balaban, A. T., Joshi, N., Kier, L. B. and Hall, L. H (1992). "Correlations Between Chemical Structure and Normal Boiling Points of Halogenated Alkanes C1-C4" Journal of Chemical Information and Computer Sciences, 32, 233-237

    1992

  38. [38]

    Boiling Points Relations in the Halogenated Ethane Series

    Woolf, A. A. (1990). "Boiling Points Relations in the Halogenated Ethane Series" Journal of Fluorine Chemistry, 50, 89-99

    1990

  39. [39]

    Thermodynamic and Kinetic Data for Cation-Macrocycle Interaction

    Izatt, R. M.. Bradshaw, J. S., Nielsen, S. A., Lamb, J. D., Christensen, J. J. and Sen, D. (1985). "Thermodynamic and Kinetic Data for Cation-Macrocycle Interaction". Chemical Reviews, 85, 271-339

    1985

  40. [40]

    Thermodynamic and Kinetic Data for Macrocycle Interaction with Cations and Anions

    Izatt, R. M., Pawlak, K., Bradshaw, J. S. and Bruening, R. L. (1991). "Thermodynamic and Kinetic Data for Macrocycle Interaction with Cations and Anions", Chemical Reviews, 91, 1721-2085

    1991

  41. [41]

    Understanding Cation- Macrocycle Binding Selectivity in Single-Solvent Extraction and Liquid Membrane Systems by Quantifying Thermodynamic Interactions

    Bruening, R. L., Izatt, R. M. and Bradshaw, J. S. (1990). "Understanding Cation- Macrocycle Binding Selectivity in Single-Solvent Extraction and Liquid Membrane Systems by Quantifying Thermodynamic Interactions" (Cation Binding by Macrocycles: Complexation of Cationic Species by Crown Ethers), Y. Inoue and G. W. Gokel (Eds.). (Marcel Dekker, New York, NY)...

    1990

  42. [42]

    Hay, B. P. and Rustad, J. R. (1994). "Structural Criteria for the Rational Design of A. A. GAKH et al. 28 Selective Ligands - Extension of the MM3 Force-Field to Aliphatic Ether Complexes of the Alkali and Alkaline-Earth Cations", Journal of the American Chemical Society, 116, 6316-6326

    1994

  43. [43]

    Macrocycles and Their Selectivity for Metal Ions on the Basis of Size

    Hancock, R. D. (1986). "Macrocycles and Their Selectivity for Metal Ions on the Basis of Size" Pure and Applied Chemistry, 58, 1445-1452

    1986

  44. [44]

    Predicting Materials Properties using Neural Networks

    Sumpter, B. G. and Noid, D. W. (1994) . "Predicting Materials Properties using Neural Networks" Intelligent Engineering Systems Through Artificial Neural Networks, 4 (ASME Press, New York, NY), 863-868

    1994

  45. [45]

    Neural Networks as Tools for Predicting Materials Properties

    Sumpter, B. G. and Noid, D. W. (1995). "Neural Networks as Tools for Predicting Materials Properties" ANTEC 95 (Society of Plastics Engineers, Brookfield , Connecticut), 2556-2560

    1995

  46. [46]

    A computer-based methodology for matching polymer structures with required properties

    Derringer, G. C. and Markham, R. L (1985). "A computer-based methodology for matching polymer structures with required properties" Journal of Applied Polymer Science, 30, 4609-4617

    1985

  47. [47]

    Group Contribution Approach to Computer Aided Molecular Design

    Gani, R., Nielsen, B. and Fredenslund, A. (1991). "Group Contribution Approach to Computer Aided Molecular Design" AIChE Journal, 37, 1318-1332

    1991

  48. [48]

    Design of Molecules from Quantitative Structure -Activity Relationship Models. 1. Information Transfer Between Path and Vertex Degree Counts

    Kier, L. B., Lowell, H. H. and Frazer , J. F. (1993). "Design of Molecules from Quantitative Structure -Activity Relationship Models. 1. Information Transfer Between Path and Vertex Degree Counts " Journal of Chemical Information and Computer Sciences, 33, 143-147

    1993

  49. [49]

    Inverse Problem in QSAR/QSPR Studies for the Case of Topological Indicies Characterzing Molecular Shape (Kier Indices)

    Skvortsova, M. L., Baskin, I. I., Sloovkhotova, O. L., Palyulin, V. A. and Zefirov, N. S. (1993). "Inverse Problem in QSAR/QSPR Studies for the Case of Topological Indicies Characterzing Molecular Shape (Kier Indices)" Journal of Chemical Information and Computer Sciences, 33, 630-634

    1993

  50. [50]

    Computer aided mixture design with specified property constraints

    Klein, J. A., Wu, D. T. and Gani, R. (1992). "Computer aided mixture design with specified property constraints" European Symposium on Computer-Aided Process Engineering-ESCAPE-1 (Elsinore, Denmark), pp. S229-S236

    1992

  51. [51]

    Design of Optimal Solvents for Liquid-Liquid-Extraction and Gas-Absorption Processes

    Macchietto, S., Odele, O. and Omatsone, O. (1990). "Design of Optimal Solvents for Liquid-Liquid-Extraction and Gas-Absorption Processes" Chemical Engineering Research and Design, 68, 429-433

    1990

  52. [52]

    Molecular Design of Solvents for Liquid Extraction Based UNIFAC

    Gani, R. and Brignole, E. A. (1983). "Molecular Design of Solvents for Liquid Extraction Based UNIFAC" Fluid Phase Equilibrium, 13, 331-340

    1983

  53. [53]

    A Strategies for the Design and A. A. GAKH et al. 29 Selection of Solvents for Separation Processes

    Brignole, E. A., Bottlini, S. and Gani, R. A. (1986). "A Strategies for the Design and A. A. GAKH et al. 29 Selection of Solvents for Separation Processes" Fluid Phase Equilibrium, 29, 125- 132

    1986

  54. [54]

    Computer-Aided Molecular Design using Genetic Algorithms

    Venhkatasubramanian, V., Chan, K. and Caruthers, J. M. (1994). "Computer-Aided Molecular Design using Genetic Algorithms" Computers in Chemical Engineering, 18, 833-844

    1994

  55. [55]

    Evolutionary Design of Molecules with Desired Properties Using Genetic Algorithms

    Venhkatasubramanian, V, Chan, K. and Caruthers , J. M. (1995). " Evolutionary Design of Molecules with Desired Properties Using Genetic Algorithms " Journal of Chemical Information and Computer Sciences, 35, 188-195

    1995

  56. [56]

    A Genetic Algorithm for the Automated Generation of Molecules within Constraints

    Glen, R. C. and Payne, A. W. R. (1995). "A Genetic Algorithm for the Automated Generation of Molecules within Constraints" Journal of Computer-Aided Molecular Design, 9, 181-202

    1995

  57. [57]

    Neural- Network Prediction of Glass-Transition Temperatures from Monomer Structure

    Joyce, S. J., Osguthorpe, D. J., Padgett, J. A. and Price, G. J. (1995). "Neural- Network Prediction of Glass-Transition Temperatures from Monomer Structure" Journal of the Chemical Society, Faraday Transactions, 91, 2491- 2496

    1995

  58. [58]

    Nagasaka, K. (1992). Chemical Design Automation News, 7, 26-30

    1992

  59. [59]

    Designing Molecules Possessing Desired Physical Property Values,

    Joback, K. G. and Stephanopoulos, G. (1989). "Designing Molecules Possessing Desired Physical Property Values," in Foundations of Computer-Aided process Design: Proceedings of the Third International Conference on Foundations of Computer-Aided Process Design, Snowmass Village, Colorado, July 10-14, 1989, pp. 363-387 (Elsevier, New York, NY)

    1989

  60. [60]

    Predicting physical and physical-chemical properties of molecular-based materials using computational neural networks

    Bibliographic reference: Gakh, A.A., Sumpter, B.G. and Noid, D.W., (1998). "Predicting physical and physical-chemical properties of molecular-based materials using computational neural networks" International Journal of Smart Engineering System Design, 1, 255-272. APPENDIX Systematic (e.g., IUPAC) chemical names of the organic compounds represent a valuab...

    1998