This article is written by Berend Gort & Bruce Yang, core team members of the Open-Source project AI4Finance. This project is an open-source community sharing AI tools for finance, and a part of the Columbia University in New York. GitHub link:WeFund announces that it has reached an agreement with FTM Launch in order to pool together resources and help each other pursue big milestones. WeFund has conducted several partnerships lately and the FTM Launch partnership is amongst the most promising.
FTM Launch is a fully decentralized and permissionless crowdfunding platform launching the next generation of disruptive applications on Fantom.
Launchpads built on blockchain ecosystems are an essential part of the survival of the blockchain. Launchpads help projects reach their potential and grow from an infancy stage to product/project launch.
WeFund and FTM Launch are both top-tier launch platforms that will provide a number of benefits to each other and to their respective blockchain ecosystems.
Some of the benefits include collaborating on fundraising and incubating projects together. The project sponsor can choose which launchpad they want to fundraise on. When it’s time to move forward both parties can do so based on the rules of the platform and the desires of the hosted project.
Both communities will drive up engagement, resulting in increased interest from potential investors and then converted into an investment opportunity.
Working together will create an opportunity for both FTM Launch and WeFund to share technologies and resources in order to put their best foot forward in the marketplace.
Finally, it will draw attention from the Fantom ecosystem to the Terra ecosystem and vice versa. As blockchains develop, it makes sense to shine light wherever possible.
Ika Co-Founder and CMO Comments, “We are so proud to be a partner with FTM Launch. Together we will merge our armies to help grow and support each other.”
FTM Launch Team, “A partnership and collaboration between two solid community-focussed launchpads will not only boost synergy in the crypto space but also build strong foundations for providing innovative opportunities to both investors and developers.”
We hope that in the future the FTM Launch partnership will nurture many projects and launch them to the moon and beyond.
First Decentralized and Permissionless crowdfunding Platform launching the next generation of Disruptive Applications on Fantom.
FTM Launch aims to provide a realistic, transparent, and fair launch platform for projects looking to launch on Fantom. We prioritize our community and the public by our fair distribution model and also reward long-term holders.
Website:
Social Media Links: Twitter: https://twitter.com/FTMlaunch
Telegram: https://t.me/FTMlaunchchat
Email: contact@ftmlaunch.com
WeFund is a cross-chain, decentralized incubator crowdfunding platform for the crypto-startup project industry and beyond implemented for a real-life use case. The vision of WeFund is to become the connector of the blockchain ecosystem that exists on the market.
To fulfill this vision, WeFund’s initial development stage would be in the Terra ecosystem and will develop to use additional ecosystems such as Solana, Etherium, Cardano, etc. in the near future.
Website:
Social Media Links:
Twitter https://twitter.com/WeFund_Official
Telegram — https://t.me/talkwithwefundhttps://t.me/wefundofficial
Email: info@wefund.app
github.com
Results
In my new article we present some results using this pipeline:
medium.com
Introduction
This tutorial aims to perform a feature importance analysis for a neural network that is used during deep reinforcement learning (DRL). Traditional machine learning techniques are deployed to decide on the best features.
So simply stated, this pipeline is built to:
The pipeline is a combination of ideas presented by Lopez de Prado, M. (2018) in Advances in financial machine learning. This tutorial follows a similar methodology but differs in the fact that a neural net is used to classify to do nothing, short or long. This article is subdivided into five main sections:
A functional Google Colab notebook is available on the AI4Finance foundation FinRL-Meta repo:
github.com
Ch2: Adapted Binance downloader
If you have no experience with candlestick (or klines) data we recommend this video:
First of all, unlimited data is required. This processor has no limitations on the amount of klines you want to download. It does, however, require data API keys from Binance.
Ok, back to the processor. The class has five important methods:
To use this class one simply creates an instance of it and then applies the run() method. Therefore, we will explain the run method in sequential order. and go from there. An instance of the BinanceProcessor() requires only the API information.
After that, you can call the run method on the instance. The run method requires:
The reader now understands the inputs of the run() method. After that, the download_data() is called.
Which loops through ticker_list and creates a separate data frame for every ticker and appends them together. It is recommended for this Google Colab notebook to restrict your analysis to one cryptocurrency. The rest of the code is not yet built for a multi-crypto analysis.
The loop through ticker_list applies get_binance_bars() for each ticker in ticker_list. This method uses the Binance client to obtain the klines, drops the ones we are not interested in, and changes the timestamp label to the index. Alright, now we go back up to the run() method. We obtained our data frame. We clean it and apply the method add_technical_indicators.
This is where TA-lib comes into action (technical analysis library). A list of supported indicators is present here. You can add any of the TA-lib features in the data processor, just as the Relative Strength Index (RSI), the Moving Average Convergence Divergence (MACD), Commodity Channel Index (CCI) and the Directional Movement Index (DX). Most of these indicators are calculated from the klines (or Open-High-Low-Close-Volume data, OHLCV). This part is left as an exercise for the reader.
The pandas data frame is now ready for further analysis. The method df_to_array() converts the data frame to Numpy arrays which can be a direct input to the deep reinforcement learning (DRL) algorithms present in the AI4Finance GitHub. This is not the goal of this article, and therefore our data is ready now as is.
Ch3: Triple barrier method (p.43, Lopez de Prado, M.)
Consider a features matrix X with a I rows. The features matrix exists of the OHLCV data and the technical indicators added. An example:
Let’s first work on your intuition. Stay with me here. Consider the BTC/USDT graph below. We draw a red box from 30–08–2017, the price is 4580. How we are drawing the box does not matter for now. The box has three important lines:
Imagine we have performed this labelling, now we have something to predict based on the features we have! 0: go short, 1: do nothing, 2: go long.
Now let’s find a reasonable way to get the box’s shape. The equivalent labelling rule in mathematical form is: