corso|33574

Lectures: • Introduction to Computational Cognitive Science: Goals, frameworks (Marr's levels), and key modeling paradigms (Connectionist, Bayesian overview, Symbolic overview). • Computational foundations: Algorithms, probability, and computational thinking. • Connectionist Modeling: Principles, foundational models (Perceptron, MLPs, RNNs, CNNs), and learning algorithms (Backpropagation). • Network Applications in Cognition: Modeling Memory and simulating aspects relevant to neurodegeneration via network dynamics and ablation. • Computational Vision: Problems and models for Object Recognition and Categorization, including Neural Network approaches. • Computational Language: Problems, statistical models (N-grams), and vector-space/Neural Network approaches (Embeddings, RNNs). • Linking computational models to empirical data and cognitive neuroscience; current research frontiers. Labs: • Foundational R Skills: Environment setup, programming constructs, data handling, and visualization. • Implementing Neural Networks in R: Building, training, and using foundational models (Perceptron, MLPs) with libraries like keras. • Network Modeling of Memory & Neurodegeneration in R: Building simple memory networks and simulating network ablation/damage to explore functional impacts. • Using Experiment Builder Software: Learning and applying tools (OpenSesame/PsychoPy) to design cognitive experiments for Perception and Language research. • Computational Analysis & Modeling in R (Applications): Applying computational techniques to analyze data and explore models (categorization, CNNs, N-grams, Embeddings, RNNs) related to Perception and Language. • Integrating learned techniques for analysis or modeling projects.

Students with a three-year degree in Psychology or equivalent according to the international regulation can attend this class. Basic knowledge in computer programming, general psychology, statistics and psychophysiology are essential to fully follow the lessons, even if, we will try to accommodate differences in starting levels as much as possible.

Sun, R. (Ed.). (2023). The Cambridge Handbook of Computational Cognitive Sciences. Cambridge University Press. Chollet, F., & Allaire, J. J. (2018). Deep Learning with R. Manning Publications. Grolemund, G., & Wickham, H. (2017). R for Data Science. O'Reilly Media.

Frequency is strongly advised for all frontal lectures, whereas participation in laboratory activities is mandatory.

Aim of the evaluation Summative assessments will evaluate theoretical knowledge as well as critical competence about the topics covered during the entirety of the module. Such assessments, detailed below, will be distributed into three tests: a written exam, a programming assignment, and a computational project, with the following weighting. Written Exam (30%) Assesses: Theoretical understanding of core CCS concepts, principles of major modelling paradigms (Connectionist, Bayesian, Symbolic), conceptual understanding of specific model types and algorithms, ability to compare and contrast models, understanding the link between theory, models, and empirical findings. Programming Assignment (40%): Assesses: Practical R programming skills, ability to implement/use specific modelling techniques taught in the labs (Perceptrons, MLPS, Bayesian calculations, network simulations, N-grams, embeddings), data handling, visualisation, basic analysis of model outputs. Computational Project (30%): Assesses: Application of multiple learned skills to a specific problem, synthesis of theoretical knowledge and practical ability, independent work, in-depth analysis, critical evaluation of models/results, ability to link computational work to cognitive/neuroscience questions, written (and potentially oral) communication of scientific findings. For non-attending students, the exam session will inevitably be longer than a typical attending student's to allow sufficient time to cover the breadth and depth required. They will be examined in a single written exam that will cover the following sections: Section 1: Core Concepts and Paradigms (Approx. 20-25% of total marks) Question Types: Short answer, definitions, compare and contrast. Section 2: Model Principles and Mechanisms (Approx. 25-30% of total marks) Question Types: Explanatory, interpreting diagrams or pseudocode, problem-solving (conceptual). Section 3: Applications and Interpretation (Approx. 30-35% of total marks) Question Types: Scenario-based questions, interpreting results, and explaining application. Section 4: Critical Evaluation and Synthesis (Approx. 20-25% of total marks) Question Types: Essay-style, critical analysis, comparing approaches. Final evaluation The final evaluation is achieved on all the assessment components described above under the detailed weighting schema. Grade 28-30: Students use empirical and theoretical material excellently, offering structured arguments in their work. Students write comprehensive essays/exam questions, and their work shows strong evidence of critical thinking and extensive reading. Grade 24-27: Students show a sound theoretical and practical understanding of the problem and offer sufficient critical analysis. Grade 21-23: Students produce an acceptable work demonstrating an elementary understanding of the theoretical and practical concepts being discussed. However, the work falls short on its structure (e.g., organisation of findings) and/or logic of arguments, so it requires improvement. Grade 18-20: Students barely pass because the work covers the most elementary points touched upon. However, there are more serious concerns about the depth at which issues are understood, and so it shows a superficial commitment to the module content. Grade <18: Students produced insufficient work, clearly showing that the content was not understood. There is little or no critical analysis, the structure is very poor, and the logic of the arguments does not flow. The quality of the work highlights a poor commitment to studying the theories and applying them to the practices being taught.

Sun, R. (Ed.). (2023). The Cambridge Handbook of Computational Cognitive Sciences. Cambridge University Press. Chollet, F., & Allaire, J. J. (2018). Deep Learning with R. Manning Publications. Grolemund, G., & Wickham, H. (2017). R for Data Science. O'Reilly Media.

The course combines lectures and hands-on laboratory sessions. In lectures, students will explore core concepts, major computational modelling paradigms (Connectionist, Bayesian), and their application to key cognitive domains (Memory, Perception, and Language). They will learn about the theoretical underpinnings of different models and how they relate to empirical findings in cognitive neuroscience. In the labs, students will gain practical experience with the R programming language and experiment builder software (OpenSesame/PsychoPy). They will implement, simulate, analyse, and apply the computational models discussed in lectures, working collaboratively in groups to build their programming and modelling skills and design cognitive experiments.