Kazuki Iwamoto

Title: Kazuki Iwamoto: Inventor in Anomaly Detection

Introduction

Kazuki Iwamoto is an inventor based in Osaka, Japan, associated with Fujitsu Corporation. His work includes patent inventions concerning the training of machine-learning systems for anomaly detection. He is credited with 2 patents.

Latest Patents

Iwamoto’s listed patents focus on training an image generator and a discriminator, which evaluates whether generated images appear real or fake.

One patent, “Training Apparatus, Training Method, and Non-Transitory Computer-Readable Recording Medium,” describes training the generator to produce image data that challenges the discriminator while also generating specified images that reduce its discrimination error at a controlled rate. The discriminator is then trained using both the generated images and the specified images to minimize that error.

A second patent with the same title describes a training process in which the generator seeks to maximize the discriminator’s error while the discriminator seeks to minimize it. During training, the apparatus stores the state of a partially trained generator that meets a preset condition. It then uses an image produced by that stored generator state to retrain the discriminator.

Career Highlights

Iwamoto’s listed patent work connects adversarial machine-learning training methods with anomaly detection. The inventions describe ways to manage the interaction between a generator and discriminator, including how training data and intermediate model states can be used.

Collaborations

Iwamoto’s listed coworkers include Takashi Katoh and Kento Uemura. The provided information does not specify their roles in the patents or their working relationship with Iwamoto.

Conclusion

Kazuki Iwamoto’s two listed patents address methods for training generator-discriminator systems used in anomaly detection, reflecting his documented innovation work at Fujitsu Corporation.