<h1>Research</h1>
<h2 id="current-project">Current Project</h2>
<p> My current research project is to investigate "how the visual contrast
modulates the learning responses in larval zebrafish".
To automate processes of both data acquisition and data analysis,
I developed two software named <a href="https://github.com/young24/BLITZ">BLITZ</a> (Behavioral Learning In The Zebrafish)
and <a href="https://github.com/young24/ABLITZER">ABLITZER</a> (the Analyzer of BLITZ Results). </p>
<p>As the main result, we found that when the visual conditioned stimuli were more contrasted to the background,
the larvae showed more learning responses in terms of the position and escaping turns.
Now the results have been published on BioXRiv: <a href="https://www.biorxiv.org/content/early/2018/08/28/401000">Visual intensity ratio modulates operant learning responses in larval zebrafish</a></p>
<p>I'm now with other members of WenLab working towards learning the neural substrate
of this visual modulation effect. We'll first try to get the related neural activities
in a head-fixed paradigm. Then monitor the entire learning process
in freely swimming larval zebrafish with our <a href="https://elifesciences.org/articles/28158">imaging technique</a>.</p>
<h2 id="research-interests-and-future-plan">Research Interests and Future Plan</h2>
<blockquote>
<p style="color:gray; text-shadow: 1px 1px 1px #666;">What I cannot create, I do not understand <br>
<em>- Richard Feynman</em></p>
</blockquote>
<h3 id="research-interest1">Decode functional neural circuits with the large-scale recording/imaging</h3>
<p>As a further step, I want to understand the underlying neural substrates of specific goal-directed behaviors.
To this end, it would be more informative and less biased to answer this question
with the large-scale recording/imaging tools (e.g., the whole-brain imaging in zebrafish,
the multi-electrodes array recording in mammal cortex). Furthermore, machine learning technique
is naturally suitable to learn from the large-scale dataset.
Eventually, I may develop some useful software for the large-scale dataset
as well as gain insights about the underlying neural mechanism.</p>
<h3 id="research-interest2">Comparative study of neural coding among different model organisms</h3>
<p>From C.elegans to human beings, the structures of the brain change dramatically.
However, they share some basic behaviors and functions. Is there any
neural coding conservative across all species, all vertebrates, or all mammals?
And if different, how different are they?</P>
<p>I want to answer this question by comparing the neural codings in
different organisms when perform the same task.
It would be better to incorporate the large-scale recording/imaging into the skeme.</P>
<h3 id="research-interest3">Develop software/websites for neuroscientists to share datasets, tools, and results in a more intelligent way</h3>
<p>As a programmer, I believe the spirit of open source is vatal for the scientific community
to develop in a more collaborative, reliable and efficient way.
First, the nature of science is to share the knowledge and tools to the entire human-beings.
Open source is the best practice for this purpose.
Second, there exist some platforms for open source in the artificial intelligence,
such as <a href="www.github.com">GitHub</a> and <a href="https://toolbox.google.com/datasetsearch">Google Dataset Search Engine(beta)</a>.
Also, neuroscientists/institutes in specific displines have developed dataset platforms,
such as <a href="http://portal.brain-map.org/">Brain Atlas</a> by Allen Institute,
<a href="https://engertlab.fas.harvard.edu/Z-Brain/">Z-brain</a> by Engert Lab,
and <a href="https://kristinbranson.github.io/BABAM/">BABAM</a> by Branson Lab.
However, they all require specific software/codes to analyze and add new dataset with many parameters to configure.
So there still lots of things to do to build a simplified and esay-to-use framework and platform.</p>
<p>I am developing the Perfect-Lab system based on GitHub to help labs
to manage all their projects on GitHub. In this way, this Perfect-Lab system
would help labs 1. manage the projects in a more intelligent way and
build up bedrocks for inter-labs' communications. 2. it would practically decrease
the difficulty to share and open their data and code. However, I have to admit,
this Perfect-Lab system is still at its infancy.</p>