Yoshitatsu Matsuda

In baseball, the pitch sequence is a significant factor in determining the outcome of a game. However, previous research has primarily focused on the direct relation between outcome and pitches rather than pitch sequences. In order to improve the game strategy in baseball, this study investigates the difference between the pitch sequences in outs and those in hits. For this purpose, pitching data of six pitchers of Nippon Professional Baseball (NPB) in 2022 was collected. Sequential pattern mining was then applied to extract common patterns in the pitch sequences. The results showed that the accuracy of the final phase pitches is crucial for achieving an out and that a pitch sequence incorporating fast-breaking is effective. Furthermore, the extracted patterns in hit suggested the appropriate inning to change pitchers.

近年、機械学習技術を活用した競馬の勝馬予測のAIが発達している。機械学習には様々な手法が存在しているが、既存研究では、勝馬予測のタスクに適合したある一つの手法を選択して活用することが多かった。本研究では、適切な機械学習手法がデータによって異なる可能性を考慮し、勝馬予測のタスクを勝馬のオッズによって分割し、各オッズにおける最適な機械学習手法を調査した。具体的には、オッズを利用して出走馬データを4分割し、scikit-learnを用いて様々な手法を適用した。その結果、オッズによって有効な機械学習手法が異なることが明らかになった。更に、様々な手法を組み合わせて予測回収率を高くする予測アルゴリズムを構築した。

In this paper, we extract various patterns of the spatio-temporal distribution from Foursquare data. For extracting patterns, we employ ICA (Independent Component Analysis), which is a widely-used method in signal processing and feature extraction. Because the Foursquare data consists of check-in's at some time points and locations, ICA is not directly applicable to the data. In order to smooth the data, we estimate a continuous spatio-temporal distribution by employing a diffusion-type formula. The experiments on the actual Foursquare dataset showed that the proposed method could extract some plausible and interesting spatio-temporal patterns. They also suggested that the estimation with the proposed method is promising when the sample size is limited.

Theoretical researches on the formation of brain topographic mapping have been carried out by computer simulation based on Hebb's law, which shows that self-organization based on Hebb's law can model the formation of local column structure as ocular dominance column or orientation column. However, the self-organization based on Hebb's law cannot model the formation of global topographic mapping such as retinotopic mapping in visual system or somatotopic mapping in somatosensory system. In this paper, we examine some well known facts of human cognitive and neuron system and introduce a principle for forming a global topographic mapping: "Global topographic mapping is formed by the law that draws cortex neurons spacially nearer, which correspond to input neurons that frequently change their values simultaneously. "We model this law, carry out computer simulation of the formation of visual system global topographic mapping (retinotopic mapping), and discuss this law's relevance and generality.