MonkeyLogic-Bonsai alignment /

[HNXJ]

[Hamed Nejat] Following up on the meeting;

Resolving ML-Bonsai alignment
Sequence GLO task in Bonsai
Laser management, recording, digital channels and intanDAQ-intanInterface

[jeromelecoq]

Ok. can you give us a little more details about your set-up?
how coupled is it to the stimulus code?

[HNXJ]

I'm a little bit confused, do you mean the hardware-software system or the generation and adjustment of stimulus itself?

Here is our setup summary, let me know if there was any other information needed regarding the experimental setup:

1. Behavioral control PC (runs on monkeylogic); stimulus is generated and saved as a .avi file (moving grating video). We just use monkeylogic to present the stimuli; tuning and adjustments (contrast, brightness, gamma, etc...) is performed prior to the experiment. The subject's display is a video-projector receiving task scene from the behavioral control PC.
2. Recording-INTAN PC receives event codes from monkeylogic and electrode signals from NI-DAQ hardware. After the session ends, we run the preprocessing pipeline and get the NWB file.

[jeromelecoq]

I understand. I am not familiar with MonkeyLogic obviously. How coupled is your system to it? We chose Bonsai because it is able to do all of the stimulus presentation, is fast and is very good at closed-loop control. This is important in this case, since we have a wheel controlling the gratings.

I assumed the recording PC just receives pulses on a DAQ card? Bonsai can also write to such a card.
It seems like you could use Bonsai but you are concerned this would be additional work? How do you anticipate to run the experiments otherwise?

[jeromelecoq]

Just a note that I discussed with Bruno Cruz today on this. He mentioned of some past effort to transition Monkey Logic to Bonsai so this has been looked into. I am not sure how much it involves. Bonsai is a general purpose tool so it should be a better Swiss knife later on. It all depends if your lab is interested in making the transition.

[bruno-f-cruz]

Heya,

Not sure how applicable my experience is here, but I’ll share in case it’s useful.

In our case, the use-case involved a consortium of labs and companies trying to acquire physiology data across different modalities (e.g., EEG, MRI), alongside various sensor data (e.g., temperature, perspiration, cameras) and multi-sensory stimuli (e.g., 2D rendering, virtual reality, auditory stimuli). While MonkeyLogic handled 2D stimulus presentation well for some labs, we found it difficult to scale across the other modalities—particularly virtual reality and closed-loop audio (it worked but hard to update continuously if I remember correctly). It’s not that it was impossible, but the domain-specific language (DSL), and MATLAB in general, lacked robust async support and mature SDKs, which made integration challenging.

Another major issue we faced was synchronization. Some groups used TTL pulses parsed offline, but with a growing number of sensors, this became hard to scale. Others relied on software timestamping, which introduced temporal jitter. LSL was also used in some setups, but not all hardware vendors supported it, which limited its usefulness.

Eventually, we moved most of our setup to Harp and Bonsai, which addressed many of these issues:

Bonsai is a complete language designed for working with asynchronous event streams. It already supports 2D and 3D rendering for visual stimuli (see this paper) and integrates well with cameras, audio, mouse/keyboard input, etc.

Harp allowed us to synchronize all our sensor streams at the point of acquisition, which was much more scalable.

During the project, we also had to integrate additional devices—like pupil tracking via Pupil Labs—and this was relatively painless. Since Bonsai is built with C#, we could use their existing SDKs and implement it as "yet another stream". (We also ended up combining it with online inference using YOLO to find what "What people were looking at").

To Jerome’s point: if your goal is to present passive 2D stimuli and you’ve already got that working in MonkeyLogic/Matlab + an extensive infrastructure for analysis, it may not be worth switching. But if you’re looking for a more extensible framework with better async and integration support, Bonsai is definitely worth considering.

Hope it helps and let me know if you have any questions!