Pinned Repositories
Classification-of-Consumer-Complaints-by-Llama3
This is a classification model for customer complaints in financial industries. It can classify six classes of products. Number of samples is 500. Accuracy is 88.6%(70B model), 78.0%(8B model).
Classification-of-Japanese-news-with-BERT
日本語ニュースタイトルの判別-BERT&TensorFlow
Feature-extraction-by-BERT
fine-tuning-with-synthetic-data-by-GPT4
This is fine tuning process with synthetic data created by GPT4. Fine tuned model can be used for classification for customer complaints in financial industries. It can classify six classes of products.
Japanese-News-Genre-Classification-Task-Solution-Using-Llama3-8B_20240520
Japanese News Genre Classification Task Solution Using Llama3-8B_20240520
PDF-QAsystem_20230510
By open AI LLM and LangChain, I create QA system to extract information of PDF.
Probability-of-default-model-by-Tensorflow-Keras
Probability-of-default-PD-model-by-Tensorflow-Keras
RAG-with-Claude3-Haiku
Information Extraction from Toyota Motor Corporation's Financial Results Briefing トヨタ自動車の決算短信からの情報抽出
Semantic-segmentation-by-Fully-Convolutional-DenseNet
Semantic segmentation by Fully Convolutional DenseNet56
Wagashi-Collection-Web-Generation-agent-by-GPT3.5
GPT3.5による「和菓子コレクションweb生成」エージェント
TOSHISTATS's Repositories
TOSHISTATS/Semantic-segmentation-by-Fully-Convolutional-DenseNet
Semantic segmentation by Fully Convolutional DenseNet56
TOSHISTATS/Classification-of-Consumer-Complaints-by-Llama3
This is a classification model for customer complaints in financial industries. It can classify six classes of products. Number of samples is 500. Accuracy is 88.6%(70B model), 78.0%(8B model).
TOSHISTATS/Classification-of-Japanese-news-with-BERT
日本語ニュースタイトルの判別-BERT&TensorFlow
TOSHISTATS/Feature-extraction-by-BERT
TOSHISTATS/fine-tuning-with-synthetic-data-by-GPT4
This is fine tuning process with synthetic data created by GPT4. Fine tuned model can be used for classification for customer complaints in financial industries. It can classify six classes of products.
TOSHISTATS/Probability-of-default-model-by-Tensorflow-Keras
Probability-of-default-PD-model-by-Tensorflow-Keras
TOSHISTATS/Japanese-News-Genre-Classification-Task-Solution-Using-Llama3-8B_20240520
Japanese News Genre Classification Task Solution Using Llama3-8B_20240520
TOSHISTATS/PDF-QAsystem_20230510
By open AI LLM and LangChain, I create QA system to extract information of PDF.
TOSHISTATS/RAG-with-Claude3-Haiku
Information Extraction from Toyota Motor Corporation's Financial Results Briefing トヨタ自動車の決算短信からの情報抽出
TOSHISTATS/Wagashi-Collection-Web-Generation-agent-by-GPT3.5
GPT3.5による「和菓子コレクションweb生成」エージェント
TOSHISTATS/AI-101
TOSHISTATS/Autoencoder-of-MNIST-by-Conv-FCN
Autoencoder of MNIST by Conv&FCN
TOSHISTATS/Car-classifier-by-transfer-learning-with-ResNet
TOSHISTATS/Car-classifier-with-tf-keras-EfficientNetB0
TOSHISTATS/Classification-of-Consumer-Complaints-in-multilingual-environment
This is a classification model for customer complaints in financial industries. It can classify five classes of products. This model is based on universal-sentence-encoder-multilingual-large, followed by a simple NN classifier. Inside the model, transfomer is used as critical components in it.
TOSHISTATS/Deep-Learning-on-TPU
Computer vision and Natural Language processing on TPU
TOSHISTATS/Evolution-strategy
TOSHISTATS/How-to-use-Google-colab
TOSHISTATS/Investment-model-by-reinforcement-learning
Investment model by reinforcement learning
TOSHISTATS/Japanese-consumer-complaints-classification
Classify 5000 samples of Japanese consumer complaints in banks. These are classified as one of 5 classes. Samples are increased by back translation.
TOSHISTATS/Japanese-News-Genre-Classification-Task-Solution-Using-Gemma2-2B_20240811
TOSHISTATS/livedoor-newstitle-5class-classification
livedoor-newsデータを用いた5クラスのニュース・タイトル判別問題を 日本語専用BERTと多言語対応BERTでそれぞれ90%超の精度を達成 / Achieve more than 90% accuracy with Japanese-BERT & multilingual-BERT for 5 class-newstitle-classification problem using livedoor-news data
TOSHISTATS/Malay-consumer-complaints-classification
Classify 5000 samples of Malay consumer complaints in banks. These are classified as one of 5 classes
TOSHISTATS/MNIST-with-tf.Keras-and-TPUs-on-colab
MNIST with tf.Keras and TPUs on colab
TOSHISTATS/Research-Paper-collection
This is my collection about Deep learning
TOSHISTATS/Stable-Diffusion_Car-classifier_20220919
car-classifier based on synthetic car-images by Stable Diffusion
TOSHISTATS/Transition-matrix_RCI_GPT4_LCEL
TOSHISTATS/Universal-sentence-encoder