|
Mission Statement:
- Apply computational method
- to explore the theories of operations of the brain;
- to analyze the functions of the brain;
- to simulate the operating functions of the brain;
- to engineer the potentials of the brain.
Research Areas:
- Computational neuroscience
- Spike train analysis
- Multiple spike train analyses
- Stochastic statistical analyses
- Corrleation analyses
- Spike time and phase relationship analyses
- Neural encoding/decoding analyses
- Neural information processing
- Neural simulation
- Generic object-oriented realistic neural simulator
- Spike train simulator
- Neural Networks
- Learning rules -- Hebbian associative learning
- Time-delayed network architecture
- Neural network analyses
- Artifical neural net algorithms
- Autonomous robotic control
- Nagivational control
- Insect-like mini-robot for automonous path searching, path finding and path
planning
- Sensori-motor integration
- Decision making
- Experimental neuroscience
- Emotional components
- Theoretical emotional model in humans
- Experimental analysis of parameters regulating emotions in rats
- Limbic system
- Mesolimbic system
- Motor control
- cerebellum neurophysiology
- motor integration
- sensori-motor integration
- Innovative Teaching Methodology
- Excellence in higher education
- Science education
- Critical thinking techniques
- Creative thinking techniques
- Positive reinforcement techniques
- Learner-centered learning methodology
- Brain-compatible learning methodology
- Psychology and neuroscience of procedural learning and associative learning
- The art and science of education
- Education philosophy
- Assessment, evaluation and accountability
- Research Projects:
- Autonomous Robotic Control Project (sole P.I.)
- Project Overview:
- Autonomous robots are smart robots that are capable of making decisions to interact
with the environment in exploration and learn from the
experience.
- Rationale:
- Autonomous robots can mimick the basic functions of simple organisms, such as
insects, such that the basic principles of operation used
to interact with the environment can be explored.
- In designing an autonomous robot, we need to address the issues of how an organism,
such as an insect, solves the problems encountered in the
interaction with the environment.
- That is to say, the organism needs to "know" or "acquire knowledge" about the environment so that it can respond appropriately in the changing environment
in order to survive.
- Thus, an autonomous robot faces the same issues as an insect would have encountered.
- Advantages of using autonomous robots as a model:
- The reason why autonomous robot is a good model for investigating the evolution
of the nervous system is that we can explore various principles
of operation without being constrained by the inherited
evolutionary path.
- The evolutionary paths of insects are often stuck due to its heritary past, whereas
we can explore many more scenarios in neuroengineering.
- Real robots vs. Simulation:
- Real robots need to deal with the unexpected environments, such as wheels getting
stuck. These real-world issues pose a set of challenging problems for the organism
to solve.
- Simulations often assume that the robot will move without spinning the wheels,
but that is not a real-world problem. Insects need to solve the same problem when they are stuck while they encounter
obstacles.
- Research Objectives:
- Find the simplest/minimal set of solutions to solve the most/complex problems – this is called finding an "elegant solution."
- That is to ask: What is the simplest nervous system an animal can have in order
to solve the most complex problems encountered by a simple
organism, such as an insect.
- We choose not to study complex high-level vision, such as scene analysis, because
it is an overkill to the problem.
- Instead, we choose to study simple compound eye without even forming a retinal
image, yet insects can detect a great variety of visual
objects, and escape from predation without even having
a brain!
- Autonomous robots are similar in using the simplest set of algorithms to solve
the most complex problems without relying on a super-computer
to solve these problems.
- Basic components of autonomous robots:
- Sensory inputs (sensors for detecting light, touch, sound, etc.)
- Motor outputs (actuators for producing motions, such as wheeled robots or legged
robots)
- Controller (integrators of inputs to produce outputs, including memory modules
to store "experience")
- The Challenge:
- Design a robot without any a priori knowledge of the environment.
- That is, the robot has no knowledge of the external environment, and there is
no pre-programming done to allow the robot to know beforehand
what to expect or how to respond in a given solution.
- The robot has to learn from the exploratory experience to acquire knowledge about
how to respond appropriately in the environment in order
to survive.
- If we pre-program the solution of how to solve the problem, we would be actually
cheating, and defect the whole purpose of studying the
robot's behaviors.
- So the question becomes: What is the minimal set of assumptions and constraints
that we need in order for the robot to acquire knowledge
from the exploration in the environment such that the internal
circuitry will be self-organized to produce the final appropriate
response similar to those we found in the reflexes in insects.
- The Solution:
- Neural network algorithm is used to "learn" from experience. Exploration by the robot provides the exemplary set of data
to build an internal map of the environment.
- Neural nets are known to adapt to the environment, and learn from the environment
without any pre-programming. The task of the neural net
essentially construct the nonlinear mathematical mapping
functions from the input set (external environment) to
the output set (motor responses), with the internal synaptic
weights stored as the internal map.
- Neuroprosthetic Project (collaborative project)
- Project Overview:
- Neuroprosthetic is an example of neuroengineering resulted from the application
of the knowledge in decoding the brain signals to assist
paraplegic patients to move by volitional control signals
recorded from the brain.
- Rationale:
- The neural code used in voluntary movement execution is known in neuroscience
as "population vector code."
- Arm movement trajectory can be predicted based on the the population vector code
of a large ensemble of neurons recorded in the motor cortex.
- With the ability to decode the motor-command signal in the motor cortex, arm
movement direction and velocity can be specified based
on the neural code.
- Paralyzed patients will be able to move a robot arm based on the neural code
extracted from the neurons in the motor cortex.
- Research Objectives:
- Primates are used as the experimental model to test these neuroengineering techniques
to make voluntary arm movements by "thinking about it" without actual execution of arm movements by making vitrual 3-D arm movements
via virtual-reality display feedback fitted in front of
the animal. [Primate research is done by Dr. Andrew Schwartz
at the University of Pittsburgh. Neural spike train data
are provided by Dr. Schwartz.]
- The goal is to develop real-time decoding algorithms to extract information from
the implanted multi-electrode array in the motor cortex
to drive the robotic arm.
- Specific Goals:
- The neural code used in voluntary movement execution is known in neuroscience
as "population vector code."
- Arm movement trajectory can be predicted based on the the population vector code
of a large ensemble of neurons recorded in the motor cortex.
- Real-time implementation of the algorithm is essential. It is computationally
expensive to analyze a hundred channels of neural data
simultaneously. Efficient computational algorithms are
needed. Off-line analysis can be performed without any
constraints on the real-time performance. But if the neuroprosthesis
were to be practical, real-time responses are essential.
- Incomplete data set. Ideally, there should be enough data points in the neural
recordings to compute the predicted arm trajectory. In
reality, the firing rates of the neurons are very low in
the motor cortex. This puts constriants on the statistical
sampling size problem to extract enough information about
the intended trajectory to drive the robot arm. Ingenious
methods using adaptive predictive algorithms can overcome
these limitations.
- Adaptive algorithm is used to "learn" from prior movement trials to adjust for the unknown coefficients of the parameters
needed to generate the predictive movement trajectory.
- Predictive algorithm is used to "project" what the future neural code would be before the actual code are filled in from
the neural recordings. This allows us to overcome the incomplete
data sample problem.
- Adaptive-corrective algorithm allows us to correct the on-going errors from the
estimations, and adjust the predictive algorithm accordingly.
- Spike Train Analysis Project (sole P.I.)
- Project Overview:
- Definition:
- Spike train:
- Spike trains are the time-series electrical signals recorded from individual
neurons in the brain. They are essentially the action potentials (nerve impulses) generated by neurons. Spike trains are the signals generated by neurons used to communicate with one another.
- Neurons use a series of "pulse-coded" signals (i.e., action potentials) to represent the information encoded by a
neuron. The message encoded by a neuron is embedded by a time-series of spike train. Since all action potentials are essentially identifical to one another (i.e.,
same amplitude and same width), they represent the
digital signals used by neurons where the signal is
conveyed not by the amplitude of the signal, but by
the time-of-arrival of the signal. These pulse-coded digital signals are hybrid between the binary-code (used by
modern-day digital computers) and the time-code. It is the time-of-occurrence of the spike that encodes the parameter/content
of the signal.
- Mathematically, spike trains belong to a class of a process called "point process." A point process is a natural process that is characterized by the occurrence
of a point-event. A point event is an event that occurs as a point in time or a point in space. Mathematically, a point does not occupy any finite time or finite space, rather
it signifies the onset of an event in time or the limit
of an event in space. In other words, a point is infinitestimally small. Usually a point is used to signify the onset of event.
- Although action potentials do occupy finite time, the time of occurrence (or
the onset of an action potential) can be considered
as a "point." Thus, the analysis of the signal contents encoded by neurons can be treated
as a point process, which allows us to simplify the
complex problem into elegant mathematics.
- Spike Train Analysis:
- Mathematically, spike train analysis is essentially an analysis of the point
process encoded by the spike train.
- Physiologically, spike train analysis is used to deduce the functions of a neural
circuitry based on the spike train signals recorded
from neurons. In other words, it extracts the underlying functional circuitry of a neural
network based solely on the spike train signals.
- Reverse-Engineering:
- Reverse-engineering is a branch of engineering that extract the underlying principles
of operation of an unknown machine by analyzing the
(signal) contents of the machine. It is the principles for cracking the code, or figuring out what is not known
inside the box (or under the hood).
- In many ways, biology is essentially reverse-engineering the principles of biological
systems. The unknown machines are the biological organisms. We dissect them to figure out how it works, that is essentially opening up the
hood and figure out how a car works (if we didn't know
how cars work before).
- Black-Box Approach:
- Black-box approach is a classical reverse-engineering approach to deduce the
working principles of a "black-box" (an unknown box) based on the input and output signals applied to the black-box
without opening up the black-box. That is, we can deduce the principles of operation mathematically by analyzing
the signals going into the black-box and the signals
coming out of the black-box. Based on these input/output signal relationships, we can deduce what the black-box
is computing without the need to open up the black-box
or look into the content of the black-box.
- What is essential in the black-box approach is the input/output relationship. By knowing the input/output relationships, a mathematical formulation of the
black-box can be deduced. Although the specific details inside the black-box can be different from implementation
to implementation, the overall function is the same. For instance, the input/output function of a lamp is: given the input of some
electrical energy, the black-box will produce light-energy
as output. This lamp can be implemented as a incandescent lamp or a flourescent lamp or
a laser lamp, but the function is essentially the same. What is important is figuring out "what it does," not "how it does." There can be many different ways to accomplish the same function.
- In other words, a black-box approach is to deduce what is inside the black-box
just by analyzing the input-output signals of the black-box
without opening up to see what is inside. In biology, it is a nice approach to study what an organism does without dissecting
the organism. Dissecting an organism is not only an invasive technique, but also destroy the
normal operating function of the organism and perturbing
the system in such a way that prevents us from finding
out the true unperturbed functions.
- Therefore, the black-box approach to spike train analysis allows us to deduce
what is the principles of operation of the brain without
opening up the brain to see what is inside, just by
examining the neural signals generated by the neurons.
- Rationale:
- The basis behind spike train analysis is to deduce the principles of operation
of a neural network (black-box) by the spike train signals
recorded from these neurons. That is, given a set of spike train signals representing the input/output or
intermediate signal of a network, how can we deduce what
the network is computing?
- Mathematically, we are looking at the input/output mapping function of the system. Now, this mapping function is not a simple one-to-one mapping function, rather
it is a many-to-many mapping. Furthermore, the mapping function is non-unique, i.e., it is non-deterministic. In other words, the mapping function is a probablistic function. This is why given the same stimulus to an animal, the response is variable – not always the same each time. The stimulus-response function is variable because the underlying neural network
producing the mapping function is probablistic.
- Research Objectives:
- The objective is to analyze a set of spike train signals recorded from a large
number of (~100) neurons in the brain to deduce the function of the underlying neural circuitry.
- Specific Goals:
- The goal is to derive the probablistic input/ouput mapping function of the neural
network based on the set of spike train signals recorded
simultaneously from many neurons within a network.
- The Challenge:
- Find the probabilistic mapping functions such that they represent the internal processing functions for massively parallel operation.
- The Solution:
- Neural Signal Decoding Project (sole P.I.)
- Project Overview:
- This project addresses the question of how the brain encodes information, and
what information is encoded in the spike train signals.
- From the spike train analysis project above, we can deduce what the operation
principles of the neural network are (using the black-box
approach), the next step is to go to the next-level of
abstraction, and determine what these signals "mean." Basically, we ask the question of what is represented in the spike train code
by a set of neurons.
- What is important to address is that neurons don't function in isolation, i.e.,
no single neuron determines the overal function of the
system. The output of the central nervous system (CNS) is determined by the collective
output of individual neurons. This implies that the function of each neuron is determined by the interrelationships
among the neurons within the network. So our objective is to see how the brain encode (and represent) signals not
just by a single neurons.
- Thus, the signal decoding project is to examine what signals (information) are
represented/encoded in a network of neuron by its collective
propoerties, i.e., by the population dynamics.
- Rationale:
- The question of what signals are being encoded by neurons is actually more evasive
than just figuring out what it does. In engineering terms, signal content is relative, i.e., it depends on the "eyes of the beholder." The same signal can be interpreted differently by different observers. What is that? Because the signal content is dependent on both the encoder and the decoder. That is to say, a signal is only meaningful if the encoder and decoder agree
on what the signal representation is. If they agree on a common scheme, the signal can be decoded by the decoded;
if not, the signal is "meaningless." This is precisely what signal encryption is all about. When a signal is encoded in a form usign a scheme that is unknown to the decoder,
the signal is lost (in garbage). To the decoder, the signal is random. But to the encoder, the signal is not random, and is highly meaningful.
- So, the question becomes addressing the signal relevant to the encoder and decoder. In other words, a signal cannot be considered in isolation independent of the
encoder and decoder. What is interesting is that the same signal can be interpreted differently depending
on the decoder. In other words, the same signal can have multiple representations, and these
multiple interpretations can be extracted by different
decoders. This is what engineers called "multiplexing." A multiplexed signal actually convey multiple meanings – it is all up to the decoders to extract the multiple representations within
the same signal. This is a very economical and efficent way to represent signals.
- So the task for this project is to address the encoder and decoder functions
so that the meaning (and representation) of the signals
embedded in the spike trains can be extracted/decoded.
- Research Objectives:
- To decode the signals represented by the spike train based on the relationship
between the encoded information and the decoded output
in a network of neurons. The neural signals can be encoded by a set of neurons in a network, and different
aspects of the signals will then be decoded/extracted by
other neurons. This is essentially how the brain processes
information – by extracting the different components of the original signals that is meaningful
to the subsequent stages of analysis. For instance, the sensory signal of limb position can be decomposed/extracted
into displacement (=x), velocity (=dx/dt), acceleration
(=d2x/dt2) and jerk (=d3x/dt3) for subsequent analysis of the components of the movement.
- That is to say, we need to identify the "meaning" of the signal within context (i.e., relevant to the encoder and decoder). Most often, this means relevant to the physiological context or behavioral context
of the signal being processed.
- Specific Goals:
- Given a set of spike trains, identify what these signals represent in relation
to what is encoded and what is decoded. It is a question is signal representation as well as a question of signal re-representation.
- The Challenge:
- Find the representation of pulse-coded timing signal such that the information is embedded in the timing of these signals.
- The Solution:
- Evolutionary Neuroscience/Physiology Project (sole P.I.)
- Project Overview:
- How does the evolutionary process of trial-and-error combinatorics produce the
physiological functions that provide the feedback of an
appropriate behavior in the environment?
- In other words, how does the neural circuitry, such as reflex circuitry, form
from the behavioral appropriateness in the environment?
How does this "appropriateness" get inherited in the next generation in the evolutionary process?
- Rationale:
- Inherited traits are adaptations that are similar to adaptation in learning,
yet the mechanisms for establishing the adaptation is totally
different.
- Inheritance is based on a feedforward system
whereas learning is based on a feedback system. Our goal
is to find the similarities and differences between the
two mechanisms of adaptation.
- Research Objectives:
- We will compare and contrast the difference between the feedback and feedforward systems, and identify how the feedback system in physiological adaptation interacts with the feedforward system in inheritance.
- Specific Goals:
- Determine what are the crucial factors for switching from modifiable adaptive learned system to consolidation into fixed hardwire system
- The Challenge:
- What are the criteria for this switching to occur so that the temporary learned system acquired in an organism can pass onto the next generation as hardwire circuitry?
- That is, how does innate behavior capture by organisms?
- How does an organism know what to pass onto the next generation?
- The Solution:
- Emotions in Neuroscience Project (sole P.I.)
- Project Overview:
- Emotions are the internal parameters operating in the brain to assist the recognition
of the external environment for appropriate responses from
within and without.
- This project examines what are the essential parameters that characterize different
emotions, and addresses the roles of emotions in governing
behaviors of an animal or an autonomous robot.
- Define what emotions are and how they are used functionally in organisms, and derive them based on first principles without any a priori assumptions about what emotions are and what they are used for.
- Rationale:
- We want to derive the basis of emotions from first principles in the control of behavior in an organism. That is, we do not make any a prior assumptions about what emotions are in retrospection subjectively. Rather, we will derive the necessity of emotion in facilitating the realistic operation of an organism or a robot encountered in the real world.
- That is to say, imagine that we created an organism or a robot, what are the essential autonomous controlling principles that are needed to operate in the real world successfully. How would emotions allow an organism to survive better compared to an organism without any emotions?
- Research Objectives:
- We examine and identify what emotions are and what they are for in terms of
autonomous control of an organism.
- Specific Goals:
- We investigate and identify the parameters that govern each of the specific emotions, and determine how they are related to the operation of the behavior of the organism or an autonomous robot.
- The Challenge:
- We want to identify and verify whether the specific parameters of the emotional model we created would correspond to the emotions found in various animals.
- The Solution:
|