Dissertation Announcement for Babak Saravi — 10/04/2019 at 11:00 AM
September 25, 2019
Dear Faculty, Graduate and Undergraduate Students,
You are cordially invited to my Ph.D. dissertation defense.
Candidate: Babak Saravi
Degree: Ph.D., Electrical and Computer Engineering
Dissertation Title: Evaluating the Applicability of Deep Learning Techniques in Agricultural Systems Modeling
Date and time: Friday, October 04, 2019, 11:00 am to 1:00 pm
Venue: Simrall-228 (Conference Room)
Committee:
Dr. Masoud Karimi-Ghartemani
(Major Professor)
Dr. A. Pouyan Nejadhashemi
(Dissertation Advisor & Co-Major Prof.)
Dr. Ioana Banicescu
(Committee Member)
Dr. Sherif Abdelwahed
(Committee Member)
Dr. Bo Tang
(Committee Member)
Abstract: A rapidly expanding world population and extreme climate change have made food
production a crucial challenge in the twenty-first century. Therefore, improving crop production
through agricultural management could be an effective solution for this challenge.
However, due to the associated cost and time to perform field works, researchers widely
rely on agricultural system modeling to examine the impacts of different crop management
scenarios. However, due to the complexity of agricultural system modeling, their applications
in producing practical knowledge for producers are limited. Concurrently, deep
learning techniques have been recognized as a preferred method when dealing with large
datasets. In addition, deep learning techniques are easily adopted by non-experts due to its
ability to automatically learn and discover the classifications from raw data. Therefore, the
goal of this study is to examine the applicability of deep learning techniques in crop modeling
and utilize its features to create a simple alternative to traditional modeling systems.
In order to address this goal, this study was performed in three phases. In the first phase,
an agricultural systems model known as the Decision Support System for Agrotechnology
Transfer (DSSAT) was used to evaluate the impacts of irrigation application amount and
time on crop yield. A deep learning network was utilized and trained by incorporating
a large number of datasets produced by the DSSAT model. The inputs to the model are
comprised of precipitation and irrigation amounts and dates and the output is maize yield
at the end of the growing season. The results obtained from this phase of the study were
promising with an accuracy level of around 98%. However, the developed models were
suffered from overfitting. The second phase of the study examined the robustness of the
deep learning model under a wider range of environmental factors (e.g., different irrigation
and climatological conditions) while a simpler deep learning structure was desired compare
to the first study. To optimize the deep learning structure, three variable reduction
methods were used (Bayesian, Spearman, and Principal Component Analysis). The result
of this study showed that a simpler deep learning structure could be developed that has a
similar accuracy level as the original model (with 800 input variables) while the structural
size was reduced up to 80 times. In the third phase of the study, three techniques (L1/L2
regularization, and neurons dropout) were used to address the overfitting problem in some
deep learning models. The L2 regularization was identified as the most effective method
that increased model generalization and reduced overfitting. The overall results from this
study demonstrated the effectiveness of the proposed deep learning technique in replicating
the yield results from crop modeling under different climatological (e.g., precipitation,
temperature solar radiation) and management conditions (e.g., irrigation scheduling).
Thanks,
Babak Saravi