Questionable aspects of Oliver Sacks' (1985) report.
Explore the source record for details and available documents.
Autistic Disorder↗
Biomedical subjects
Makoto Yamaguchi
Publications and source records attributed to Makoto Yamaguchi.
Multireference configuration interaction calculation of the B(2)a"-X(2)a" transition of halogen- and methyl-substituted vinoxy radicals.
Ground and second excited electronic states of halogen and monomethyl substituted vinoxy radicals were studied by multireference configuration interaction (MRCI) calculation. Optimized geometries, rotational constants and vibrational frequencies of vinoxy and 1-fluorovinoxy showed good agreement with experimental values. Differences in calculated and observed B-X electronic transition energies were less than 0.1 eV and observed trends of blue shift upon increasing the number of substituted halogen atoms were reproduced by MRCI calculation. Observed fluorescence lifetimes of the vibrationless level in B state were in good agreement with calculated values. Rotational profiles of the 0-0 vibronic bands were successfully simulated with calculated rotational constants and transition dipole moments. Energy differences between planar and nonplanar optimized geometries in B state showed good correlation with the onset of fast nonradiative decay in B state, supporting the proposed mechanism of nonradiative decay via avoided crossings from B to A state which is followed by the decay to the ground state via conical intersections.
Journal Article↗
Orthogonality is not a panacea: backpropagation and "catastrophic interference".
Connectionist models with the backpropagation learning rule are said to exhibit catastrophic interference (or forgetting) with sequential training. Subsequent works showed that interference can be reduced by using orthogonal inputs. This study investigated, with a more rigorous assessment method, whether all orthogonal inputs lead to comparable extent of interference using three coding schemes. The results revealed large differences between the coding schemes. With larger networks, dense inputs led to severer interference compared with sparse inputs. With smaller networks, all the three schemes led to comparable extent of interference. Therefore, this study proved that not all the orthogonal inputs cause the same extent of interference, and that severity of interference depends on the interaction of the input coding scheme and the network size.
Humans↗
Complete solution of the Rescorla-Wagner model for relative validity.
In several papers, useful methods for deriving predictions of the Rescorla-Wagner (R-W) model were presented. This article supplements those previous efforts, which could not solve one important problem. Those cases are considered in which infinite solutions exist. In such cases, the resulting associative strength is determined by the initial associative strength. An important example is the experimental paradigm called relative validity. Surprisingly, an easy, pencil-and-paper solution can be derived. Based on the results of this paper, along with earlier ones, predictions can be derived for fairly broad conditions.
Animals↗
Comments on the misuse of terminology in savant research: it is not the sieve of Eratosthenes.
Explore the source record for details and available documents.
Autistic Disorder↗
Imaging of water distribution in thermally fractured granites by SPRITE.
Water distribution in thermally fractured granite samples was visualized by using SPRITE sequences. Networks of intergranular fractures were observed in the coarse-grained Inada granite after heating at 873 K or above. On the other hand, bright spots were observed in the fine-grained Okazaki granite, which may be due to pore water in feldspar grains. D2O diffusion into samples saturated with H2O was also observed by 2D-projected SPRITE imaging.
Diffusion↗
Reassessment of catastrophic interference.
Connectionist models using the back propagation learning rule are known to have a serious problem in that they exhibit catastrophic interference (or forgetting) with sequential training. After the model learns the first set of patterns, if the model is trained on another set of patterns, its performance on the first set dramatically deteriorates very rapidly. The present study reconsiders this issue with three sets of simulations. With orthogonal input vectors, interference can be reasonably mild. The number of hidden units was critical for the degree of interference, in contrast to suggestions of previous studies. Output coding scheme was found to be critical. The length of input lists also influenced the degree of interference. This study suggests that the interference problem has been overstated in the literature.
Computer Simulation↗
Size invariance does not hold for connectionist models: dangers of using a toy model.
Connectionist models with backpropagation learning rule are known to have a serious problem called catastrophic interference or forgetting, although there have been several reports showing that the interference can be relatively mild with orthogonal inputs. The present study investigated the extent of interference using orthogonal inputs with varying network sizes. One would naturally assume that results obtained from small networks could be extrapolated for larger networks. Unexpectedly, the use of small networks was shown to worsen performance. This result has important implications for interpreting some data in the literature and cautions against the use of a toy model.
Algorithms↗
Are multi-layer backpropagation networks catastrophically amnesic?
Connectionist models with a backpropagation learning rule are known to have a serious problem. Such models exhibit catastrophic interference (or forgetting) with sequential training. Having learned a set of patterns, if the model is trained on another set of patterns, its performance on the first set can dramatically deteriorate very rapidly. The present study reconsiders this issue with four simulations. The model learned arithmetic facts sequentially, but the interference was only modest with random (hence approximately orthogonal) inputs. Essentially the same result was obtained when the inputs are made less orthogonal by adding irrelevant elements. Reducing the number of hidden units did not have major effects. This study suggests that the interference problem has been somewhat overstated.
Algorithms↗
Fluctuation of predictions of a connectionist model can lead to qualitatively different results.
Several methods have been proposed to derive ideal predictions of connectionist models. In contrast, there have been few attempts to explore fluctuations of the predictions around the ideal value in probabilistic environment. This void must be filled. Thus, Monte-Carlo simulations with the Rescorla-Wagner model were conducted for an experimental paradigm called relative validity. Simulations indicated that fluctuation is too sizeable to neglect and that it can even lead to qualitatively different predictions. The final associative strength was influenced by the last several trials.
Cues↗
Assessment of weight-error derivative (WED) analysis.
Lewandowsky in 1995 made an important contribution to the analysis of connectionist models by proposing WED analysis. Here it is argued that original WED analysis is not sufficient. An extension which consists of displaying not only the angles but also the norms of vectors is suggested.
Humans↗