CMOS and beyond CMOS
Discover why imec is the premier R&D center for advanced logic & memory devices. anced logic & memory devices.
Connected health solutions
Explore the technologies that will power tomorrow’s wearable, implantable, ingestible and non-contact devices.
Life sciences
See how imec brings the power of chip technology to the world of healthcare.
Sensor solutions for IoT
Dive into innovative solutions for sensor networks, high speed networks and sensor technologies.
Artificial intelligence
Explore the possibilities and technologies of AI.
More expertises
Discover all our expertises.
Be the first to reap the benefits of imec’s research by joining one of our programs or starting an exclusive bilateral collaboration.
Build on our expertise for the design, prototyping and low-volume manufacturing of your innovative nanotech components and products.
Use one of imec’s mature technologies for groundbreaking applications across a multitude of industries such as healthcare, agriculture and Industry 4.0.
Venturing and startups
Kick-start your business. Launch or expand your tech company by drawing on the funds and knowhow of imec’s ecosystem of tailored venturing support.
/Job opportunities/High Precision Neural Network Inference using Analog-In-Memory Computing

High Precision Neural Network Inference using Analog-In-Memory Computing

Research & development - Leuven | More than two weeks ago

Development of strategies for mapping state-of-the-art quantized networks (such as CNNs, etc.) onto energy efficient ternary precision AiMC hardware.
High Precision Neural Network Inference using Analog-In-Memory Computing

 Peter Vrancx, Debjyoti Bhattacharjee, Arindam Mallik

With the massive growth in size of neural networks, energy consumption of neural network inference has also risen. From a computation perspective, quantization techniques are used widely for lowering computational complexity. From the hardware perspective, analog-in-memory computing (AiMC) based solutions have grown popular in recent years. AiMC based solutions have extremely high energy efficiency of the order of 1000Tops/W.

AiMC based compute cells which store the neural network weights, natively support very low precisions (often binary or ternary). For mapping higher precision networks, each high precision weight needs to be mapped to multiple compute cells. Furthermore, the accumulation of the multiplication results using ADCs suffer from non-ideal behavior. These factors make it imperative to train the neural networks aware of the non-ideal hardware behavior to retain the original inference accuracy.

The goal of this project is to develop strategies for mapping state-of-the-art quantized networks (such as CNNs, etc.) onto ternary precision AiMC hardware and evaluate baseline accuracy of inference on such hardware. Specifically, the thesis would focus on the following aspects:

1. Develop mapping technique for transferring weights of existing quantized neural network onto AIMC hardware.
2. Report baseline accuracy after multi-bit weight transfer for different bit-widths.
3. Retrain the quantized network with noise information to recover accuracy.
4. Compare against the accuracy of other AIMC based results from literature.

Mandatory: Python, experience with PyTorch or any other deep neural network framework, fundamentals with number representation (2’s complement, 32-bit floating point, etc.).

Optional: Familiarity with quantization techniques for neural networks, multi-GPU training

Type of project: Internship, Thesis

Duration: 6 months

Required degree: Master of Engineering Technology, Master of Science, Master of Engineering Science

Required background: Computer Science, Electrotechnics/Electrical Engineering

Supervising scientist(s): For further information or for application, please contact: Peter Vrancx ( and Debjyoti Bhattacharjee ( and Arindam Mallik (

Imec allowance will be provided for students studying at a non-Belgian university.