| 研究課題/領域番号 |
16H06209
|
|---|
| 研究機関 | 沖縄科学技術大学院大学 |
|---|
研究代表者 |
ミケェエヴ アレクサンダー 沖縄科学技術大学院大学, 生態・進化学ユニット, 准教授 (90601162)
|
|---|
| 研究期間 (年度) |
2016-04-01 – 2018-03-31
|
| キーワード | animal behavior / social insects / machine learning |
|---|
| 研究実績の概要 |
We have been actively working on developing machine learning algorithms that can track the movement of individuals. Our firs task was to establish an experimental apiary, from which we draw the specimens necessary for the experiments. We then established the experimental setup, which allows us high-throughput acquisition of large-scale data and its direct transfer to OIST’s supercomputing cluster. We then used Amazon’s Mechanical Turk platform, which allows large numbers of human workers to perform analytical tasks, to acquire a massive training data set, which is helping us optimize model architecture and performance. This framework will also allow us to acquire similarly massive training data sets for other experiments.
|
| 現在までの達成度 (区分) |
現在までの達成度 (区分)
3: やや遅れている
理由
Our first round of crosses, conducted in FY2016 was unsuccessful, with all F1 hybrid queens ultimately dying before producing F2 offspring. While this is a delay, we were able to continue working on the data acquisition and analysis software, which is the most innovative and technically challenging part of the grant. We are hoping to move on to data acquisition for the main part of the project this fiscal year, though it may be delayed into FY2018.
|
| 今後の研究の推進方策 |
Having undergone additional training and practice, we are re-doing the crosses that have failed in FY2016. Our previous work has focused on the tracking of unmarked individuals, and we will expand it to follow marked individuals, which is necessary for the analysis of individual behavior. We are also preparing the first paper for submission.
|
