Periodic Reporting for period 1 - ReWoMeN (Recall dynamics of working memory networks: Modeling, analysis, and applications)
Reporting period: 2021-06-01 to 2023-05-31
The importance of understanding human memory functioning is evident from its central role in our cognitive health as well as its role as the main inspiration behind developments in artificial intelligence, in particular artificial deep neural networks (DNN). Despite considerable progress in the recent years in the area of DNNs, robustness of these networks is an important open issue. The human memory is composed of several modules responsible for processing, learning, and recalling the received information. Among the memory modules is the working memory (WM) which is responsible for holding and processing information in a temporary fashion and in service of higher order cognitive tasks, e.g. decision making. The short-term nature of the WM makes it a great example for designing dynamic DNNs, which are useful in safety critical applications in uncertain environments. The aim of this proposal is to build a combined model-based and data-driven mathematical framework for understanding Recall dynamics of human Working Memory Networks (ReWoMeN) for realization of a robust DNN as well as contributing to the mechanistic understanding of the human WM.
The main project objective ‘A combined model-based and data-driven approach for the learning and recall dynamics of WM’ was proposed to be realized within three technical work packages, and two management/personal development work packages.
The main project objective ‘A combined model-based and data-driven approach for the learning and recall dynamics of WM’ was proposed to be realized within three technical work packages, and two management/personal development work packages.
Five work packages (WP) were proposed to realize the project objectives, as listed below:
WP1: A combined model-based and data-driven approach for learning and recall dynamics of WM
Two tasks were designed for this WP. For the first task ‘Choosing an appropriate model-based module for WM: Incorporating biological aspects’, two main topics were explored. First, an abstract, yet bio-plausible model of WM in collaboration with KTH was selected. This model gives a modular WM network model of populations of neurons coupled by a Hebbian learning rule. For the WM learning mechanism, two competing theories exist: persistent and sparse plasticity. To further understand these mechanisms, we have also studied the cellular level models with Calcium current coupling.
The second task of this WP, ‘Data selection and development of the data-driven based module for the learning dynamics’, has also been performed. Within two MSc projects, we have been deriving data-driven models of memory in associative learning, using fMRI data, as well as choice making, using EEG data.
WP2: Analysis: Explanation, Validation and Prediction
The first task, ‘Control-theory based analysis of the model in WP1’, was performed by control-theory based network synchronization analysis of the biological modular model of the WM, nonlinear stability analysis of the learning dynamics, as well as stochastic stability of a general and abstract model of neuronal networks.
The second task of this WP, ‘Verification of the model in WP1 as predictor of the output data as well as a new data set’, is currently ongoing.
WP3: Implementation as a dynamic DNN/RNN
The first task of this WP, ‘Preparation of a test simulator’, has been performed by simulating a large-scale spiking model of the WM. For data-driven components, Matlab codes have been developed, also the utility of Matlab toolbox for dynamic casual modeling, has been verified. The second task of the WP, ‘Robustness comparison of the WM model obtained in WP1 and WP2', is planned as an MSc project to be completed in March 2024. Theoretical analysis on robustness of deterministic and stochastic models have been performed.
WP4: Dissemination/ Communication/Exploitation
Three levels, The Netherlands, Europe, and international have been targeted. The performed activities are:
A. Invited talks:
1)Stochastic Stability of Discrete-Time Phase-Coupled Oscillators. In the mini-symposium ‘Effects of stochasticity and heterogeneity on networks' synchronization properties’ at the International Symposium of Society of Mathematical Biology, online talk, 2021.
2)Synchronization in neuronal networks: A control theory approach. ‘Donders Institute for Brain, Cognition and Behaviour’, NL, 2021.
B. Workshop and Invited session organization:
1)Co-organizing a 2-day workshop for the Dutch Institute for Systems and Control on ‘Data Learning & Dynamics at the Intersection of Neuroscience and Control’, NL, 2022.
2)Co-organizing an invited session on ‘Control theory in Neuroscience and Brain-inspired Engineering’ at the World congress of the International Federation of Automatic Control, Japan, 2023.
C. Partnership to the European Human Brain (HBP) Project, 2023
1)Participation in the HBP summit, Marseille, France, 2023.
D. Research communication assisted by TU Delft facilities.
1)Participated in the neuroscience annual event organized by TU Delft Bio Engineering Institute, Delft, The Netherlands, 2022-2023
2)Participated in Delft- Erasmus MC Rotterdam Neuromedicine flagship.
WP5: Training and management
Training has been implemented based on the proposal plans.
T1. At the main host: TU Delft, NL
1) meetings with Prof. De Schutter on several aspects of developing a successful academic career, including supervision of students, and research communications. Moreover, we have had meetings on the technical topics of robust control and optimization techniques for uncertain nonlinear systems, which has led to co-supervision of a PhD student on this topic under the main advice of Prof. De Schutter.
2) meetings with Prof. Reinders, helpful in learning about the interdisciplinary research platforms available at the TU Delft, in particular the Bio Engineering Institute and the flagship programs.
3) I have also participated in TU Delft Personal Development Plan and obtained the certificate. I have been participating in activities for women organized by Delft Women in Science.
4) I have earned experience in both designing and teaching interdisciplinary courses within a BSc course focusing on joining control theory and biology, taught jointly at TU Delft and Erasmus MC.
T2. At the secondment: Donders Institute, NL
My secondment has been implemented as performing research one day per week for the period of the project. I have had participated in research meetings of the lab of Prof. Fernandez, had meetings with him on various aspects of human memory systems, and its several modules. Currently, we have one joint MSc student and one joint PhD student.
T3. At the visiting Institute: KTH Royal Inst. of Technology, SE
The one-month visit, the labs of Prof. Lansner and Prof. Johansson, was performed, which allowed continuation of our collaborations on model-based analysis for the recall of working memory networks
WP1: A combined model-based and data-driven approach for learning and recall dynamics of WM
Two tasks were designed for this WP. For the first task ‘Choosing an appropriate model-based module for WM: Incorporating biological aspects’, two main topics were explored. First, an abstract, yet bio-plausible model of WM in collaboration with KTH was selected. This model gives a modular WM network model of populations of neurons coupled by a Hebbian learning rule. For the WM learning mechanism, two competing theories exist: persistent and sparse plasticity. To further understand these mechanisms, we have also studied the cellular level models with Calcium current coupling.
The second task of this WP, ‘Data selection and development of the data-driven based module for the learning dynamics’, has also been performed. Within two MSc projects, we have been deriving data-driven models of memory in associative learning, using fMRI data, as well as choice making, using EEG data.
WP2: Analysis: Explanation, Validation and Prediction
The first task, ‘Control-theory based analysis of the model in WP1’, was performed by control-theory based network synchronization analysis of the biological modular model of the WM, nonlinear stability analysis of the learning dynamics, as well as stochastic stability of a general and abstract model of neuronal networks.
The second task of this WP, ‘Verification of the model in WP1 as predictor of the output data as well as a new data set’, is currently ongoing.
WP3: Implementation as a dynamic DNN/RNN
The first task of this WP, ‘Preparation of a test simulator’, has been performed by simulating a large-scale spiking model of the WM. For data-driven components, Matlab codes have been developed, also the utility of Matlab toolbox for dynamic casual modeling, has been verified. The second task of the WP, ‘Robustness comparison of the WM model obtained in WP1 and WP2', is planned as an MSc project to be completed in March 2024. Theoretical analysis on robustness of deterministic and stochastic models have been performed.
WP4: Dissemination/ Communication/Exploitation
Three levels, The Netherlands, Europe, and international have been targeted. The performed activities are:
A. Invited talks:
1)Stochastic Stability of Discrete-Time Phase-Coupled Oscillators. In the mini-symposium ‘Effects of stochasticity and heterogeneity on networks' synchronization properties’ at the International Symposium of Society of Mathematical Biology, online talk, 2021.
2)Synchronization in neuronal networks: A control theory approach. ‘Donders Institute for Brain, Cognition and Behaviour’, NL, 2021.
B. Workshop and Invited session organization:
1)Co-organizing a 2-day workshop for the Dutch Institute for Systems and Control on ‘Data Learning & Dynamics at the Intersection of Neuroscience and Control’, NL, 2022.
2)Co-organizing an invited session on ‘Control theory in Neuroscience and Brain-inspired Engineering’ at the World congress of the International Federation of Automatic Control, Japan, 2023.
C. Partnership to the European Human Brain (HBP) Project, 2023
1)Participation in the HBP summit, Marseille, France, 2023.
D. Research communication assisted by TU Delft facilities.
1)Participated in the neuroscience annual event organized by TU Delft Bio Engineering Institute, Delft, The Netherlands, 2022-2023
2)Participated in Delft- Erasmus MC Rotterdam Neuromedicine flagship.
WP5: Training and management
Training has been implemented based on the proposal plans.
T1. At the main host: TU Delft, NL
1) meetings with Prof. De Schutter on several aspects of developing a successful academic career, including supervision of students, and research communications. Moreover, we have had meetings on the technical topics of robust control and optimization techniques for uncertain nonlinear systems, which has led to co-supervision of a PhD student on this topic under the main advice of Prof. De Schutter.
2) meetings with Prof. Reinders, helpful in learning about the interdisciplinary research platforms available at the TU Delft, in particular the Bio Engineering Institute and the flagship programs.
3) I have also participated in TU Delft Personal Development Plan and obtained the certificate. I have been participating in activities for women organized by Delft Women in Science.
4) I have earned experience in both designing and teaching interdisciplinary courses within a BSc course focusing on joining control theory and biology, taught jointly at TU Delft and Erasmus MC.
T2. At the secondment: Donders Institute, NL
My secondment has been implemented as performing research one day per week for the period of the project. I have had participated in research meetings of the lab of Prof. Fernandez, had meetings with him on various aspects of human memory systems, and its several modules. Currently, we have one joint MSc student and one joint PhD student.
T3. At the visiting Institute: KTH Royal Inst. of Technology, SE
The one-month visit, the labs of Prof. Lansner and Prof. Johansson, was performed, which allowed continuation of our collaborations on model-based analysis for the recall of working memory networks
Main obtained results are:
-revealing the effects of stochastic uncertainties in an abstract phase coupled model useful in memory research. Our results has elaborated on the similarities and differences between deterministic and noisy models;
-finding out that cluster synchronization is governing recall of WM models under some realistic assumptions;
-exploring the effects of Calcium currents on information holding in WM models.
Data-driven modeling has been also performed, and will be continued within two PhD projects which are ongoing.
This fellowship had been an exceptional impact on my career path. It has significantly allowed me to earn more visibility and improve my leadership skills as a female researcher conducting a
new interdisciplinary research.
-revealing the effects of stochastic uncertainties in an abstract phase coupled model useful in memory research. Our results has elaborated on the similarities and differences between deterministic and noisy models;
-finding out that cluster synchronization is governing recall of WM models under some realistic assumptions;
-exploring the effects of Calcium currents on information holding in WM models.
Data-driven modeling has been also performed, and will be continued within two PhD projects which are ongoing.
This fellowship had been an exceptional impact on my career path. It has significantly allowed me to earn more visibility and improve my leadership skills as a female researcher conducting a
new interdisciplinary research.