CTU: Event-driven sensing and Neuromorphic computing (Giulia D'Angelo)

JavaScript isn't enabled in your browser, so this file can't be opened. Enable and reload.

Tutorial responses for the Github: https://github.com/GiuliaDAngelo/CTU-EDNeuromorphic

Email *

Name Surname

Tutorial 1 (A) - Time Window for Event-Based Data Visualization: 1. Understanding Event Cameras: How do event cameras differ from traditional frame-based cameras in terms of capturing dynamic scenes, and what advantages do they provide for analyzing fast-moving objects or changes in the environment?

Tutorial 1 (A) - Time Window for Event-Based Data Visualization: 2.

Data Extraction and Processing: After loading the event data, we extract the x and y coordinates, timestamps, and polarities of the events. How might these different attributes be useful in understanding the behavior of moving objects within a scene, and what additional processing steps could enhance the analysis of this data?

Tutorial 1 (A) - Time Window for Event-Based Data Visualization: 3.

Interactive Visualization: The script visualizes events in real-time within specified time windows. How might adjusting the window period affect the visualization of events, and what strategies could be employed to ensure that significant events are not missed or overwhelmed by noise during the visualization process?

Tutorial 1 (B) - Sliding Window for Event-Based Data Visualization: 1. Sliding Window Technique: How does the implementation of a sliding time window impact the way we visualize event data? What considerations should be made regarding the duration of the sliding window to ensure meaningful representations of the scene are captured?

Tutorial 1 (C) - Fixed Event Count for Event-Based Data Visualization: 1.

Understanding Event Grouping: How does the fixed event count visualization approach differ from time-based visualization methods? What are the potential benefits and drawbacks of using a fixed number of events per visualization window in terms of capturing dynamic scenes?

Tutorial 1 (C) - Fixed Event Count for Event-Based Data Visualization: 2.

Real-Time Visualization Implications: What challenges might arise when visualizing events in fixed batches, particularly in a dynamic environment? How could the choice of batch size (e.g., 1000 events) affect the responsiveness and accuracy of the visualization?

Tutorial 2: Loading IBM DVS Gesture Dataset: 1. Try experimenting with different values for user_trial and time_window. How do these changes affect the visualization and interpretation of the data?

Tutorial 3: Play with Neurons: 1.

Modifying Input Current: How would you change the amplitude and duration of the input current pulses? Try adjusting the values in the I_ext array or the parameters used to define pulse_times and observe how it affects the neuron's firing behavior. Exploring Neuron Parameters:

Tutorial 3: Play with Neurons: 2.

Exploring Neuron Parameters: What happens if you modify the LIF parameters, such as Cm, gL, or VT? Experiment by increasing or decreasing these values and observe how the membrane potential and spiking behavior change in the animation. Can you identify the impact of each parameter on the neuron's dynamics? Adding More Pulses:

Tutorial 3: Play with Neurons: 3.

Adding More Pulses: Can you modify the code to introduce more input current pulses within the simulation time (e.g., add more pulses or change their timing)? Try to create a pattern of input that leads to a different firing rate of the neuron and describe what you observe in the membrane potential graph.

Tutorial 4: Play with Spiking Neural Networks (Brian) 1. Effect of Neuron Count: How does changing the number of neurons (variable N) in the simulation affect the overall spiking activity and the firing rate of the network? Try modifying the value of N and observe the differences in the plots. What insights can you gather about the relationship between the number of neurons and network dynamics?

Tutorial 4: Play with Spiking Neural Networks (Brian): 2.

Time Constant Variation: What happens to the spiking behavior of the neurons when you adjust the time constant (tau)? Experiment with different values (e.g., 5 ms, 20 ms, 50 ms) and analyze how the membrane potential's response to changes in v0 and incoming spikes is influenced. How does this impact the firing rate and the pattern of spikes?

Tutorial 4: Play with Spiking Neural Networks (Brian): 3.

Baseline Potential Exploration: The baseline potential (v0) is initialized based on the neuron's index. If you were to change the equation used to set G.v0 to something more random (e.g., G.v0 = 'rand()*v0_max'), how would this affect the spiking patterns? Investigate the new patterns generated and discuss what this randomness might represent in a biological context.

Tutorial 5: Play with SNN Visual Attention: 1. Why is a time window (window_period = 100 ms) used in the attention mechanism, and how does it affect the saliency computation?

Tutorial 5: Play with SNN Visual Attention: 2. What role does the run_attention function play in updating the saliency map, and how is the most salient location determined?

A copy of your responses will be emailed to the address you provided.

Submit

Clear form

Never submit passwords through Google Forms.

reCAPTCHA

Privacy Terms

This content is neither created nor endorsed by Google. - Terms of Service - Privacy Policy

Does this form look suspicious? Report

Forms