Hao Fang
Doctor of Philosophy
Department of Electrical and Computer Engineering
University of Washington
Seattle, WA, 98195
Contact Email: haof459 at uw dot edu or haof459 at gmail dot com
Office: ECE 432
Doctor of Philosophy
Department of Electrical and Computer Engineering
University of Washington
Seattle, WA, 98195
Contact Email: haof459 at uw dot edu or haof459 at gmail dot com
Office: ECE 432
Hao Fang is a Postdoctoral Scholar at the University of Washington. He received his Ph.D. degree under the supervision of Prof. Yuxiao Yang from the University of Central Florida, FL, USA, in 2023.
During his Ph.D., he leveraged his expertise in machine learning, stochastic dynamical systems, and neuroscience to build brain-computer interfaces (BCIs) for decoding and controlling the brain state.
His research interests include BCIs, machine learning, robotics, large language models, and computer vision (CV).
Waitting connections!
I am actively seeking a full-time research scientist focusing on Machine Learning. Please feel free to contact me if you are interested in my profile.
My PhD research focuses on brain-computer interfaces (BCIs) with wide applications to decode and treat neurological and neuropsychiatric Disorders.
Image Description
Neuromodulation systems that use closed-loop brain stimulation to control brain states
can provide new therapies for brain disorders. To date, closed-loop brain stimulation has largely
used linear time-invariant controllers. However, nonlinear time-varying brain network dynamics
and external disturbances can appear during real-time stimulation, collectively leading to real-time
model uncertainty. Real-time model uncertainty can degrade the performance or even cause
instability of time-invariant controllers. Three problems need to be resolved to enable accurate and
stable control under model uncertainty. First, an adaptive controller is needed to track the model
uncertainty. Second, the adaptive controller additionally needs to be robust to noise and
disturbances. Third, theoretical analyses of stability and robustness are needed as prerequisites for
stable operation of the controller in practical applications. We develop a robust adaptive
neuromodulation algorithm that solves the above three problems. First, we develop a state-space
brain network model that explicitly includes nonlinear terms of real-time model uncertainty and
design an adaptive controller to track and cancel the model uncertainty. Second, to improve the
robustness of the adaptive controller, we design two linear filters to increase steady-state control
accuracy and reduce sensitivity to high-frequency noise and disturbances. Third, we conduct
theoretical analyses to prove the stability of the neuromodulation algorithm and establish a
trade-off between stability and robustness, which we further use to optimize the algorithm design.
Finally, we validate the algorithm using comprehensive Monte Carlo simulations that span a broad
range of model nonlinearity, uncertainty, and complexity. The robust adaptive
neuromodulation algorithm accurately tracks various types of target brain state trajectories,
enables stable and robust control, and significantly outperforms state-of-the-art neuromodulation
algorithms. Our algorithm has implications for future designs of precise, stable, and
robust closed-loop brain stimulation systems to treat brain disorders and facilitate brain functions.
H. Fang and Y. Yang, "Designing and Validating of a Robust and Adaptive Neuromodulation Algorithm for Closed-Loop Control of Brain States," J. Neural Eng., 19 (3), 036018, 2022.
H. Fang, et al., "Design and Validation of a Robust and Adaptive Brain-Machine Interface Algorithm for Control of Brain States". Annual Meeting, Society for Neuroscience (SfN), 2021.
H. Fang and Y. Yang, "A Robust and Adaptive Control Algorithm for Closed-Loop Brain Stimulation," 43rd Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2021.
Image Description
Deep brain stimulation (DBS) is a promising therapy for treatment-resistant major depressive disorder (MDD). MDD involves the dysfunction of a brain network that can exhibit complex nonlinear neural dynamics in multiple
frequency bands. However, current open-loop and responsive DBS methods cannot track the complexmultiband neural dynamics inMDD, leading to imprecise
regulation of symptoms, variable treatment eects among patients, and high battery power consumption.Here, we develop a closed-loop brain-computer interface (BCI) system
of predictive neuromodulation for treating MDD. We first use a biophysical plausible ventral anterior cingulate cortex (vACC)-dorsolateral prefrontal cortex
(dlPFC) neural mass model of MDD to simulate nonlinear and multiband neural dynamics in response to DBS. We then use o ine system identification to
build a dynamic model that predicts the DBS eect on neural activity. We next use the online identified model to design an online BCI system of predictive neuromodulation. The online BCI system consists of a dynamic brain state
estimator and a model predictive controller. The brain state estimator estimates the MDD brain state from the history of neural activity and previously delivered
DBS patterns. The predictive controller takes the estimated MDD brain state as the feedback signal and optimally adjusts DBS to regulate the MDD neural dynamics
to therapeutic targets. We use the vACC-dlPFC neural mass model as a simulation testbed to test the BCI system and compare it with state-of-the-art open-loop
and responsive DBS treatments of MDD. We demonstrate that our dynamic model accurately predicts nonlinear and multiband neural activity. Consequently, the predictive neuromodulation
system accurately regulates the neural dynamics in MDD, resulting in significantly smaller control errors and lower DBS battery power consumption than open-loop
and responsive DBS. Our results have implications for developing future precisely tailored clinical closed-loop DBS treatments for MDD.
H. Fang and Y. Yang, "Predictive Neuromodulation of Cingulo-Frontal Neural Dynamics in Major Depressive Disorder using a Brain-Computer Interface System: A Simulation Study," Front. Comput. Neurosci., 2023.
H. Fang and Y. Yang, "Developing Robust and Adaptive Neuromodulation Algorithms for Treating Neurological and Neuropsychiatric Disorders," 20th Annual World Congress of Society for Brain Mapping and Therapeutics (SBMT), 2023.
Image Description
Deep brain stimulation (DBS) systems provide promising therapy for Parkinson's disease (PD). Current closed-loop DBS algorithms use on-off or linear-time invariant (LTI) controllers.
However, PD network activity is nonlinear, non-stationary, and noisy. Additionally, a single therapeutic target shows variable efficacy across different PD symptoms at different PD stages and even at different times of the day.
The above issues impair the performance of current DBS algorithms. Here, we propose a robust adaptive DBS algorithm for treating PD.
We first develop a nonlinear state-space PD network model to describe the nonlinear and non-stationary PD network activity and then construct an adaptive estimator to regulate them in real time.
We next add a robust controller to improve the robustness of the adaptive estimator by designing a low-pass filter to reduce the sensitivity of high-frequency noise and disturbances.
We test the proposed robust adaptive algorithm using nonlinear and non-stationary cortex-basal ganglia-thalamus network models in PD.
Our algorithm achieves precise and robust control of non-stationary PD network models across different therapeutic targets and uniformly outperforms current DBS algorithms.
Our results establish a foundation for validating the robust adaptive DBS algorithm in future in vivo animal and human experiments.
Our algorithm has implications for future designs of closed-loop DBS systems to treat neurological and neuropsychiatric disorders.
H. Fang, et al., "Robust adaptive deep brain stimulation control of in-silico non-stationary Parkinsonian neural oscillatory dynamics," J. Neural Eng., 2024.
Y. Yang and H. Fang, "Towards Model-Based, Robust, and Adaptive Neuromodulation for Closed-Loop Treatment of Brain Disorders," The 3rd International Workshop on Neural Engineering and Rehabilitation (NER), 2022.
Image Description
Emotion recognition from electroencephalogram (EEG) is key in developing affective brain-computer interfaces (BCIs).
The functional interaction of brain networks encodes emotion. Thus, using functional EEG network connectivity for emotion recognition is promising but remains difficult because of three challenges.
First, EEG network connectivity is typically quantified by symmetric positive definite matrices represented on Riemannian manifolds, making existing Euclidean recognition methods not directly applicable.
Second, the large EEG channel number leads to high-dimensional and noisy connectivity matrices that are difficult to classify. Third, aggregating EEG phase and spectral information on Riemannian manifolds remains elusive.
Here, we develop a new Riemannian discriminant analysis (RDA) method for emotion recognition from high-dimensional EEG covariance, coherence, and cross-power spectrum density matrices.
RDA first projects connectivity matrices onto a low-dimensional Riemannian manifold. Such low-dimensional projection extends to the complex Fourier domain to incorporate phase information.
RDA then aggregates spectral information via Bayesian-based frequency fusion. RDA finally uses a k-nearest neighbor algorithm on the low-dimensional Riemannian manifold for emotion recognition.
We demonstrate that RDA consistently achieves better recognition performance than state-of-the-art methods on three representative Emotion-EEG datasets (SEED, DEAP, and MAHNOB-HCI).
Our results hold promise to effectively use brain network connectivity for emotion recognition in affective BCIs.
H. Fang, et al., "Emotion Recognition from EEG Network Connectivity Using Low-Dimensional Discriminant Analysis on Riemannian Manifolds," 2023, IEEE Trans. Affect. Comput., 2024 (In revision).
H. Fang, et al., "Developing Affective Brain-Computer Interfaces on Low-Dimensional Riemannian Manifolds". Annual Meeting, Society for Neuroscience (SfN), 2023 (To appear, Washington. DC, Nov.).
Image Description
Developing an efficient and generalizable emotion recognition algorithm that can quickly adapt to an unknown subject is emerging in the field of affective computing.
However, the unknown subject usually has heterogeneous underlying neuron characteristics and variable emotional activities, causing the undesirable accuracy of current recognition algorithms.
Here, we propose a plug-and-play model-agnostic meta-learning framework (Pnp-MAML) to efficiently learn the generalizable emotion decoder at the population level, which can quickly adapt to the unknown subject.
Instead of building the end-to-end emotion decoder, our algorithm involves two steps: pretraining and fine-tuning.
First, we utilize the MAML framework to pre-train a meta-emotion decoder using all historical subject's emotional data.
The design of the meta-emotion decoder is model-free and compatible with mainstream emotion recognition architectures.
Then, we fine-tune the pre-trained meta-emotion decoder on the unknown subject using one-shot adaptation to realize Pnp emotion recognition.
Finally, We evaluate our proposed Pnp-MAML framework on public Emotion-EEG datasets (SEED, DEAP, DREAMER).
Our comprehensive experimental results show that our PnP-MAML framework archives state-of-the-art inter-subject emotion recognition accuracy and outperforms classical transfer learning algorithms across different emotion recognition architectures.
Our results hold promise to incorporate the proposed Pnp-MAML emotion recognition framework to effectively improve the inter-subject generalizability in designing future affective brain-computer interfaces (BCIs).
Image Description
Reactive trajectory optimization for robotics presents formidable challenges, demanding the rapid generation of purposeful robot motion in complex and swiftly changing dynamic environments.
While existing research predominantly addresses robotic motion planning with predefined objectives, emerging problems in robotic trajectory optimization frequently involve dynamically evolving objectives and stochastic motion dynamics.
However, effectively addressing such reactive trajectory optimization challenges for robot manipulators proves difficult due to inefficient, high-dimensional trajectory representations and a lack of consideration for time optimization.
"APEX: Ambidextrous Dual-Arm Robotic Manipulation Using Collision-Free Generative Diffusion Models". IEEE International Conference on Intelligent Robots and Systems (IROS), 2024.
"A Unified Control Framework using Diffusion Variational Autoencoder for Intercepting Flying Objects with Obstacle Avoidance". IEEE International Conference on Intelligent Robots and Systems (IROS), 2024.
"RETRO: Reactive Trajectory Optimization for Real-Time Robot Motion Planning in Dynamic Environments". IEEE International Conference on Robotics and Automation (ICRA), 2024.
"Nonlinear Optimal Control Synthesis Using Basis Functions: Algorithms and Examples". IEEE Conference on Control Technology and Applications, 2021.