PPPs: 👩‍🎤 NiNa Research - Week 🚀 #002

Plans 📆

• Start exploring solutions to collect and analyze LN data structure

• Define KPI

• Connect WebLN and WebBTC with some API's (ie: ChatGPT)

• Validate data points and KPIs

.

Progress ✅

At this stage, we are consolidating the team (Get in touch if you'd like to contribute) and kick the initial research needed to create a solid base to start building on. Below, an idea of approach to start a challenging but exciting endeavor. Keep in mind that this is a simplified overview, and the real-world implementation can be more complex and require additional steps based on the specific requirements and constraints of the project.

1. Define the Problem and Goals: Before start designing our Ai model, it's crucial to have a clear understanding of the goal, in this case to optimize and automate the routing performances of lightning nodes.

2. Collect and Preprocess the Data: Gather relevant data that will be needed to train the AI model. Data related to transaction history, network topology, fee rates, and any other variables that might influence routing decisions. Data preprocessing may involve cleaning the data, handling missing data, normalization, and feature engineering.

3. Select a Model Architecture: Choose an appropriate Ai model given that try to optimize and automate a decision-making process, reinforcement learning models might be suitable. Those are designed to make a sequence of decisions that lead to a final goal, often used in routing and scheduling problems. Deep reinforcement learning, which combines reinforcement learning with deep learning, could be particularly useful.

4. Train the Model: We'll be using the collected and preprocessed data to train the model. Depending on the chosen model, this might involve defining a reward function (for reinforcement learning models), determining the optimal model parameters, and tuning hyperparameters.

5. Evaluate the Model: Use a separate test dataset to evaluate the model's performance. Measure how well it's doing based on our predefined performance metrics.

6. Deploy the Model and Monitor its Performance: Once satisfied with the model's performance, deploy it in a test environment to manage a few Lightning Network nodes. Continually monitor its performance and adjust as necessary.

As for the tools, we have a wide range of choices. Here are some of them:

• Data Collection and Preprocessing: Python libraries like Pandas, NumPy, and Scikit-learn can be very useful.

• Model Building and Training: TensorFlow, PyTorch, and Keras are common choices for deep learning models. For reinforcement learning, we can consider OpenAI's Gym or Stable Baselines.

• Evaluation and Visualization: Matplotlib, Seaborn, and Ploty can be used for data visualization. For model evaluation, Scikit-learn has many utilities.

However, it's essential to have a deep understanding of both the Lightning Network and AI modeling to take on this project. We try to stay current with the latest research and developments in both fields, as this is a cutting-edge project that can benefit from these new techniques and insights.

.

⚡️ The Lightning Network Stats
In mid-2020 the Lightning Network had about 10,000 active nodes and 35,000 channels, with a network capacity of around 1,000 BTC. The network has likely grown since then, as AMBOSS repots:

• Nodes: 18,006

• Channels: 74,890

• Capacity: 5,029.86 BTC

1ML provides slightly different numbers, but on average it confirms lightning node's growth by 180%, the number of channels doubled and over 5x in network's capacity in the last three years. This increasing numbers are definitely a consequence of the grown interest in scaling solutions for Bitcoin.

Lightning Network and Data Points and KPIs
Here are some more detailed key performance indicators (KPIs) that are commonly used to assess the health and growth of the Lightning Network:

1. Number of Nodes: This refers to the number of active nodes on the Lightning Network. An increasing number of nodes can indicate growing network adoption.

2. Number of Channels: This refers to the number of open payment channels on the network. More channels can mean higher network capacity and better routing options.

3. Network Capacity: This is the total amount of Bitcoin (or other cryptocurrencies) that can be moved across the network. It's often measured in BTC.

4. Average Channel Capacity: This measures the average amount of Bitcoin that can be sent through a single channel.

5. Node Distribution: This is a measure of how nodes are distributed across the network, which can affect the network's resilience and routing options.

6. Channel Distribution: This refers to how the capacity of the network is distributed across its channels.

7. Route Availability: This measures the network's ability to successfully find a route for a payment, which can be influenced by factors like channel balance and network topology.

The Ai should be able to independently monitor the network and gather historical data to be able to understand how to operate specifically as the node will be operating within.

.

Problems ✋

• Where to start?

• Consolidating the team (Get in touch if you'd like to contribute)

• Define the performance metrics for the model

• Define how to optimize these performances

• Define what to automate and how

.

Links 🔗

• Subscribe via email or Nostr to follow more detailed updates