Discover @harveybc Open Source projects

Harvey Bastidas (@harveybc)

harveybc

Stars
267
Global Rank 94,373 (Top 4 %)
Followers 26
Following 13
Registered almost 12 years ago
Most used languages

Python
53.3 %

JavaScript
33.3 %

C++
6.7 %

Cuda 6.7 %
Location 🇨🇴 Colombia
Country Total Rank 111
Country Ranking

Cuda 1

Python
16

C++
125

JavaScript
250

gym-fx

Forex trading simulator environment for OpenAI Gym, observations contain the order status, performance and timeseries loaded from a CSV file containing rates and indicators. Work In Progress

agent-multi

Automated trading agent for an OpenAI Gym enviroment with multiple simultaneous trading of symbols(currency pairs) using separate action and observation timeseries.

o-pow

REST API that allows extending existing evolutionary algorithms to a decentralized architecture to provide collaboration, scalability, event log and fault-tolerance in an optimization process. It also allows external clients to make remote evaluations of data using the most optimized model parameters found by the optimizers.

trading-signal

Generate an ideal trading singnal that can be used as training signal for an automated trading agent.

feature-eng

Implements classes for feature engineering including one for Singular Spectrum Analysis (SSA) decomposition, SSA prediction or an heuristic function of an input dataset that may be used as training signal.

preprocessor

Modular components for dataset preprocessing: a data-trimmer, a standarizer, a sliding-window generator and a feature selector, all optionally useable from command-line.

visualizer

Plugin-based multiuser Web platform for monitoring and visualizing variables with built in JSON-configured sqlite input and html output plugins. Has support for real time variable monitoring on multiple processes.

optimizer

Agent for decentralized multi-level optimization using Singularity and OpenAI Gym

gym-agent

Evaluator of a previously trained Deep Conv Net agent on an OpenAI gym enviroment.

membrane

Neural Network model with spiking neurons, plasticity, epigenetics and axon delay modelling

q-analyzer

Generate a csv file and a graphic with the correlations between an ideal training signal and each of the channels of a ssa decomposition of the close price and the sum of the channels with a variable number of channels. the objective is to decide the most convenient number of channels that to feed a trading agent in the gym-forex OpenAI environment.

data-logger

Remote API for configurable and secure generic data logging and visualization.

H5PHET

Implementation of Neuroevolution of Augmenting Topologies(NEAT) using a distributed architecture for training ANNs with Forex application in MQL4

SGDOC

Document management system with ISO 15489 field compatibility.

feature-extractor

Plug-in based feature extractor, includes modules for a configurable keras model trainer, evaluator and a training/evaluation visualizer with Web interface and serverless database.