Ali Bahramisharif

Let’s have AI solve our quality challenges, it worked in the demo, right? AI requires very different thinking vs traditional mechatronic and software systems, especially in keeping its running costs in check. This talk will walk through some less intuitive pitfalls with AI for quality assurance and look into some future trends that further improve the feasibility of this long-running AI promise for higher quality at lower cost.

In diagnostics we want to find cause-effect relationships. One common way to find such causal interactions is through experimental design. During experiments we can observe whether a controlled change in one factor leads to a change in a different one. Alternatively, if we have some level of understanding of the underlying mechanisms, we can use reasoning based on this knowledge to determine or at least predict possible cause and effect. Experimentation can increase our level of understanding of underlying mechanisms, and as a result lead to more accurate diagnoses. However, such experiments can be too expensive, too impractical, or even impossible to perform. Besides, they rely on hypotheses that tend to be the result of the current mechanism-based knowledge. Instead, it is possible to use state-of-the art machine learning methods to let the data do the talking. When we have access to sufficient amounts of data, we can learn cause-effect relationships that are difficult to predict solely using mechanism-based reasoning. Using statistical models based on directed graphs we can indeed extract causal relationships from purely observational data based on measures for independence between variables [1-3]. In addition, we can use unsupervised machine learning methods such as clustering and latent variable modelling (e.g [4]) to discover subtypes within the population or discover biomarkers even without any domain or mechanism-based knowledge or even detect underlying causes that cannot be observed directly. In the context of the Repo4EU project, Machine2Learn will use the above techniques for use in AI-driven patient recruitment that could lead to the development of precision medicine [5] down the line.

A key component of working memory is the ability to remember multiple items simultaneously. To understand how the human brain maintains multiple items in memory, we examined direct brain recordings of neural oscillations from neurosurgical patients as they performed a working memory task. We analyzed the data to identify the neural representations of individual memory items by identifying recording sites with broadband gamma activity that varied according to the identity of the letter a subject viewed. Next, we tested a previously proposed model of working memory, which had hypothesized that the neural representations of individual memory items sequentially occurred at different phases of the theta/alpha cycle. Consistent with this model, the phase of the theta/alpha oscillation when stimulus-related gamma activity occurred during maintenance reflected the order of list presentation. These results suggest that working memory is organized by a cortical phase code coordinated by coupled theta/alpha and gamma oscillations and, more broadly, provide support for the serial representation of items in working memory. In our daily lives, we often have to keep multiple items active in our working memory—for instance, a series of digits in a phone number. How does the brain perform this task? In 1995, Lisman and Idiart proposed that multi-item working memory was supported by a mechanism in which individual memory items are repeatedly activated in sequence. The temporal coordination of the activation was supported by coupled oscillatory brain activity in the theta/alpha (7–13 Hz) and gamma range (>30 Hz). Using intracranial recordings from the human brain, we now report experimental support for this scheme. In particular, when subjects remembered a list of working memory items, it activated theta/alpha oscillations. Simultaneously, the brain showed item-specific activations of gamma activity that appeared at a theta/alpha phase corresponding to the item’s position in the sequence. The work demonstrates that coupled cortical oscillations are important for coordinating the maintenance of multi-item working memory.

We used intracranial recordings to study brain oscillations during a working memory task. To analyze sites involved in working memory, we focused on sites at which the elevation of the broadband gamma signal depended on which letter was presented. We tested a previously proposed model according to which different items are active at different phases of the theta cycle (in different gamma cycles within the theta cycle). Consistent with this model, the theta phase of letter-induced gamma elevation during maintenance reflected the order of letter presentation. These results suggest that working memory is organized by a theta-gamma code and provide strong support for the serial representation of items held in working memory.