3.1 IMMT AI assistant (character)

The Deep Learning (DL) AI character introduced by IMMT is an AI technology designed to assist users with their financial activities. IMMT has developed an investment algorithm with a high level of accuracy through Deep Learning (DL) using cryptocurrency transaction data from the period of 2014 to 2021. This investment algorithm is applied to the AI character, providing support to users in achieving their investment goals through advanced technology. By engaging in growth and strengthening activities, the performance and range of services offered by the AI algorithm can be expanded. With an AI character equipped with a broader range of services and increased accuracy, users can achieve meaningful results in reaching their investment goals. Through continuous strengthening and collection of AI characters, users will have the ability to create their own investment portfolios.

What technology will IMMT utilize to accelerate its AI Character models?

IMMT will leverage NVIDIA's deep learning technology 🚀 to accelerate its machine learning models. Harnessing NVIDIA's RAPIDS™ suite of open-source libraries 📚, IMMT will enable GPU acceleration using Python 🐍 in the development process of its machine learning models. Additionally, frameworks such as TensorFlow, Keras, or PyTorch will be utilized to configure the framework for GPUs, ensuring efficient deep learning model development. This approach will accelerate data preparation tasks and ETL (Extract, Transform, Load) functions using NVIDIA A100 GPUs and NVTabular.

Moreover, IMMT will utilize high-performance PyTorch and NVIDIA's HPC hardware for daily model updates and fine-tuning using real-time big data 📊. This will involve continuous cross-validation and validation of model accuracy through statistical p-values. The training of state-of-the-art NLP deep learning model BERT will be conducted using 10GB of raw text data from cryptocurrency-related news and 500GB of raw text data from cryptocurrency-related communities.

IMMT's self-trained CryptoBERT is expected to outperform vocabulary-based language models and existing models like SK Telecom's KoBERT. Additionally, real-time cryptocurrency-related news and online community data will be scraped from the web for sentiment analysis, with the results analyzed using CryptoBERT. This will enable IMMT to make informed decisions based on sentiment values derived from the analysis. 📈